Systems Thinking Tools: A User's Guide

Transcript

1 TOOLBOX PR INT S ER IES RE SYSTEMS THINKINGTOOLS  AUser’sReferenceGuide  DANIELH.KIM BY

2 T H E R E P R I N T S E R I E S T O O L B O X Sy st em s Ar che t ype s I : Dia gno si ng Sys t emic Issues and Designing Hi gh-Levera ge Inter ventions Action Sy em s Ar che t ype s I I: st Effective ake T to Archetypes ms te Sys ng Usi Und II: I s ype t che Ar s em st Sy y Dela and or Behavi of Patterns ing nd ta ers o fer em s Th i n k in g T st ol s: A Us er ’s Re Sy ence Guide Essential The “T hi nk i n g” i n Sys tem s T h ink ing: Seven Skills Tools : A Us e r's Re fe re nce Guide Sy s T hi nk in g stem b ni el H. Kim Da y 1 9 94 , 2 0 00 by Pegas us Commun ic ations, I nc. © nu rst i nti ng Ja pr ary 19 94 . Fi t rig h ts rese rved. No par itted of t his book may be reproduced o r trans m All in any form or by any i me s, e lect roni c or m ec hanic al, an ncl uding phot ocopying and recording, or by any informat ion ri sto ra g e o r r etri eva l s yste m, wit hout w r. publishe the from ission perm n tte irin e g Kelli Acqu y or: e War dman O dit ’Re ill nc ty ctio n: Na du y Daugh er ro P I SBN 1- 8 83 82 3 -0 2-1 . NI PE SUS COM MU GA CAT ION S, IN C. re On e M oo dy St et 0 Wal US 9 453-533 2 A A M , ham t 781 0945 2- 7 -2 0 0 8 e n / Pho -3 9 8- 97 00 F 7 ax 8 1 - 8 94 -7 1 7 5 info cust er ser vi m egas us com. com / om @pega suscom.co [email protected] m w ww.p eg asu scom .co

3 2

4 T A B L E O F C O N T E N T S 3 In trod ucti on PAR T I: AN OV ER VIEW 5 6 by Mic ha el R. Good man Sy ste ms Thin ki ng as a La ng ua ge 8 Lev el s of Un der st an di ng: “F ir e-F ight ing” at Mul tipl e Leve ls A Pa le tt e of Sys tem s Th in king Tool s 10 PAR T II: DYN AMI C THI NKI NG TOOLS 13 Rein fo rc in g an d Ba lanci ng Lo ops: Bui ldi ng Bl ock s of Dyna mi c Syst ems 14 Bal an cin g Loo ps wi th Del ay s: Teet er- Totter ing on Sees aws 16 Gu id elin es fo r Dra wi ng Ca us al Loop Diag ram s 18 Sy ste ms Arch ety pes at a Gl ance 20 URAL 23 PAR T III : ST RUCT THI NKI NG TOOLS Fro m Cau sa l Lo ops to Gr aphi ca l Funct ions : Arti culat ing Chaos 24 ica l Funct ion s: “Se ei ng ” the Ful l St ory Graph 26 St ru ctur al Th in kin g: Th e Wor ld Accor ding to Accumula tors and Flows 28 Acc um ulat ors: Bat htu bs, Bat htubs, Ever ywher e 30 Acc um ulat io n Man ag eme nt: Av oidi ng the “Pa ck Rat ” Syndrome 32 Delay s: Accu mul at or s in Di sgui se 34 36 S- Sha ped Gr ow th an d the La w of Dim inis hing Ret urns TOOLS 39 PAR T IV: COM PUT ER-B ASED Mo de li ng fo r Wha t Pur pos e? by Ja yW.For rest er 40 Man ag eme nt Fl ig ht Si mul ator s: Fl ight Tr aini ng for Manager t I 42 s—Par Man ag eme nt Fl ig ht Si mul ator s: Fl ight Tr aini ng for Manager s—Par t II 44 Learn in g Lab or at or ie s: Pr acti ci ng Bet ween Perform ances 46 PAR T V: REFERENCE 49 GUIDE The Vo cab ul ary of Sy stem s Th inki ng : A Pocket Guide 50 The Do ’s an d Do n’ t’s of Sy stem s Thi nk ing on the Job by Mi ch ael R.Go odma n 52 Fu rthe r Re ad in g on the Ten Tool s of Sys tems Thinki ng 54 In dex to THESY ST EMS THIN KER 55 3

5 U C T I O N T R O D I N e are in the midst of a cha ng ing of an age— from the ag e of machine s to the sy s- W as a machine tha t co uld be tem s age. Our pas t was de fined by a view of the world ag e view of the unders too d by br ea kin g it into sma ller and smalle r pa rts. In the machine ea ch of the part s, we anding wor ld, the par ts are wha t is mo st im por tant —by underst of the la rg er who les. In the sy st em s view , it is the whole tha t bu ild up ou r under sta nding ha ve no me aning in and of the mselve s. Syst ems is mos t impor tan t—pa rts in is olation thin king embod ies the ide a that the inter re la tionship s among part s rela tive to a com mo n pu rpo se of a sys te m ar e what is imp or ta nt . Th ere is nothing more power ful tha n an idea whos e tim e ha s com e. But idea s wi thout prac tic al tools ca n take us only so fa r in making any me aning ful cha nge s th at will s the idea s tha t ca n help us see the have an impact on the wo rld. Sys tem s think ing provide wor ld in ne w ways , as we ll as the tools tha t ca n he lp us take ne w actions tha t are sys tem ic an d more eff ect ive . This bo okle t pro vides a bas ic introduct ion to the various too ls of sys - tem s thi nkin g th at have be en deve lo pe d and us ed ove r the last 50 yea rs . OWL NTS ACKN EDGME Mu ch of this wo rk has be en dev elop ed ov er the ye ars throug h the effo rts of many sys tem part of the Too lbox Re print Sys tems Think ing Tools :AUser’s Refer en ce Guid e, dyn am ici st s. Se ri es, was cr eat ed and co mp iled by Kellie Wardma n O’ Reilly. L I N K S A N D L O O P S O F T H E L A N G U A G E T H E L A N B A L A N C I N G L O O P G U A G E O F E X A M P L E O R S U L AT A C C U M Aca usa llink betw ee ntwo var iables , “clouds” represent the s + s where achange inXcaus es ach ang e Desired boundaries of what we want to Desired Level Level o Gap – + Gap inYinthesame dir ec tion, or wh er eX include in the diagram s + adds to Y. Actual Adjust- Actual Adjust- flow regulator Aca usa llink betw ee ntwo var iables , B B o Level ments Level ments accumulator where achange inXcaus es ach ang e _ y s + a inYintheopposi tedire ct ion ,orwhere l e D Xsu btra cts from Y. R population A“re inforci ng” fee dba ck loo ptha t If the re is agap be tween the des ired leve l amp lifieschange. births deaths and the act ua llev el ,adj us tmen ts ar emad e A“balan ci ng” feedbac kloop that seeks B to connector indicate unt ilthe ac tual equ als thedes ired level .The causal connection equi librium. pipe flow st ar tingvar iable is gr ey. 3 PEG ASU S COMMU NICA TIONS , INC. THI NK IN G TOOLS WWW.P EGAS USCOM. COM SYSTEMS

6 2

7 P A R T I R V I E W A N O V E yst em s thinki ng can be thou gh t of as a lan gu age. As a langu age, it is a spec ific S wa y of view ing the wor ld; it affect s th ou ght, and th ough t in turn affec ts ho w we (p. 6) offers insi ght in to ho w sys - loo k at the wor ld. “S yste ms Thi nki ng as a Language” tems th ink ing can be a usef ul frame work for comm unic at ing abo ut com pl ex is sues. of fee dback loops, we can learn to bet ter arti cul ate By “co nver si ng ” in the language the com pl ex in ter co nnect ions of circu lar causalit y in which we li ve . Learni ng the lan - our worl d on at leas t fo ur lev els— gua ge of sys tem s think ing requi re s us to understand stru cture, and share d vi sion . “Lev el s of ev ents , pa tter n of event s, systemic Unde rs tand ing : ‘Fir e-F ighti ng’ at Mul tipl e Level s” (p. 8) de sc ri bes thes e lev el s and the spec if ic act io n mo de associ ate d with each one . Thi s secti on cl os es wi th “A Palet te of System s Thinki ng Too ls ” (p. 10 ), whi ch out - line s ten tool s of sys te ms th inki ng. Seven of these tools are cov ered in th e subs eq uent sect io ns of this bo okl et. 5 WWW.PE GASUS COM.COM SYST EMS THI NKING TOOLS PEG ASU S COMMU NICA TIONS , INC.

8 L B O X O O T S Y S T E M S T H I N K I N G A S A L A N G U A G E E BY MI CHA N GOODMA R. L Dia gram s als o faci litat e lear ni ng. per cep tions of a probl em int o black-and - an gua ge has a sub tle, yet power ful L Stud ies have sh own that man y people lea rn whi te pictures tha t ca n rev eal subt le di ffer - ef fe ct on the way we view the wo rl d. best thro ugh rep res entat ional ima ges , such enc es in vi ewp oi nt . Englis h, like most other Wester n lan guage s, as pictu res or sto ri es. A sy stem s di agr am is Exa mp le : In one syst ems thi nki ng co ns tru ctio n, is lin ear —its basic sentence a po werf ul means of com muni ca ti on cour se, a tea m of ma nag er s was wor ki ng on no un- ve rb-n ou n, tr anslates into a wor ld - beca use it di sti lls the es sence of a prob le m an issue they ha d bee n wre st ling wi th for predi s- view of “x cau se s y.” This linearity in to a fo rmat that can be ea si ly remem - mont hs. One ma nag er wa s expl ai ni ng his pos es us to fo cu s on one- way relat io nsh ip s bered, yet is ri ch in im pl icat ions and posi ti on , tr aci ng thr oug h th e lo ops he ha d rathe r tha n circu lar or mutually causati ve in sig ht s. hi m. dr awn, when a te am member st opped ones , whe re x in flu ences y, and y in tur n “D oe s that mode l rep resent your thi nki ng many of the in fl uen ce s x. Un fort unately, ab out thi s probl em ?” he as ked. mo st vex ing prob lems confronting man - The present er hesi ta ted a bit , revi ewed ager s an d corpor at ions to day are cau sed by agr i s em st y s A is am d a hi s di ag ra m, and fi na lly answer ed, “Y es.” circu lar a we b of tigh tly int erconnected m f owe r f ul p e ans o The first ma n, ev ident ly rel ieved, rela tio nsh ips . To en hance our und erst an d- u m om c it e n eca io cat us ni b re sp onded, “After all of these mont hs, I of su ch pro bl em s, ication in g and co mmun is essen ce o t he d f t i ll s a your tho ught s on re al lyunde rsta nd fi na lly we nee d a lan gu age mor e nat ural ly su ite d thi s issu e. I di sa gr ee wi th it , but at lea st now to the tas k. f a o t in em l an b ro p c t ha t t ma or tha t we ar e cl ear on our di ff er ent vi ew - emb s b e e asi ly r em i er ed , y et EL EM EN TS OF THE poi nts, we can wor k together to cl ar if y the LANGUA GE nd in li ns cat p a im ri c h io prob lem.” Sys te ms thinkin g can be tho ug ht of as a exam ina ti on and inquir y. • Allows i s. ht i g ns for commu la nguage about co m- nicating Sys tem s di agra ms can be power fu l means plex itie s and in ter de pende nc ies. In part icu - fo r fost er ing a co ll ect ive under st andi ng of a la r, th e follow in g qualities make syst em s The spe cific set of “s yn - • Add s pr ecis ion. prob lem. Once indi vi dua ls have st at ed thei r th ink ing a use fu l framework fo r discu ssin g ta ct ica l” rul es th at go ve rn syst ems di agram s under st and ing of the pro bl em, th ey can co l- x issues: and ana ly zin g comple grea tl y re duce the am bi gui ti es and mi scom - lab or ate on add ress in g th e chal len ges it ende ncies .” Foc us es on “c lo sed interdep • mu ni cat ions th at can occur wh en we ta ckle poses. An d by fo cusi ng the di scussio n on is circu - The lan guag e of sy ste ms thinking co mp le x issue s. the dia gr ams , syst ems thi nki ng def uses la r rathe r th an lin ear. It fo cuses on clo sed : In drawi ng out the rel ati on - Examp le muc h of th e def ensi venes s tha t ca n ar ise in y, y in ter de pe nden cies , whe re x influences key asp ect s of a prob lem, shi ps be tween a hi gh-l evel deb at e. in fl uen ce s z, and z influe nces x. ca usa l li nks are no t onl y indi cat ed by ca rr yi ng on a syst ems : When Exa mp le Offe rs a “vi sual ” la nguage. • Man y of the arro ws, but ar e labe le d “s ” (sa me) or “o” di sc ussi on, di ffer ing op ini ons ar e no longer syste ms th in kin g tools —causal loo p dia - (o ppo si te) to sp eci fy how one va ri abl e vi ew ed as “huma n re sour ces’ vi ew of ou r gr am s, be havior ov er time diag rams, sy s- aff ects anothe r. Such label ing ma kes the produ ctivi ty prob le m” or “m ar ket ing’ s te ms arche typ es, an d str uctural na ture of the rel ati onsh ip more pr eci se, desc ript ion of decr easi ng cust omer sat isf ac- di agra ms —h av e a st rong visual com pon ent. ensu rin g onl y one poss ibl e int erp ret at ion. ti on ,” but si mp ly di ffer ent st ructu ra l repr e- The y help clarify issues by su m- complex • Forc es an “exp lic itne ss ” of menta l mod - sent at ions of th e syste m. Thi s sh ifts the the key el e- min g up, con cisely and clearly, el s. The syste ms th ink ing langua ge trans - fo cus of the di scus si on from wheth er d. me nts involve la tes “wa r sto ri es” and indi vi dua l huma n re sour ces or ma rket ing is rig ht, to 6 PEGASUS , INC . 781.398.9 700 COMMUNICATIONS SY STEM S THIN KING TO OLS

9 cons tru cting a diag ram that bes t cap tur es st ruct ed by linking to ge ther ke y varia ble s sk il ls by us ing the lang uag e as of ten as po s- the be ha vior of the system. the ca us al rela tions hips an d indicating si ble . The sa me hol ds tru e fo r syst ems • Embodies a wo rl dv iew that lo oks at be twee n the m. By str inging tog et her sev - thi nki ng. and that re cog - who le s, rathe r th an parts, eral loo ps, we can cr eate a “paragrap h” When sit tin g in a me eti ng , see if you nize s the import ance of understan din g ho w th at tell s a co her ent sto ry ab out a pa rticu - can in for m you r under st andi ng of a pr ob - the diffe re nt seg men ts of a system are inte r- lar pr obl em unde r study. lem by ap plyi ng a syst ems perspect ive. conn ect ed. An in he rent ass ump tio n of the If there we re a Be rl it z gui de to syst ems Look for key wo rds that sugg est linear sys tem s thin king wo rldv iew is that prob - thi nki ng , ar chetyp es such as “F ix es That thi nki ng is occur ring— sta te ment s such as hat we lem s ar e in te rn ally generated—t Fa il” or “Shi fti ng the Bur den” woul d be “w e ne ed mor e of the sa me” or “t hat sol u- ofte n cre ate ou r ow n “worst nightm ares.” list ed as “com mo nly us ed phr ases.” They tion wor ked for us the last time thi s hap - : At syst ems th inking Exa mple co urs es pro vide a ready-m n ade libr ar y of commo pened, why not us e it agai n? ” Yo u ca n al so at Inno va tion As sociates , par ticip ants play a st ruct ure s and behav io rs tha t ca n apply to creat e low-k ey practice sess io ns by wor ki ng bo ar d game kn ow n as the Bee r Game, ma ny sit uati on s. Me mo riz in g them ca n with a sma ll tea m of col leag ues to di agr am of re tail er, whe re th ey assu me the position help you reco gniz e a busi ness sit ua tion or a pa rt icul ar prob lem or issue. who le sale r, dist ribu tor, or produce r. Each pro blem that is ex hibi ti ng co mmon sym p- BEC OM ING FLUEN T play er trie s to achie ve a car eful balanc e tom s of a syst emi c break down. We say so meone is flue nt wh en they begin be tw ee n carry in g too much inv en tory or to thin k in a par ti cul ar languag e and no go wro ng , things ed . When be ing ba cklogg long er have to tra ns lat e. But fl uen cy means their sup plie r (“I kep t ma ny people blame i nh er ent assu m pt ion of t he A n mo re tha n just an ab ilit y to commu ni ca te in ord er ing mo re, bu t he didn’t re sp ond” ) or l or w g n nki i h t ie m te s y s is w s dv a la ngu age ; it me ans und erst andi ng the su r- the buye rs (“fickle co nsu mers —on e day t l em s are int e rn al ly h a b pr o t roundi ng cul tur e of the langu age— the they’ re bu yin g it by the tr ucklo ad, the nex t wor ldvi ew . As wi th any forei gn lan gua ge, day the y won ’t even touch the stuf f”). In t ea r c n our e r at ed— we of e g e n te ma st er ing syst ems thi nki ng wi ll allow us to reality , neithe r the buyer s nor the sup pl ie rs ht g .” ni st “wor wn s o e ar m ful ly enga ge in and ab sor b the wor ldv iew are re sp on sible for the wide fluc tuat io ns in that perv ades it. By le ar ni ng the lang uag e inve nto ry —th ey ar e a natur al con sequ ence of sy st ems thi nki ng , we wi ll hopef ull y of the str uct ure of th e system in whic h th e cha nge no t onl y the way we discu ss co m- play ers ar e fu nct ion ing . Of cour se, the ke y to becom ing mor e plex issues , but th e way we thi nk about The syst ems thinking wo rld vi ew di s- prof ici ent in any la nguage is to pr actice — them as wel l. • pels th e “us ve rsu s the m” menta lit y by an d practic e ofte n. Whe n re ading a news - ex pand ing th e boun dary of our th in ki ng . paper art icle, for exampl e, tr y to Mic hael Good man is an as socia te dir ector of thi nki ng , ork of systems Wi thin th e framew “t ran sl ate” it into a sys te ms pe rs pe ctive : In nova tio n As so ci at es, In c. (Ca mbri dg e, MA) .The “us ” and “t hem” are part of the sam e sy s- • Tak e eve nts repor te d in the news pa per his 20 ye ars mat eria l in this art ic le wa s draw n from of experien ce in the fie ld , as well as from busi ness tem an d thu s resp ons ible for both the pro b- an d try to tra ce out an unde rly ing pa tt er n co urs es dev elo pe d by Inn ov ation Assoc iates . lem s and their so lu tions. th at is at wor k. • Ch eck whe the r the stor y fit s one of the LE AR NIN G TH E sys te ms arche typ es , or whethe r it is pe r- E LA NGUAG haps a com bina tion of sever al ar che ty pe s. Lea rni ng syst ems think ing can be lik ened • Tr y to ske tch out a ca usa l loop or tw o to ma st er ing a fo rei gn langua ge. In school, th at captur es the str uctur e producing tha t by fir st we stud ie d a fo rei gn la nguage patt ern . mem orizing the essential vocabulary Do n’t ex pe ct to be flu ent in sys tem s wo rds an d verb conj uga tions. Then we thi nki ng ri ght away. Re membe r, af ter your beg an pu tting toget her the pieces into sim - fir st few Lati n cl ass es, you sti ll coul dn’t pl e se nte nces. In the la nguage of sy stems For that ma tt er , you prob - The Odyss ey. read th in kin g, sy stem s di agrams such as caus al ab ly knew onl y a few key phr ase s and con - loop s ca n be tho ught of as sentences vo ca bu la ry wor ds, bu t you imp roved your 7 PEGAS US COMMUN THI NK IN G TOOLS SYSTEMS M WWW.P EGAS USCOM.CO , INC. ICATIONS

10 L B O X O O T L E V E I N G : E R S T A N D L S O F U N D ” A T M U L T I P L E H T I N G “ F I R E - F I G L E V E L S t’ s anoth er bu sy nigh t in the THE SYS TE MSTH IN KE R, V3N7 ). This fo ur di st inc t le ve ls —ev ent s, pa tterns of I hospi tal eme rge nc y room . Sever al ry hist ory , is co ns is tent wit h our evolutiona eve nt s, sy st emic st ruct ure, and sha red vis io n car ac cident victim s have been rush ed int o whic h wa s gea red towa rd resp onding to (see “L eve ls of Unde rst anding ”) . Even ts are surge ry , one li ttl e boy is having a br ok en those thing s tha t pos ed an imm ediat e dan - the thi ngs we enc ount er on a da y- to-da y arm set, a dru g over dos e victi m is bei ng ger to our well- being . ba sis : a ma chine br eak s, it ra ins, we eat din - tr eate d, and nu me rou s other people fill the ate response. Event s re qui re an immedi ne r, see a mo vi e, or writ e a report . Pa tterns chai rs in the wa iting room . Eac h nigh t is If a ho use is bur ni ng, we re act by ta king of eve nt s are the ac cumul at ed mem ories of dif fer ent, and yet eac h one is also th e sam e. ac tion to pu t out the fire. Put ti ng out th e eve nt s—w he n st rung to get her in a serie s The doc to rs and nur ses must ac t fast to fir e is app rop ri ate , but if it is the only acti on pa tterns. ove r time , th ey re veal recurring tr eat the most se rious ly inj ured, wh ile the th at is eve r ta ken, it is inadequ at e fr om a Sy st emic st ruc tu re can be view ed as “e ve nt other s wa it th ei r tu rn. Li ke an assem bl y Alt hou gh it sy st em ic pers pe ct ive . Why? ge ne rat or s” be caus e th ey are re spons ibl e fo r li ne of defe ctive parts , pati en ts are diag - sol ved the imm edi ate pro bl em (the bur ning pr oduc ing the ev ent s. Sim ila rly, sha re d Each inj ury is nosed, treate d, and released. house ), it ha s done no thi ng to al ter th e fu n- vis ion can be vie we d as “sy st emic st ruc ture a cri sis that de mand s imm ed iate at te nt ion. dam ent al st ruct ur e tha t ca used that even t ge ne rat or s” be caus e th ey are the guiding So what ’s wron g with this pict ure? (e.g ., ina dequa te bui ldi ng cod es, lack of fire fo rc e be hin d the crea tion or chang e of al l Af te r all, isn ’t th is what eme rgency ro om s det ec to rs, fi re pre vent io n ed uca ti on) . The kin ds of struc tu res. are mea nt to do? Th e ans we r dep en ds on “L evel s of Un ders ta ndi ng” di ag ra m an d and We li ve in an even t- orie nt ed world, at which we ar e the lev el of un derst anding k ca n hel p us go beyo nd typica l fra mewor our la ngua ge is ro ot ed at the le vel of eve nt s. loo ki ng at the sit uat ion. even t-or ient at ion resp onses and begi n to At wo rk, we enc ount er a serie s of ev ent s, look for hi gher lever age acti on s. whic h oft en appe ar in the form of prob lem s LE VEL S OF tha t we mus t “sol ve .” Our solut ion s, how - UNDE RSTANDI NG FR OM FIR E- FI GHT ING TO eve r, may be sho rt -liv ed, and the sym pt om s FI RE PRE VE NT IO N There are multip le level s from whic h we can eve nt uall y ret ur n as se em ing ly ne w From a the world. can vi ew an d un de rstand At th e even t leve l, if a hou se is on fire , all pr oblem s (se e “Us in g ‘Fix es Th at Fa il’ to sy st emi c per sp ective , we are inte re ste d in we ca n do is rea ct as quick ly as pos sib le to Tre adm ill, ” Ge t of f the Pr obl em- Solving put the fire ou t. The on ly mode of action tha t is ap prop ri ate an d ava ilab le is to be O F L E V E L S S T A N D I N G U N D E R react ive. If we rea cted to fires only at the ev en ts level , we wou ld pu t all of our ene rg y Mode Understanding of Levels Action Typical Questions Time Orientation into figh ting fire s—a nd we wou ld pr ob a- bl y have a lot mor e fire sta tions than we do What are the stated or Future Generative unstated visions that generate Shared Vision toda y. the structures? If we loo k at the probl em of fire s at the are the mental or What pat tern of even ts leve l, we ca n beg in to Creative Systemic Structure organizational structures that create the patterns? an tici pat e wh er e the y ar e mor e lik ely to kinds of trends or What occu r. We ma y not ice th at ce rtain neig h- Patterns of Events patterns of events seem to be Adaptive bo rhood s seem to have mor e fir es tha n oth - recurring? er s. We are abl e to be adaptive by loca tin g What is the fastest way to mor e fire sta tion s in those areas , and Present Events react to this event NOW? Reactive 8 PEGASUS , IN C. 781.398.9700 COMMUNICATIONS SY STEM S THI NKIN G TOO LS

11 ma y try an ad ap tive respon se and incr ea se on pas t st affi ng the m accor din gly (based and can al so chal len ge ou r sha red vis ion . ER ca paci ty in those reg ion s. If diver sion patt er ns of usage ). Si nce th e stat ion s are a To be most eff ec ti ve, the full ran ge of levels rat es are hi gh, we ca n al so find out whe re lot clo se r, we can be more eff ec ti ve at mus t be cons id ere d sim ul tan eou sl y. The the am bu lan ces are bei ng dive rt ed from to th em soon er. putt ing out fi re s by getting dang er lie s in op erati ng at an y on e level to and try to en hance ca pa ci ty the re . Yet whil e be in g adaptive all ow s us to be the excl usion of the oth er s. At the system ic str uct ur e level , we can mor e effect ive fire -fi ghters, it does not hi ng Our ab ility to infl uence th e future does begi n to exp lor e why ce rt ai n regi on s have to re duc e the actual oc currence of fir es. inc rease , howe ver, as we move fr om the an inc rea sed need for ERs. We may dis - At the sy st emic st ruct ure le vel we beg in level of eve nt s to sh ared vi si on . Does this cove r, for ex ample, that 40 per cent of th e : “Are smok e de tec tors ask ing questions act ion s ca n onl y mea n th at high -l everage ER adm issi ons ar e chi ldr en who ar e poi - bein g used? What kind s of bui ldin g ma teri - be fou nd at high er leve ls? No, beca use son ed, be cause a lar ge per cent age of the als are less flammab le? What safe ty fe atu res lever age is a re lat ive con cept, not an ab so - com muni ty ca nnot rea d Engl ish and all redu ce fa tal ities ?” Actions tak en at this le vel lut e. Wh en someon e is bl eed ing, the hi gh- warni ng la bels are print ed in En gl ish . By ca n actua ll y red uce the numb er an d se ve rity es t leve rage act ion at that momen t is to stop redr awi ng the bounda ry of the ER issue to of fires. Estab lis hin g fi re codes with requ ire - inc lude the commu ni ty, we ca n take act ions men ts such as automatic sp ri nkl er systems, ing de ep er ro p e h T cess o f g ai n that wi ll cha nge the inf low of pati ent s. fir ep roo f mate rials , fi re wal ls, and fire alarm Elec tr ical ut ilities hav e bee n doi ng thi s for syst ems sa ve s live s by pre ven ti ng or co ntai n- n i g n s n i d a stan r e d n ot l i r a e u som e ti me . In stea d of bui ldi ng anot her s taken at th is lev el are cre - in g fi res. Action st e . Ou r u n der n an d in o a g o f exp ens iv e power plant to sup ply more ative beca use th ey he lp cre ate a differe nt ca f e ed e ev l ne o at n o i t ua t i l s n pow er , they ar e wor ki ng wi th cust omer s to fut ure. nf an k c a b ss ne e d i r o r m ou r aw a reduc e the dema nd for power . Syst emic str uc tu re inclu de s not onl y At a com muni ty-wi de level , we may the or gani zational struc tures an d ph ysi cal no a . t a level er h t want to exp lor e th e que st io n, “Wha t is ou r bui lding s, bu t pe opl e’ s ment al mod el s an d sha re d vi si on of the rol e our hea lt hcar e sys - habi ts as we ll . Whe re do th e syst em ic str uc - tem plays in ou r lives?” Per ha ps the tur es come fr om? They are usu al ly a ref lec - the ble ed in g– –an y oth er action wou ld be resour ces that are goi ng int o ERs coul d be tion of a share d vis ion of what is val ue d or ina pprop ri ate . As we mo ve up the leve ls bet ter ut ilized el se whe re , such as comm u- desi red. In the ca se of fire- fighti ng, the from eve nts to sh ared vi si on , the focu s ni ty educ at ion and prevent ion pro gra ms. es (e .g., fi re codes) are bor n new structur moves from bei ng prese nt -or ie nt ed to The hi ghest lev er ag e li es in clar ifyi ng the out of a shar ed value of the im por tanc e of bei ng fu tu re -ori ente d. Con seq ue nt ly, th e qua lit y of lif e we env is io n fo r our sel ves, an d ed wi th the prot ecting hu man li ves , combin act ion s we tak e at the hi gh er level s have then usi ng tha t as a guid e for cr ea tin g the desi re to live an d work in safe bu ild ings. more impact on futu re ou tcom es, not pr e- sy st em ic st ruct ur es that wi ll hel p us achi ev e At the level of sh ar ed vis ion, ou r acti on s se nt eve nt s. that vi si on. can be gener ati ve, br ing ing somet hi ng in to AT THE BACK The basi c me ssag e of the “Level s of bei ng that di d not exis t bef ore. We begi n EMERGENCY RO OM Unders ta ndi ng” di agr am is the impo rtance the role of aski ng ques tions li ke “What’s Wh il e th e emer gen cy room (ER) of fe rs a of rec og ni zi ng the level at which you are the fir e-fi ghtin g fun ction in thi s com mu- gra phic examp le of a sit ua ti on in wh ic h op er at ing, and ev al uat ing wh et her or no t it ni ty ? What are th e trade- offs we are wi ll - peop le mus t be focu se d on th e pr esen t, it pro vi des the hi ghes t leve ra ge for th at si tua - ing to make as a com mun ity betwe en th e also re veals th e li mi tat ions of the even ts- tion. Eac h leve l off er s di ffer ent oppor tun i- amount to fi re- fi ght - of re so ur ces devoted ori en te d resp on se. ER trea tm en t of fer s ties fo r hi gh-l ever ag e act ion, but th ey al so ed to oth er th ings?” ing compar maxim al le verage to af fect the pr esen t si tu- ha ve thei r limi ts. Th e cha ll en ge is to ch oose It is impor tant to re mem ber th at th e ation wi th each pat ien t, bu t it pr ovi de s very at e the ap propr iate resp onse for the immedi proc ess of gaini ng deeper un derst and ing is lit tl e le verage for chan gin g th e future. If si tuat ion and find ways to alt er th e fu tur e not a linear one . Our understandi ng of a we go up on e leve l an d exam ine ER use oc cur renc e of tho se ev ent s. si tuation at one leve l can feed back an d • from a patt ern s of eve nt s level , we ma y dis - inform our awar en es s at another level . co ve r that cer tai n areas of a ci ty seem to Events and patte rn s of even ts, for exam pl e, have hig her eme rgen cy room need s. We can cause us to ch ange sys tem ic stru ct ur es 9 PEGAS US COMMUN THI NK IN G TOOLS SYSTEMS M WWW.P EGAS USCOM.CO , INC. ICATIONS

12 O O L B O X T T T E A P A L E O F S Y S T E M S T H I N K I N G T O O L S her e is a full array of sys tem s th ink - tur n ha ve sub-f actor s. Ma ny layer s of su b They ser ve as a sta rting poi nt fro m whi ch T ing tools that you ca n thin k of in the one ca n bui ld a cle ar er arti cul at ion of a nest in g, ho wev er, may be a si gn tha t one of sam e wa y as a paint er views color s—m any the sub -facto rs sh oul d be turne d int o a busi ness st or y or issu e. Spe cif ic ar chetypes shad es can be create d out of thre e prim ary inc lude: “Dr ift ing Goa ls, ” “S hi ft ing the ma jor factor . colo rs , bu t having a full rang e of read y- Burde n,” “L imi ts to Success, ” “Success to THINKING DY NAMIC ma de co lors make s painting much eas ier . the Suc cessf ul ,” “Fi xe s That Fai l, ” TOO LS The re are at least 10 dis tinct types of “T rage dy of the Commo ns,” “G rowt h and ar e Beh av iorOve rTi me (BO T) Di ag ra ms syst em s thin kin g too ls (an abbr ev iated sum - Under inv est ment ,” and “Es cala ti on” (see mo re than si mp le li ne pro ject ions— the y ma ry dia gram ap pears on the facin g pag e). “S yst ems Ar chet ypes at a Gl ance, ” p. 20). ca pt ure the dynam ic re lat ionshi ps amo ng Th ey fall un der four bro ad catego rie s: STR UCT URAL THINK ING va ri abl es. For exam pl e, sa y we we re trying thin kin g br ains tor ming tools, dynamic TO OL S to proj ect th e rel ati on shi p be twee n sa les, to ols, and co m- tool s, st ruct ur al thinking Grap hi cal Func ti on Di agr ams,St ruct ur e- in ven to ry , and pro ducti on. If sal es jum p 20 each of the pute r- bas ed tools. Although and Beh av iorPai rs, Po licySt ruct ure pe rcen t, pro ducti on cann ot jump insta nt a- tool s is de sign ed to stand alone, th ey al so Diag rams can be vi ewe d as the bui lding neo usl y to the new sal es numbe r. In ad di - bui ld up on on e an ot her and can be use d in blo cks for co mp uter mo del s. Gr aphi cal ti on , inv ento ry mu st dro p bel ow its com bina tion to ach ieve deep er in sig ht s in to Func tions are us eful for cl ari fyi ng no nli n- pr evi ous le vel whi le pr od ucti on ca tche s up beh avio r. dynamic ea r rel at ionshi ps betwee n var ia bl es. They wit h sa les . By sk etch ing out th e beha vi or of TOOLS MING BRA INSTOR are part icul arly he lp ful fo r qu ant ifyi ng the di ff erent vari abl es on the same gra ph, we ef fec ts of variabl es tha t are di ffi cul t to mea - ca n gain a more exp lici t under st andi ng of is based on The Do uble- Q(QQ) Diagram sur e, su ch as mo ra le or time pr essur e. ho w the se vari abl es int err el ate. what is co mmon ly kno wn as a fish bon e or Str uc ture -Beh avi or Pairs lin k a speci fic provide a Ca usal Loop Di agram s(C LDs) caus e- and -effect diag ram. The Qs stand fo r st ru ctur e wi th its cor res pondin g beh avi or . usef ul way to rep res ent dyna mi c inter rel a- qua lit ativ e and qu an titative , and the tech - the Pol icy Structur e Diagra ms represent ti on shi ps . CLD s mak e exp lici t one ’s unde r- nique is desig ne d to help par tic ipan ts beg in proc esses tha t dr ive pol icie s. In a sen se, sta nd in g of a syste m’s st ruct ure, pr ov ide a to see the whole syst em. Dur ing a stru c- when we us e the se to ol s we ar e mov ing vi sua l rep res entati on to hel p com muni cat e ture d bra inst ormin g se ssion with the QQ fro m pai nt in g on ca nvas to scu lpt in g thr ee- tha t und ers tandi ng, and cap tu re com pl ex diagra m, bot h side s of an issue remai n sy stem s in a su cci nct fo rm. CL Ds ca n be di men si onal figur es. equa lly visible an d prop erly balanced, co mb in ed wit h BOTs to for m st ruct ur e- avo idin g a “t op-h eavy” per spectiv e. Th e CO MP UT ER -B AS ED TO OLS be ha vi or pai rs, whi ch prov ide a rich frame - diagra m also provid es a visual map of the This cl ass of tool s, inc lu ding comput er work fo r de scri bin g co mpl ex dynami c ke y factor s in volve d. Once those facto rs are tor s, and mod el s, ma nagem ent fligh t simula ph eno me na. CLD s are the system s thi nk er’ s pinp oin ted , Beh av ior Over Time Diag rams lea rning lab or at ori es, dema nds the hig hes t equ iv alent of the pai nter ’s pri ma ry co lo rs. and/ or Ca usal Lo op Diagrams can be used level of tech nical profi ci en cy to cre ate. Syst em sArc het ypes is the na me giv en to to expl ore how th ey interact. On the ot her han d, ver y little adva nce certa in co mm on dyn ami cs tha t see m to A QQ dia gram be gins with a heav y trai nin g is requ ired to us e them once the y recur in many dif fer ent set ti ngs . These to th e issue hori zo ntal arro w that points are deve lop ed . • archet yp es , con si sti ng of va ri ous comb in a- be ing ad dresse d. Major (q uant it a- “hard” ti on s of balanci ng and re info rci ng lo op s, ar e tive ) fa ct ors bran ch off along the top and the sys tems thi nke r’s “p aint -b y- numbe rs” “s of t” (qu alit at ive) factor s run al on g the set— users can tak e re al- wor ld exa mp les bo ttom . Arrow s le ad ing off of the maj or and fit the m into the appr op riat e ar chet ype . fa ctor s rep resen t sub -facto rs, which can in 1 0 PEGASUS , IN C. 781.398.9700 COMMUNICATIONS SY STEM S THI NKIN G TOO LS

13 - BASED TOOLS UTER COMP RU S TOOLS G IN K THIN RAL U T CT T S TOOL HINKING C NAMI Y D Computer Mo del Gr aph ical Funct ion Diagr am Beh avi or Ove r Tim e Di agr am f(x) A B C x Time Le ts you tr ans lat e all re lat ionships the be havior Can be used to graph of Captur es the way in wh ic h one var iable ident ified as rele vant int o mathematical vari ab les over tim e and gain ins ig hts into affec ts anothe r, by plot ting the relation - an y inter re la tionsh ips be twe en them . equat ions. You can the n run polic y ship bet we en the tw o ove r the full range of rel evant values . analys es thr ough multiple . (BO T diagra ms ar e als o kn own as simulations re fe rence mod e dia grams .) Cau sa l Lo op Diagr am St ru ct ure -Be havi or Pai r Ma nageme nt Fl ight Si mul ator COCKPIT B DECISION INFO s STOCK o HIRING STOCK C R HIRING B s A Time s Prov id es “f light tr ain ing” for ma na ger s Used in conju nct ion wit h be havior Cons is ts of the bas ic dynam ic struct ure s over throug h th e us e of inter acti ve co mpute r block s for tim e diagr am s, ca n he lp you iden tify th at ca n serve as building gam es bas ed on a compute r model. User s re infor cing (R) an d balanc ing (B) er models (for exam - de ve loping comput nce s of can rec ogn ize lo ng-ter m co nseque ple , expone , pro cesses. ntial grow th, de lays , smooths dec ision s by formul ati ng str ateg ie s and S- sh ape d grow th, os cillat ions, and so on) . mak ing dec ision s based on those str ate gies. Po lic y Str uct ur e Di agr am Syst ems Arch ety pe Learning Labora tory Reflection Experimentation He lps you recognize common sys te m r’s pr actic e fie ld. Is equivalent A manage A conc ep tual map of the dec is ion- making beh av io r pa tt ern s su ch as “Dr ifting ion. pr oc es s embe dded in the or ganizat to a spor ts te am’ s expe rie nce, whic h Goa ls,” “Sh if tin g th e Burd en,” “Li mits to with ble nds active exper ime ntation Focu ses on the fac tors that are weighed Gro wth,” “F ixes Th at Fail ,” and so on— re fle ction and dis cus sion. Uses all the for eac h de cis io n, and ca n be us ed to al l the co mpel ling, rec urr ing “s tor ies ” of sy stems think ing tools , fr om behavior bu ild a libr ary of ge ne ric str uc ture s. org an iz ation al dy na mic s. over time diagr ams to MF Ss . 1 1 , INC. THI NK IN G TOOLS WWW.P EGAS USCOM.CO M ICATIONS SYSTEMS PEGAS US COMMUN

14

15 P A R T I I T O O L S D Y N A M I C K I N G T H I N BEH AVI OR OVER DIAGRAM TIME CAUSAL LOOP DI AGRAM S SY ST EMS ARCHE TYPE g’ at Mul ti ple Level s” on p. s di sc uss ed in “Le vel s of Unde rstan di ng: ‘Fire -fightin A 8, we ne ed to de vel op our capabi li ty to see beyond the even t-to -event vi ew of the wor ld. The dynam ic thi nking tools provide the means to repres ent the patterns of ev ent s tha t occur ov er time and also map the stru ctu res that are pro duc ing those dyna mi cs. Th is sec ti on be gins by discu ssi ng rein forc in g an d bal anc in g loo ps, whi ch are k loop stru ctu res the funda ment al buil ding bl ock s th at help us repres ent the feedbac re spo ns ibl e for gene rat ing the dyn amic patt erns th at we obs erve. Al thoug h the bas ic conce pt of rein forcin g an d bal anc ing loo ps is simpl e, actu all y ma pp in g out one ’s ow n issues in a fre e-form cau sal lo op di agram min g ses sion req uires s” (p. 18) can a fa ir amo unt of ski ll. “Gui del ines for Drawi ng Causal Loop Diagram give you some heu ri sti cs to follow in tryin g to cons tru ct yo ur own di agrams . Even with ex pe ri enc e, it can be rat he r daun tin g to stare at a bl ank page an d try to cons truct a sys tem ic pic tur e of you r issu e from scrat ch . Thi s is wh ere sy stem s arc he typ es can be ver y he lp ful in provi di ng the in iti al story lin e fro m wh ic h to eli cit unde rst andi ng of an issue. The arch etyp es repres ent generi c st ory lin es and struc tures tha t ha ve been fou nd to be prevale nt in our sys tem s. “Sy ste ms Arc hety pes at a Glan ce” (p. 20) off ers des crip ti ons and guid el in es for each of th e arc hety pe s (fo r a more com - pl ete cov erage of all th e ar chet ype s, se e Sys tems Arch etyp es I: Di agnos ing Syst emi cIs sues andDes ig ni ng Hi gh -L eve rage Inte rventions, al so publ ish ed by Pegasu s Com muni cati ons) . By deve lop ing an understan di ng of each of the arch etypes , we can beg in to enri ch our intui tive se nse of how com plex struc tures wo rk. 1 3 THI NK IN G TOOLS SYSTEMS M WWW.P EGAS USCOM.CO , INC. ICATIONS PEGAS US COMMUN

16 O O L B O X T I N G B A L A N C A N D N G F O R C I R E I N O F I N G B U I L D L O O P S : B L O C K S S Y S T E M S D Y N A M I C for ms, Wa ts on and Cr ick exp lor ed mo re n the book th e infinit e va ria tio ns we see in na ture ca n The Do uble He lix, I simple geo me tr ica l des ig ns . They event u- al l be produce d by one sim ple , ele gant Jam es Wats on des cribes the pro ce ss al ly rec eiv ed the Nobe l Pr ize fo r reve aling struct ure . thr ough which he and Robert Cric k helix struc tur e tha t is the the double “crack ed” the DNA code . While othe r Simila rly, two basic lo op s— re inforc ing genetic bas is for all life . Thr ough the ir and bala ncing —can be seen as the equiva for comp le x - res ear cher s we re sea rching re sea rc h, Wa tso n and Cr ic k prove d that es to expla in the div er sity of life structur lent building blocks of com plex soc ial and econo mic syst em s. The se simple str uc ture s comb ine in an inf init e va riety of way s to the comp lex sy st ems that we as pr oduce F O R C I N G L O O P - Y E E E M P L O R E I N S O R E R V I S U P ma nag er s are exp ecte d to contr ol. Structure Time Over Behavior REI NF OR CI NG LOOP S: EN GINES OF GROW TH Employee AND DE CA Y Perf. Performance s Supportive Level Behavior Reinf or cing lo op s produc e bot h gr owt h cha ng e and de ca y. Tha t is , they com pound R1 in one dir ect ion wit h ev en mor e cha ng e. Unsupportive Supervisor’s Behavior For ex am ple , in the “E mploy ee-S up ervisor s Supportive Behavior Time Reinf or cing Loo p” diag ram , enc our ag e- me nt from the supe rvis or is ca pab le of Re inforcin gloops compo un dchange in on edirect ion wit heven more chang e. For example, good em ploy ee per for ma nce — pr oducing behav - while critical or unsupportive enco ura gemen tcan en han ce an employeeʼs performance, tra te s tha t is , as the supe rvis or demons iorca nlead to poo remplo yee performance over time. sup po rt ive beha vior , the emp loy ee’ s per for - will lea d the ma nce will imp ro ve, which sup er visor to be ev en mo re sup por tive. At B A L A N C I N G L O O P R O L C O N T I N V E N T O R Y ca n the sa me time , uns up por tive be havior poor emp loy ee pe rfor ma nce ove r pr oduce tim e— if the supe rv isor is not sup por tive, Over Time Behavior Structure pe rfo rm ance will lik ely decr ea se, lea ding s Desired the sup er visor to be eve n less sup por tive . Actual Discrepancy Perf. Inventory o Inventory The sam e loop ca n cr ea te eithe r kind of Level reinf or cing cycle . s B2 Desired Inventory Actual Inventory Actual BAL ANCI NG LOOPS : Inventory Adjustments Inventory GOAL- SE EK ING s Time PR OCE SSE S Of co ur se, most thing s in li fe cannot con - Ba lan cin gloops try to bring asystem to adesired state and keep it there. In an inventory co n- trol syste m, the de sired inventory is maintain ed by adjusting the actual inve ntory whene ver tinue gr owing for eve r. Ther e are other the re is toomuch or toolittle. for ces— ba la ncing lo ops —t hat re sist fur - 1 4 PEGASUS NS, INC. 78 1.398.9700 COMMUNICATIO SYSTE MS THIN KIN G TO OL S

17 SU MMA RY POI NTS he r in cr ea se s in a gi ven direc ti on. t diagram ). If this continues , at som e po int to a Ba la nc ing loo ps try to bring things the sup er vis or’ s supp or tive beha vior will be is pr oduce d All com ple x dy nam ic be havior des ired stat e and keep them the re, muc h ec lipsed by the she er ene rgy dr ain of by two loop s: reinf or cing and bala ncing . li ke a ther most at regu late s the tempe rature wor king long hour s. Imp ro ved perfor - Behind ever y gr ow th or dec ay is at leas t in a hou se. be offset by the man ce will gr adually one reinf or cing loo p. For eve ry goa l- An equi va lent exa mp le in manufactur - until fina lly the bala nc - ef fec ts of burnout, see king be hav ior , the re is a bala nc ing loop . y in g involves mai ntain ing buf fe r inventor ing loop co nnect ing ene rg y le ve l and hours A per iod of growth follo we d by a le vels betw een prod ucti on stage s. In this At this point wor ked be com es do minant. wn in growt h is us ua lly cause d by a slowdo leve l sit ua ti on , there is a desir ed inventory the emplo ye e’s pe rf or mance will eithe r shif t in domina nce from a reinfor cing to a tha t is maint ai ned by adjusting the ac tual pl at eau or de cline . ba la ncing lo op . • in ven to ry wh en ev er the re is too muc h or too li tt le (see “I nv en to ry Control Balanc ing Lo op”). R E I N F O R C I N G L O O P C O U P L E D W I T H USING TH E BUILDI NG A B A L A N C I N G L O O P BLOCK S To see how th ese tw o ba sic loops can co m- Structure Behavior Over Time bi ne to form more co mplex str uctur e- s Energy “Burnout” be ha vio r pai rs, let ’s re visit the Level Perf. o Employee s loop. emp lo yee-su pervi sor fe edback Level Performance R Cle arly the em pl oyee’s pe rformance will B2 R1 Hours B s e the not impr ove in def initely jus t becaus Worked Supervisor’s Diminishing Positive sup ervi sor is su pp ort iv e. The employe e Supportive Returns Reinforcement s ma y have been pu tti ng in longer hours in Behavior Time orde r to con ti nue impr es sing th e sup erv i- sor. Ov er a perio d of tim e, the inc reas ed Reinforci ng and balancing loops can be combin ed to describe more complex behavior. For work hours ma y begin to wear do wn the by the supervisor co uld lead the employee to work lon ger and longer example ,encouragement hours in ord er to continue impressing the supervisor, eventually leading to bu rnout and a leve l (s ee “R einforc ing emp lo yee’s energy decrease in performance. Lo op Cou pled wi th a Balancing Loop ” 1 5 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

18 L B O X O O T B A L A N C I D E L A Y S : L O O P S W I T H N G S E E S O N N G T E R I R - T O T T E E T E A W S ost of us ha ve pl ay ed on a se es aw at proc es s are produc ed by two bala ncing Th e goa l of a see saw ride is to alwa ys M on e ti me or ano th er and can rec all loop s tha t try to sta bilize on a pa rt icula r in a sta te of im bala nce (it kee p things - th e up an d down motio n as the momen price. But the proc ess is comp lica ted by wou ld be pr etty bo ring to sit on a per fe ctly tu m shi ft ed from on e end to the othe r. The th e presence of sig nific ant de la ys (se e bala nc ed one). But the goal in the mar ket - mo re eq ual th e wei ghts of bo th pe ople, the “Supply and De ma nd”) . pl ac e is exactly the oppo site— to bring sm oo th er th e ri de. At a ver y ba sic lev el, a supply in bala nce with de mand. Unfor tu- BAL ANCI NG SUPPL Y AND fr ee mar ket eco no my is a lot lik e a se es aw nate ly , the supply and de mand bala ncing ND DEMA on the with sup pl y at one en d and demand pr oce ss fe els a lot mo re lik e a see saw ride Tr acing thr oug h the loo ps you can see that ot he r en d. Pri ces in dica te the imbalance to a stab le equi - than a smo oth adjus tment if dema nd rises , price tends to go up (a ll be tw een the tw o, lik e a ne edle po sitione d libr ium. As shown in a caus al loop dia - else rema ining the same), and as pr ice goes at the pivot poi nt of the see saw . gram, the dyna mic s of this adj ust me nt up, dem and tends to go down (Bea nie Ba bie s not withs ta nding) . If ther e is enough D S U P P L Y A N D D E M A N invent or y or capacity in the sy ste m to Price ab sor b the incr ea sed de ma nd, prices ma y not go up im media te ly. As de mand out - st rip s supp ly, howe ve r, price will rise . Lo Hi d n a m e On the sup ply side of the see sa w, an D incr ea se in price provides a profit ince ntive for fir ms to produce mor e. Of cour se, it y l p p u S ta ke s time for fir ms to exp and. The leng th of the dela y de pe nds on how close they y a l e s o D alr eady are to full ca pac it y and how quic kly they can add ca pacit y to pr oduce Demand Supply Price B2 B1 mor e. Hir ing new worker s may ta ke only s o additiona l ca p- a fe w day s, while obt aining y a l e D ita l equip me nt or facto ry floor space may ta ke mont hs or even yea rs. While fir ms are ma king sup ply adjus tme nts, the ga p be twe en sup ply and dem an d wide ns and pric e pr ice go es even hig her . The higher Demand sp ur s com pa nie s to incr ea se their produc - tion pla ns even mo re. Supply As sup ply ev entua lly exp ands and ca tches up with de ma nd, price beg ins to fall. By this tim e, fir ms hav e over exp ande d Time the ir product ion ca pa cit y and sup ply ove r- Afree market economy is alot like asee saw with supply at one end and demand on the other. shoot s de ma nd, ca using price to fa ll. When Th edynamics that result from trying to balance supply an ddem and are produced by two balanc - ing loop sthat try to stabil ize on aparticu lar price. Due to the presence of signi fican tdelays, a the pr ice fa lls lo w enoug h, the produc t cycle of oversho ot and col lapse occurs. be come s mo re at tr act iv e aga in and 1 6 PEGA SU S COMMUN IC ATI ON S, INC. 78 1.398.9700 SY STEM S THI NKIN G TOO LS

19 ema nd pic ks up, st ar tin g the cyc le all d fue led a str ong de mand for airp la nes. Tha t This exam ple ma kes it clea r that pieces ov er ag ai n. in tur n sp arke d an inc re ase in air pla ne of the air line le asing indus tr y have ope r- lea se rat es as air lines scr amb le d for addi - at ed wit hin a se esa w str uct ur e. Alt hough ON SEESAWS NES IRPLA A tio nal ai rpla nes . The high lea se rates le d to the ext ended period of air tr affic growt h Th e supply an d dema nd see saw is playe d incre ase d pr of its and a sur ge in air pla ne kep t dema nd ahea d of sup ply for se ve ra l ou t in al l bu t the most tightly regula te d ord ers. Since air pla nes ta ke many months ye ar s, it did not cha ng e the na tur e of the markets. of this balanc ing A goo d ex ample to bu ild, the sup ply of le asable air pla nes dela ys in the sup ply line. Whene ver the ac t is des cribed in a Forbe s ar tic le entitl ed did not adj ust right away, ma king lea se sup ply adjust ment s bring the se esa w bac k “Fa sten Seat Bel ts, Ple as e” (April 2, 19 90 ), rat es go ev en hig her . Thi s led to hig her down, air plane lea sing comp anie s will be s. ab ou t air plan e leasi ng companie pro fi ts, which attr ac ted mor e ca pit al, . in for a bum py landing Lea sing com pani es, which ac count for wh ich wa s the n plo we d into even mor e SU MM AR Y rou ghl y 20 perc ent of all co mmer cial jet ord ers for air plane s. ai rcraft curren tl y on order, hav e enjoy ed The ba lanci ng loo p wi th del ay st ruct ur e is As th e sup pl y catches up to dema nd, eno rmou s prof its during booms in air at onc e si mp le an d comp lex: si mpl e, beca use ho wev er, the ai rp lane le ase rat es wi ll fall tr ave l. At on e time, on e car rier alone put si ng le loop it see ms to be an innocuous (t he sl owi ng of air traff ic growt h wi ll accel - in an order to lease 50 0 plane s. Base d on str uc ture tha t is ea sy to co mpr eh end; co m- era te this pro cess ). With so ma ny ai rplane s le as ing an d bu yi ng rates in the indu stry, pl ex bec ause the resul ting beha vi or is nei - in the pi pel ine , the sup pl y wi ll lik el y beg in was ex pec te d th e to ta l num ber of airplanes the r si mp le nor ea si ly predi cta bl e. The to outs trip dem and and dr ive leas e ra te s to increase by 50 perce nt be tween 199 0 and del ay s in a typ ical syst em are rar el y con si s- do wn ev en furthe r. Th is put s a squeeze on , air tr affic 1995. Bu t in th e meantime in adva nce, an d th e ten t or wel l known prof it s and fo rce margi nal fi rms out of busi - gr owt h had slo wed in the late 1980 s. The cumul at ive ef fect s are usual ly beyo nd the ness. Some orde rs wi ll be ca ncel ed; othe rs le as ing compa ni es, ho wev er, did no t se em cont rol of an y one per son or firm. • wil l be re nego tiat ed. to o wor ried. Ac cordi ng to the article, “ei ght years of un brok en pro sperit y hav e cre ate d th e illu - A N E A I R P L L E A S I N G I N D U S T R Y si on tha t ma ny cyc li ca l bus ine ss es ar en’t cyc lic al an y lo nger .” But, as one airline Traffic Air exe cu ti ve warn ed , “This is a cyclical bus i- Growth y Airplanes a l s e D for Lease will be. s has bee n, always ne ss. Alway s D e l a s o y Wit h a sm al l change in load fac tor, the air - for Demand Airplane Airplane li ne s ca n go from sp ill ing cash to ble eding Airplanes B4 B3 Orders Lease Rate red ink like th e Miss issippi River going o s th ro ugh th e del ta.” Profits s If you dr aw out a causal loop diagr am in this way, you of th is in dustry operating Acausal loop diagram of the airp lane lea sing indu stry shows the same seesaw stru cture at se e the sam e su pp ly and de mand structure work. at work. An in crease in air tr affic grow th 1 7 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

20 L B O X O O T G U I D E L I N E S F O R D R A W I N G P D I A G R A M S C A U S A L L O O It is also he lpful to de te r- Time ho ri zon. • he old ada ge “if th e only too l yo u • Boundar y issue . How do you know T mine an app rop riat e time ho rizon for the have is a hammer, eve rythin g beg in s when to st op adding to your dia gra m? If is sue —one lo ng eno ug h to see the dy nam ics to loo k lik e a nail” can also app ly to lan - yo u don’t sta y focused on the issue , you pla y out . For a cha nge in co rpora te strat egy , gua ge. If our lan gu age is linear and stati c, ma y quickly find yours elf over whelme d by the time hor iz on ma y sp an sev era l years, wit h our we will tend to view and interact the numb er of connec tions pos sible . wh ile a cha nge in adv erti sing campa ig ns wor ld as if it were linear and stati c. Tak ing Reme mb er , you are not tr ying to dra w out may be on the or de r of mont hs. an d circular a com plex , dy namic, wo rl d the whole sys tem —o nly wh at is cr itica l to Time it se lf sho uld no t be includ ed as a it int o a set of snap sho ts and line arizing the the me being addre sse d. When in caus al age nt, ho weve r. Aft er a hea vy ra in - ma y make thing s se em simple r, bu t we may or ha lve doub t, as k, “If I wer e to double fal l, a rive r leve l ste adily rise s over tim e, but to tal ly mis read the ver y reality we were this var ia ble , would it ha ve a sig nifica nt we would no t at trib ut e it to the pa ssage of see king to un derst an d. Making su ch in ap - ef fect on the is sue I am map ping?” If not, time . You ne ed to identi fy wha t is ac tual ly pro pri ate simplificat ions “is like put ti ng on it prob ab ly can be omitt ed. dr iv ing th e cha nge . In com put er chips , yo ur br ak es and then lo oking at yo ur Lev el of aggrega ti on. How det ai led • $/ MIPS (millio n ins tru ction s pe r secon d) spe edo met er to see how fa st you we re shoul d the di agra m be? Aga in , the level de cr eased in a str aigh t lin e in the 199 0s. It DIA go ing,” say s Bill Isaacs of •log os. shoul d be det er min ed by the issu e it self . wo uld be inc or rect, ho we ver, to dra w a The ti me hor iz on also ca n hel p det erm ine AR TI CUL ATI NG REALITY caus al conne ction between tim e and how det ai led the va riabl es need to be. If the Cau sal loop diagr ams pro vide a lang uag e $/ MIPS. Ins te ad, inc re asin g inve st me nt s and ti me hor izon is on the or der of week s (f lu c- g of the fo r ar ticu lat ing ou r under standin le ar nin g cu rv e eff ect s were lik ely ca us al tua tio ns on the pr odu ct ion line), var ia bl es of our dyn am ic, int ercon necte d nature for ces . tha t cha nge sl owl y ove r a peri od of man y wor ld. We can thin k of the m as sent ences Beh avi or over ti me chart s. • Ident ifyi ng yea rs ma y be assume d to be const ant (su ch th at ar e co nst ru cted by linkin g tog eth er key and dra wi ng ou t th e beha vi or ove r ti me of as bui ldin g ne w fact or ie s). As a rul e of var iable s and in dicat ing the ca us al rel at io n- key variab les is an imp ort ant fir st ste p thu mb, the va ria bles shoul d no t descr ibe toge ther ship s be twee n them. By stringing to ward art icul ati ng the cur rent under stand - sp ec ifi c eve nts (a br oke n pum p); they sev er al loops, we can cr eate a coher ent sto ry in g of the system . Drawi ng out fut ur e shoul d rep re sent patter ns of beh avi or about a pa rt icular proble m or is su e. beha vio r me ans tak ing a ris k— the risk of (p ump br eak do wns thr ougho ut the pl an t) . Follo wi ng ar e some more ge neral bein g wr ong. Th e fact is, any pr oje cti on of • Ma ke sur e to ide ntify Signif ica nt de lay s. gui deli nes th at sho uld help le ad you th e fu tur e wil l be wro ng, bu t by maki ng it dela ys (if any ) links ha ve sig nificant which th rou gh the proces s: tions and ex pl ici t, we can te st ou r assump rela tive to the res t of the di agram . Dela ys Cre ating causal lo op Th em e selec ti on. • un cov er incon sisten cies tha t may ot her wi se ar e imp or ta nt beca use the y are ofte n the di agra ms is not an end unto its elf, but part . Fo r exam pl e, dra wi ng nev er get surfaced sour ce of imb ala nces tha t accum ula te in of a pr oce ss of articu lating and co mm un i- proj ecti on s of steady pr od ucti vi ty gr owt h the sy ste m. It may he lp to visua lize pres - ca ting de epe r in sigh ts about comp le x issues . whi le tra ini ng do ll ars ar e shr inki ng raise s up in the sys tem by viewing sur es building a caus al lo op It is po intles s to beg in creating th e que sti on, “I f trai ni ng is no t dr ivi ng our the dela y co nnect ion as a relie f valve tha t a th eme or di agra m wit hou t having selected growt h, what wil l?” Th e beha vi or over eit her op ens slowly as pres sur e builds or is sue tha t you wish to under stand bet ter . ti me dia gram al so po in ts out key va ri abl es op ens abrup tly whe n the pr es sur e hits a “To un de rst and the implicatio ns of chang - th at sho ul d be includ ed in the dia gr am, cr itica l va lue. An exa mp le of this might be in g fr om a techn olog y-d rive n to a mar ket - suc h as Trai ni ng Budget and Product iv ity. a de lay bet ween lo ng wo rk hours and in g-o rie nte d st rate gy, ” for examp le, is a Yo ur diag ram sho uld try to cap tur e the bur nout : Af ter sus ta ined per iods of work - d be tte r th eme th an “To better understan stru ctu re that wi ll prod uce th e pr oje cted ing 60+ hour s per we ek, a sudden colla pse our stra te gic plann in g proce ss.” beha vio r. . mig ht occur in the fo rm of burnout • 1 8 PEGASUS , IN C. 781.398.9700 COMMUNICATIONS SY STEM S THI NKIN G TOO LS

21 LE XAMP E ID INE EL GU S Litigation Use nou ns when ch oos ing a vari abl e nam e. Avo id verbs and ac tio n phr ase s, bec ause the E cti on is co nv eye d in th e loo p’s arrow s. For exa mp le, “C osts” is better tha n “Inc re asing a 1 M Costs A Costs, ” bec aus e a de creas e in In creas in g Costs is co nfusi ng. The sign of the ar row (“s” for same Increasing Costs N or “o” fo r oppo site ) ind ica tes whe th er Costs incre ase or de crease re lative to the oth er variable. E L Rewards se var iable s th at re pre se nt quantitie s th at can vary over tim e. It does no t make sense to U B A I 2 on the othe r say tha t “St ate of Mind” inc re ases or de creas es. A ter m like “Happiness,” Happiness R hand, ca n va ry . A V State of Mind G Demand N For example, Wh ene ve r pos sible, choose the more “pos itive ” sense of a variable name. I T or decrea se the co nce pt of “Growth” inc re asing or decreas ing is clea rer than an increase 3 C Growth E in “Co ntr action.” L E Contraction S s for outcomes Think of the pos sible uninte nde d conse qu en ces as well as the expected Production Output an increase For example, ev er y cours e of ac tion inc lude d in the diagram. in “Production 4 s Production Pressure S tress Pr es sur e” ma y inc re as e “Pr oduc tion Output,” but it may also inc rease “Stress” and de crea se o “Qua lity. ” Q uality, etc. Quality s All bala ncing loops ar e goal-s ee king proc esse s. Try to make explicit the goals driving Quality s , Loop B1 may raise ques tio ns as to why increa sing “Quality” th e lo op. For example 5 o By explicitly Quality.” would le ad to a de cr ease in “Ac tions to Improve ying “Desired identif Desired B1 Quality Gap in is really dr iving impr ove - Qu ality ” as the goal in Loop B2, we see that the “Gap in Quality” B2 Quality Actions to Actions to ment act ion s. s Improve Improve o Quality Quality s N D s O Quality” Distin guis hing bet wee n pe rc eive d and act ual stat es, such as “Perceived e Actual I l a s y T Quality than reality slower Per cept ion s often change ver sus “A ctual Qua lity ,” is important. does, 6 C Perceived and mis ta king the perc eive d status for curr ent rea lity can be misleading and create undesir - U Actions to Quality R Improve abl e res ults. B2 R1 T Quality o S s N s in Gap Desired O Quality Quality C s P O the m into one term while If a varia ble ha s multiple conseq uenc es, start by lu mping O s many can represent Strategies” “Coping co mple ting the re st of the loop. For example, 7 L diffe re nt way s we respond use, etc.). to stre ss (e xe rcise, medit ation, alcohol Stress Coping B Strategies o and short-te Dra w rm cons equences. Actio ns alm os t alwa ys have dif fer en t long-term the Loop B1 shows la rge r loo ps as the y progr ess fr om short- to lon g- term processes. 8 s out the long- draws stress . Loop R2, however, sho rt-te rm beha vior of us ing alc ohol to combat Stress B1 Alcohol Use o term cons equ ence s of this be havior, stress. increases that it actually showing o o R2 Productivity Health s o two te rm s re quire s a lot of explanation to be clear, redefine the vari - If a link between Demand Quality ab les or ins er t an inte rme diate te rm. Thus, between the relationship and “Demand” 9 S s o Production is inserted Pressure” whe n “Pr od uction them. “Qua lity ” ma y be more obvious between Quality P Demand Pressure I T L o whe the r a loop is balancing is to count the or reinforcing A sho rt cut to dete rmining Bank A Failures R loop (i.e., an a balancing of “o’s” indicates numb er of “o’ s” in the loop. An od d number 10 E o an even number or no od d numbe r of U-tu rn s kee ps you he ade d in the opposite direction); N Depositors’ Solvency E R th e loop, you should : Af ter labeling always “o’s” mea ns it is a reinf orc ing loop. CAUTION Confidence G re ad thr ough it to make su re the story agrees with your R or B label. o Withdrawals o Banks from 1 9 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

22 O O L B O X T S Y S T E M S A R C H E S AT A G L A N C E T Y P E GUIDEL INES TION SCRIP DE ARC HET YPE Dr if ting Goals • Drifti ng pe rfo rman ce fi gures ar e usu- In a “Dri fti ng Go als ” ar ch ety pe, a gap o ally in dicat or s th at th e “D rift in g Go al s” bet wee n th e go al an d cu rrent real ity can Pressure to arch ety pe is at wo rk an d that real cor- be re so lved by taki ng co rr ecti ve act ion Goal Lower Goal rect ive ac ti on s are not bein g ta ken . (B1) or low er ing the goal (B2 ). The B2 s • A crit ica l aspect of av oidi ng a pote nt ial cri tic al di ff er enc e is th at lower ing the s Gap “Dr ift ing Goals” scena ri o is to dete rm ine goal imm ed iat ely closes th e gap, wh er eas o what drive s the set ti ng of the goa ls. corr ec tiv e ac tio ns usual ly tak e time. (See s 1 B • Go als locat ed ou ts ide the syste m wi ll be Sys T , em t r he T inke Octo ber 1990.) h s orrective C Actual Action les s susc eptible to dr if ti ng goa ls pr es su re s. y a l e s D In the “E sc al ati on” ar chet ype, on e To break an es cal ati on str ucture, ask the tio n Escala par ty (A) take s ac ti on s th at are per - fol low ing ques ti on s: cei ved by th e ot he r as a th re at. The tha t pits • What is th e rel ati ve measure oth er par ty (B) res pon ds in a si mi lar one par ty ag ai ns t th e oth er and ca n B’s Result Result A’s s s o s manner , in creas in g th e thr eat to A an d you chan ge it? Activity Activity re su lt ing in more th reat eni ng acti on s • What ar e the signifi can t de lays in the Quality of A’s Position B2 B1 B by by A Relative to B’s by A. Th e rein for cing loo p is tr aced sys tem th at may di sto rt the tr ue na tur e s s Threat Threat out by fo llo wing the outli ne of th e fi g- of the th reat? s o to B to A ure- 8 prod uced by th e two bal an ci ng • What ar e the deep- rooted assumptio ns er st k in Th ms , loops. (See Sy e h T e that lie ben eath th e acti on s ta ken in Novem ber 199 1.) res pons e to th e threat? That Fail Fixes In a “F ix es Th at Fai l” situ ati on, a • Br eak ing a “Fi xes that Fai l” cyc le usu- pro blem symp tom cries out for res olu- al ly req uires ac knowl ed gi ng th at th e s tion . A solu tion is quickl y impl e- fix is mer ely al lev iati ng a sympto m, Problem me nt ed th at all eviat es th e symp tom an d mak in g a co mmi tmen t to so lv e Fix Symptom B1 (B1 ), bu t the un inten ded con seq uen ces the real probl em now . o s ate the pr ob lem of the “f ix” exacerb • A two-pr onged att ack of app ly ing th e D e (R 2). Ov er time, the pr ob lem symp tom fix an d plan ning ou t th e so lu ti on wi ll l a R2 y retu rn s to its pr evi ous lev el or be comes hel p en su re th at you don ’t get cau gh t Unintended s Consequence Sys Th e , t e ms Thi n ke r worse. (S ee in a per pet ual cycle of so lvi ng yest er - Nov emb er 1990 .) day s “so lu ti ons .” th and Und er inve st me nt Grow an d Un de rinve st men t” In a “Growth • Dig int o th e assu mp ti ons whi ch driv e ar che ty pe , gr owt h ap proach es a limi t s capac ity inves tmen t dec isi on s. If past s that can be eli min at ed or push ed int o per forman ce domi na tes as a con sider - Growth Demand R1 B2 Effort the fut ure if cap acit y inve st me nt s are ati on, tr y to bal an ce th at per spec tiv e s Performance o made. In stead , per form an ce sta nda rds with a fres h loo k at deman d an d th e Standard Performance s ar e lowere d to jus tify un der inve st me nt , facto rs that drive it s gr ow th. o s lead ing to lower perform an ce which • If ther e is pote ntial for gr owth, build Need Perceived Capacity B3 to Invest furth er jus ti fie s und er inve st me nt . capac ity in antic ipati on of fu ture s D y e a l l a y e D er , June/ July (S ee T h e Sys t em s Th in k de man d. Investment s in Capacity 199 2.) 2 0 PEGASUS NS, INC. 78 1.398. 9700 COMMUNICATIO SYSTE MS THIN KIN G TO OL S

23 GUIDELI NES TION DES CRIP PE AR CHETY • The ar chety pe is mo st helpful when it is In a “Li mits of Suc cess” scen ar io, co n- Lim its to Suc cess used wel l in ad van ce of any pro blems, tin ued ef for ts in iti al ly lea d to to see how th e cumul ati ve effe ct s of imp ro ved perf orm an ce. Over ti me, on ti nu ed succes s might lea d to futur e c owev er , the sy stem enco un ter s a h onstraint C pro bl em s. lim it whi ch caus es th e per for man ce to o s s • Us e the ar chety pe to ex plor e que stions slow down or ev en decl ine (B2 ), even Limiting What ki nds of pres sures ar e suc h as T as eff ort s co ntin ue to rise. (See he erformance P Efforts 2 1 B R Action building up in th e org an iza tio n as a nk te ys S T er , h i ms Dec ember s s resul t of the grow th ? 1990/ Jan uar y 1991 .) • Look for ways to rel iev e pre ss ures or before remov e limi ts an org an izati onal gas ke t bl ows. • Pr ob lem symptom s are us ually eas ier to In a “Shi ft ing the Burd en,” a proble m is Shift ing the Bur den /Add ic tio n recog nize than the ot he r ele ment s of the “so lve d” by app ly ing a sy mpt om at ic struc ture. sol ut ion (B1), which dive rt s at tention ymptomatic S Solution • If the sid e-ef fec t ha s be com e the problem , s aw ay fr om mor e fundamenta l solut ions you ma y be dealing wit h an “Addic tion” t e m s T h i nk e r , T h e Sys (R 3). (S ee B1 s struc ture. Se pt emb er 1990. ) In an “Ad dict ion” o Problem • Wh eth er a sol ut ion is “symp toma tic ” or str uctu re, a “S hift ing the Burden” Side-Effect R3 Symptom o “fun dame nt al” oft en depe nd s on on e‘s per - de grade s into an addict ive pa ttern in D e l spec ti ve. Exp lor e the prob le m fr om a dif - a whi ch the si de-e ffect ge ts so ent re nche d y B2 feri ng per spe ct ive in order to come to a th at it ove rwhelms the origina l proble m s Fundamental more comp re he nsi ve un ders tan din g of e S yst e m , s T symp tom . (S ee T h i r h e nk Solution o what the fu nd am en tal sol ution may be. Apr il 1992. ) • Look for reas on s wh y th e syste m wa s set l” In a “S uc ces s to the Su ccessfu Success to the Su cc ess fu l up to create ju st one “wi nner. ” ar chet yp e, if one per so n or group (A) • Ch op off on e ha lf of the archet yp e by is gi ven more res our ces, it has a foc usin g ef for ts an d reso ur ces on one hig her like li hood of su cceed ing th an Success s o Success B of A of gro up , rat her th an cre ati ng a “winne r- B (assum in g th ey ar e eq ual ly capab le). s Allocation A to s R1 R2 tak e- al l” com peti tion . The init ial suc ces s justi fi es devo ti ng Instead B of • Find way s to ma ke teams co lla bor ato rs more res our ces to A, an d B’ s su ccess Resources Resources s rath er th an com peti to rs. o dimi nis hes , fu rt her just ify ing mo re to B to A • Iden tify goal s or ob jecti ves tha t define re so ur ce all ocat ions to A (R 2). (See succes s at a level high er than the indiv id - T e Sys t em s hi nk er , Mar ch 199 2.) T h ual player s A an d B. s Tragedy of th e Co mmon s” str uc- In a “T ra gedy of th e Common • Effec ti ve sol ut io ns for “Tr agedy of the s tur e, eac h pers on pur su es acti on s Co mmon s” scen ar io nev er lie at the Net Gains s for A whi ch ar e ind iv idu al ly ben efi cial (R1 indi vid ual level . R1 B5 an d R2). If the am ou nt of acti vi ty are the • Ask ques tio ns such as: “What gr ow s to o lar ge for the sy st em to su p- incent iv es for in div idu al s to persist in A’s o s Activity Resource port , ho wev er, the “co mmo ns” thei r ac ti ons?” “Can the long -te rm col - s Limit R3 bec omes exp er ien ces dimin ishi ng ben - lec ti ve los s be mad e mo re rea l and Gain per Total y Individual a l Activity e s o D Activity ef its (B5 an d B6) . (S ee Th e Sys te ms immed iat e to th e indivi dua l acto rs?” s R4 r, h i nke T Augus t 199 1.) • Find ways to rec on ci le sh ort -term B’s s Activity o cumulat iv e conseq uen ces. A gov erning R2 bod y th at is char tered wi th the susta in - R2 B6 ab ility of th e res our ces limit ca n he lp. s Net Gains for B s 2 1 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

24

25 P A R T I I I S S T R U T H I N K I N G T O O L C T U R A L ION DIAGRA IC AL FUNCT GR APH M ST RUCT URE-BEHAVIO R PAIR tool s can hel p us bec ome ev en mor e ex pli cit abou t th e struc - truc tura l thinking S ic be haviors we are try in g to und erst and. “Fro m tures th at cre ate the dynam Caus al Loop s to Gr ap hical Function s: Artic ul ati ng Chao s” (p. 24) an d “G raphi cal Fu nct io ns : ‘Seei ng’ the Full St ory” (p. 26) desc ri be how graphi cal fu nctio n di agrams ar relati onsh ips . Th ese rel ati ons hi ps ch aracte ri ze th e can easi ly rep res en t nonline natur e of mos t int er conne cti ons in com pl ex sy stem s (as oppo sed to th e sim pl e, linear rela tio nshi ps th at ar e of te n assum ed ). The sec ond half of thi s sect ion focu ses on stru ctu re- behav io r pai rs. Ac cu mu lato rs and fl ows pr ov ide a rigor ous frame work for represe nti ng sy st emi c struc tures in a mo re preci se way than th rough cau sal loop di agrams al on e. They can bet ter represen t a sys te m’s nonl inea rity, as we ll as dist in gui sh betwe en thi ngs that ac cum ul ate (water in a bathtu b) and things that flow (w at er flo wi ng throu gh a fau cet) . Th e arti cles on ac cu - mu la tors and fl ow s (p . 28–3 7) sho w how thes e con cepts ad d fu rth er prec is ion to ou r th ink ing and unde rstandi ng about th e li nk bet ween stru ct ure and behav io r. 2 3 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

26 O X L B O O T C A U S A L L O O P S T O F R O M G R A P H I C A L F U N C T I O N S : N G C H A O S A R T I C U L A T I Cha os: Ma king aNew Scien ce n that way. It is not enoug h to have a model invol ved . The que sti on of how ma ny fac - I (P engu in Bo ok s, New Yo rk ), Jame s that re pro duce s som e re al-wor ld phenom - tors to incl ude al ways depends on the pur - Gle ick de scribes a rela tively ne w branc h of en a if we cannot ide ntif y the st ructur es pose of exa mi ning th e sys te m. Since the sci ence tha t has prof ound implic ations for e the be hav ior of the actua l that produc x, det ai ls of an y sy st em are infinit ely comple ho w we view our world . Chaos, simp ly sy st em . That is why sy stem s thinking dia - it is futile to st ri ve to “mod el the syst em. ” order and pat - pu t, is th e sc ienc e of seeing gram s focus on captur ing realit y in a for - In ou r sam ple case , the pu rp ose is to repre - te rn wher e fo rmerl y only the ra ndom, mat that tap s into our int uit ive sen t as conci se ly as possi bl e the im port ant er rat ic , and unpredi ct able had been of the sy stem s in which we unde rs tanding fac tor s that affect the bala nce of a typica l obs erved . In a way , systems think ing also man age and live . savi ngs accou nt, so we wan t to look at sa v- de al s in th e sci enc e of chao s. Dia grams ing s, in com e, intere st ea rned, and spe nding FRO M CA USA L LOO PS TO su ch as ca usal loo ps, accumulator s and (see “S avi ngs and Spend ing Loops” ). If we PHICA L FUNCTI ON GRA are all ways fl ow s, and gra phi ca l functions wer e on ly inter est ed in capt uring the fact DI AGRAMS structure of extra cting the underlying tha t th er e is a ba lan ci ng loop that expla ins To see how a range of sy st ems think ing fr om the “no ise” of ev ery day lif e. the sl ow down in the growt h of our sa ving s tool s can he lp ca ptur e the str uct ur e of a accou nt, we cou ld st op at th is point . On the RE LA TING BEHAVI OR TO sy st em at incr eas ing le ve ls of de ta il, le t’ s other hand , if we wan t to be more explicit ST RUC TURE look at a sy ste m we ar e all fam iliar wit h— ab ou t the st ruct ure behi nd the beha vior , we a savi ngs account. If we plot out the st ruc - Bot h syst ems think ing and chaos insist that need to tr ansl ate ou r diag ra m in to accumu - ture of a sav ings acco unt using a caus al re al -wo rld phenom ena need to be lat ors an d flow s. loop diagr am (se e “S avin gs Loop ”), we se e de sc ribe d in term s that reflect our intu - that an incre as e in sa vings will lea d to iti on . Wr iti ng pa rti al differential equations more inter est ear ne d, which incr ea ses our to de scribe cl ouds, for example, miss es the S A V I N G S A N D sav in gs bala nce still fur the r. The gra ph of clouds in po in t, bec ause we do n’ t perceive N D I N G L O O P S S P E the be ha vior over time look som ething would Spending s P S L O O S A V I N G lik e the exp one ntia l gro wth curv e show n on B2 the righ t of the dia gr am. s Savings $ “Wa it a mi nute ,” you o Savings s ma y pro te st, “I don’ t R1 kn ow wh ose ban k ac cou nt that is, but it ce r- R1 Interest s ta inly does n’t look like Earned Time Interest s mi ne !” Tha t’s tr ue— Earned ra rel y is a sy st em so si m- If webegin to explore ou rsavings account “syst em” bydraw ing pl e in re al life; nor ar e acau sal loo pdiagram, we se ethat an increase in savi ngs will Abalancing loop explains the slowing growth ba nk accou nts th at wel l- lead to more intere st ea rned, wh ich increase sour savings bal - of the savings account: As our savin gs ance still further (left). The graph of this behavior over time will increases, we are more like ly to increase be have d. The re are usu - look somethin glike the exponen tial growt hcurve (righ t). spendin g, which will reduce our savings. al ly man y ot her fact ors 2 4 PEGASUS CA TIONS , INC . 781 .398.9700 COMMUNI SYSTE MS THIN KI NG TOO LS

27 ACC AND UM UL ATORS mont h on discr etionar y exp ense s.” S AV I N G S P O L I C Y FLOWS G R A F U N C T I O N P H I C A L Gra phical fun ct ion s al low us to expan d D I A G R A M our exp lorat ion of a syst em to in cl ude pol i- Whe n we tran slate CL Ds into accu mu la - ci es an d in terre lat ion sh ips betwee n va ri- to rs and flow s, we ar e be coming ev en mo re able s. If we tried to captur e the sa vi ngs pre cis e abou t the st ructures prod ucin g the pla n we descri bed above in an analyt ical dyn am ics. The bat ht ub as a metap ho r for s Savings fo rm, we wou ld have to do quite a bit of ho w co n- accum ula tio ns help s us visualize Spending work in orde r to come up wi th a su itab le ce pt s as dive rse as savi ngs, pol lutio n, cu s- 500 s e eq ua tion . An d wh en we wer e don e, it to mers, an d cor por ate rep utation sha re a s n e p x E . r woul d be hard to te ll if th e eq uat ion repre - (see sim ila r und erly ing structure c s i D 0 20000 sen te d our savi ngs ac coun t or the number s, Batht ubs “A cc umu la tors: Bathtub Savings of wid get s on sal e at Wal -Ma rt . Th e truth Everyw here ,” p. 30). is , mos t of us don ’t th ink in ab st ract mat he - Ac cumu lators an d flows add more This Graphical Function Dia gram captures ur savings pol icy in an intuitive way by map - o matical conce pts, bu t in images and st ruc- det ai l and unde rs tan di ng to our cau sal loop ing out the relationshi pbetween savings p tu res grou nd ed in our eve ryd ay experi en ce. di ag ram by di ffer entiatin g betw een thos e and discretionary expenses. Tha t is why gr aph ical fu nct ions ar e usef ul. var iab les in the diagram that “ac cu mu late ” The y cap tu re pol icie s in an int uitive way (our savi ng s bal anc e) and those that jus t th rou gh a si mple graph th at maps ou t the (inc om e and “fl ow” throu gh th e system rel ati on sh ip bet ween on e var iabl e in rel a- sp ending ). In th e “S avin gs as an as ite ms on our product ion rep ort s, finan - ti on to an ot her (see “Savin gs Pol icy Acc umul ator ” di ag ram, we can vis ual ly see cia l st atem ent s, and cust omer sur ve ys. To Gra phi cal Fu nc ti on Di ag ram ”). In ou r sa v- money fl owin g in to an d ou t of savi ngs in the ext ent tha t those det ails do not ca ptur e in gs pol ic y plan , for exam pl e, we see at a the form of inco me and spendi ng. Mor e the cor e st ructur es tha t ar e im por ta nt, we gla nce th at savin gs has no impa ct on ou r im port antly , we can re late to thi s stru ct ur e ma y be the unwitt ing produ ce rs of our di sc ret ion ary exp en se s unt il savi ngs hi ts int ui tivel y bec aus e we experienc e mon ey in own cha os. As one sy st ems think ing $50 00. After th at , discr eti on ar y expen ses term s of fl ows an d accumul at ors (or lack ma xim warns, “I t ain’t what you don’ t ri se unti l savin gs reach es $20, 000, at wh ich there of!). know tha t hur ts you, it ’s wh at you DO poin t th ey level ou t at $500. know tha t ain’ t so. ” • GR APHIC AL FU NCTIO NS: MAP PING POL ICI ES AR TI STIC MA NAGE RS good idea of both So now we have a pretty Ph ysi ci st Mitche ll Fe ige nb aum sug ges ts the ba sic dy namic behav ior of the savings that art is a the or y about the way the acc ount , and a feel for the important wor ld loo ks to huma n is still infl ow s and out fl ow s. But our model bei ngs. “I t’s ab undantly A C C U M U L A T O R S AV I N G S A S A N pr ett y ele menta ry. Supp ose now yo u obvi ou s tha t one doe sn’t and us e a sy s- wa nt ed to go a li ttl e further kn ow the wor ld ar ound s your family ’s te ms dia gra m for describing us in deta il. What artis ts Interest Rate Interest s Earned “Our po li cy for ma na ging yo ur savings. have acco mplishe d is rea l- Fixed Expense R3 dis cre tiona ry spending depends on how izin g th at ther e’s only a mu ch savi ngs we hav e,” you ex plain. “I f sm al l am ount of stuf f the ba la nce in our sav ings acco unt is below that ’s impor tant, and se e- s s Savings $5000, we don’t spend a dime. As our sav - s ing wh at it is .” Income Spending ing s ri se abov e $50 00, we may inc rea se dis - Wh ethe r we re co gniz e Salary cret io na ry sp endi ng by, say, $15- 20 per it or no t, we ar e artis ts as mo nth . If our sav ings tops $10,000, the n wel l, sel ective ly picking Accu mulators and flows add more detail and unde rstanding to we ’re lik ely to sp end sev era l hundred do l- our causal loop diagram by differentiating between those vari - ou t det ails of the wor ld (sa vings) and those that able sin the diagram that “accumulate” lars a mont h. But in any case, we don’t see that we choo se to foc us “flo w” through the system (income and spend ing). ou rs elve s sp endi ng more than $5 00 per on . Th ose de tails app ea r 2 5 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

28 O O L B O X T : G R A P H I C A L F U N C T I O N S “ S E E I N G ” T H E F U L L S T O R Y pe rsp ectiv es. Gr aphic al funct ions ca n help frame of ref ere nce fo r th eir di sc uss ion, each n execu tiv e of a la rge aut omot iv e A us go bey ond me re ly obse rving cor rela - compa ny tells the story of two eng i- had ass umed the othe r’s view po int was wr on g. tiona l rela tio nship s (whe n X hap pens, Y ne ers who were arg uing ab out the co rrect DUAL WOR LD S INDIVI hap pens) to exp lor ing our under st anding angle of an eng ine moun t. The two had io n be twe en two va ri - of the caus al connect be en at it for more tha n ha lf an ho ur—o ne rs point s out Th e story of the two enginee ab les (X ca use s Y). In constr uct ing GFD s, engine er swe ar ing that the ang le was 40 an age -old com munic atio n prob lem . Each follow the 60 per cent rule —it ’s one should that it was de gre es while th e other fumed of us car rie s our own set of assum pt ions be tt er to get it 60 per cent right ve ry 50 de gr ee s. After seve ral civil attemp ts to abou t reality —our l pict ure own individua quic kly and sp end time modify ing it tha n co rr ect ea ch ot he r’s vie wp oint, they had ju st of the wo rld. Ofte ntime s, we mist akenly sp end a gr eat dea l of ef for t try ing to get it star ted atta ckin g each other’s inte ll ig ence, as sume that our vie wpo int is the only way 10 0 per cent rig ht the fir st time. when the ex ecut iv e ability, and ch aracter of look ing at a situation. Bot h eng ineer s, ha pp ene d to walk by. for exam ple , be liev ed the ot her perso n’ s GRA PH ICA L FUNCT ION S “What ax is of ref er ence are you pos it ion was base d on the sa me axis as CHA RT S VS. SC AT TER us in g?” he ask ed. thei r own—they ne ve r ev en ques tione d it . exclaimed “Th e ver ti ca l, of co urse!” Grap hica l funct ion s ar e be st de scr ibed by If we do n’ t acknow ledg e our assum ptions on e engin eer . fir st est abl ish ing wh at the y are . no t at the outse t of a discus sion, we ris k ex pe - “Th e ho ri zonta l!” sa id the othe r. Alth ou gh they ma y look si mila r, gra phical ri enc ing the same fr us tra tions as the two in amaz ement Bo th stopped as they fun ct ion s ar e not the sam e as sca tt er cha rt s, en gi nee rs . reali zed they ha dbee nsay ing the same thi ng! which plot on e va ri ab le’ s dat a ag ain st In ma ny ins tance s, spoke n lang ua ge ed a common Bec aus e they had no t establish an ot her’ s. If we we re to look at the rela - e rather than a he lp in can be a hindranc tion sh ip betwe en sales and de liver y dela y our mental pict ur es of com municating plot some usi ng a scat ter ch ar t, we would re al it y be caus e wor ds, un like pict ur es, do . V S D E L I V E S A L E S R Y dat a poi nts an d then dra w a reg res sion line not forc e us to be explicit when ex pla ining AY D E L thr ou gh them (see “Sa le s vs. De liver y Graphics , beca use the y ca n ou r rea soning. Dela y”). re pre se nt ideas mor e cle arly, ca n be a Sales - 25 = 1000 Sales Delay * Delivery ($) Fr om the sca tt er cha rt, we can see tha t much mor e powe rful and ef fect ive mea ns in weeks one thr oug h fo ur , sal es fa ll by ing tion (se e “S ys te ms Think of com munica 1000 K del ay. We can then $25K for each one -week as a Lang uage ,” p. 6) . Tr ite as it ma y ext rapol at e be yond the his tor ical dat a to sou nd, the sa ying “a pictur e is wor th a 900 K predi ct tha t a five- week de lay wi ll resul t in thous an d wor ds” still ho lds tr ue . Had the an addi ti on al $25K dr op in sa les. In gener al , two enginee rs sim ply dra wn tw o ax es and sc at ter di agram s answer the qu est ion, a line, th ey would hav e save d a grea t many K 800 “W hat ha pp ened hist or ical ly, wa s ther e a an gry wor ds. When the issue is mo re com - cor rel ation , and bas ed on th at in for mati on pl ex than a single angle , the use of gr aph - what can I exp ect to ha pp en in th e fu tu re?” ics can beco me eve n more imp or ta nt fo r 3 012 4 (weeks) Delay Delivery They tend to be retr os pect ive. re ach ing a shar ed unde rst anding . A gra phical function , on the ot her func tion dia gra ms Us in g gra phical Scatter ch arts plot one var iableʼs data han d, is ver y mu ch pr ospect ive in na tur e. (GFDs) , it is much eas ie r to cap tur e how ag ain st ano ther and answ er the question, “Wha thappe ned historicall y? ”The vi ew is By includ ing the full spect rum of pos sib le two vari able s re late in a for ma t tha t is con - retrosp ective. the val ues, GFD s ca n hel p you se e beyond ci se an d invite s othe rs to shar e their own 2 6 PEGA SU S COMMUN IC ATI ON S, INC. 78 1.398.9700 SY STEM S THI NKIN G TOO LS

29 st il l com e from cap ti ve cust omer s wh o have hi st oric al ran ge of operating val ues an d as much as 80 per cent . no wh er e els e to turn in the shor t ter m. The “E ff ect -of ” versi on of the GFD ask , “Giv en my un ders tandi ng of the sys - fo cuses attent ion on the re lativ e im pac t of tem , wh at do I th ink wi ll happen at each BU IL DIN G SHA RE D the deliv er y dela y on sa le s ins te ad of on the pos sib le poin t?” UNDE RSTA ND IN G spe cific num bers th ems elv es. In this way , we CRE ATIN G A GFD The res ulti ng diagram is a con ci se ca usal ca n com pa re acr oss differe nt variab les, suc h hy pot he sis th at stat es th at cu st ome rs wi ll A gra ph ica l func tio n dia gra m can help as the rela tive effe cts of qua lit y on sa le s vs. rewa rd shorte r del ays wit h sl ightly highe r model ex plic at e your (o r a team ’s) mental ma rket ing sp end ing on sale s, and ma ke ord ers , bu t wi ll seve re ly pe nalize dela ys of the re la ti onshi p between two critical ex plicit our un de rst anding of which fa ct or is th at exte nd bey on d an ac cept able ra nge. va ri ab les . Unl ik e behav ior charts, GFDs the dom inant dr iver . • The GFD conve ys a mu ch ove r do no t show how varia bles change ri ch er des cri pt ion abou t the time , bu t how two variables interr elate. To D E L I V E R Y D E L A Y G R A P H I C A L F U N C T I O N D I A G R A M rel ati on sh ip bet ween del iv - crea te a GFD, it is best to begin by answ er- ery del ay and sales th an a Sales ($) ing th e foll owi ng questions: K 1000 sca tte r ch art based on hi stor - • Wha t do we kno w from the outse t ic al dat a. Th e di agram hel ps abo ut the causal rel atio nship betwe en the se K 900 vis ua lize th e fu ll range of tw o va ri abl es? K 800 im plicat ions and min imi zes zo nes” where • Are there any “neutral 700 K th e dan ger of re mai ni ng the varia ble on the y-a xis is not affe cted by 600 K myop ical ly focu sed on a nar- cha nge s in the x-variable? ro w ban d of pos sibl e out- • Wha t are the ex trem e val ues that both 500 K co mes. Devel op ing the va ri ab les ca n assum e? 400 K di agram as a grou p can al so If we looked at the sa les and delive ry delay K 300 he lp sur face diffe ri ng ment al star t by ex amp le usi ng a GFD, we would assu mp tions abou t th e poten- 200 K as kin g wha t we think the gener al natur e 0 20 (weeks) Delivery Delay ti al im pact of de te rior ati ng of the re la ti onshi p is between the two vari - or impr ovin g del ive ry per - abl es—is it fla t, is it upw ard slo ping, or is fo rman ce (r eme mber th e it down ward slop ing? With mo st products, E F F E C T O F D E L I V E R Y D E L AY O N S A L E S en gin eer s!) . lon ge r de liv ery del ays mean lowe r sales , so So me time s it is he lpfu l to would we ca n assum e that the relationship co nv er t the relat io ns hip in to a slo pe downw ard. s e 1.4 l a the mor e gene ral for m where Maximum Advantage Usi ng avail ab le histori cal dat a an d pas t S 1.2 n y- va riab le is conv erte d to an experienc e, we can th en take a fir st cu t at o Neutral Zone y a “e ffe ct -of ” va ri abl e. Inst ea d ident ify ing a ne utr al zon e where sale s may 1.0 l e D of “S ale s” on the y- axis, fo r be inse nsi tive to diff eren ces in the len gt h of y .8 r e ex amp le, we wou ld hav e v ic al the delay (s ee “De livery Del ay Graph i l e .6 D “E ffe ct of De liv ery Del ay on Funct ion Di agram ”). Pas t experi enc e may Maximum Disadvantage f o .4 Sal es ” (s ee gra ph), which sug ge st tha t sal es wi ll inc rease st ead il y as t c e f show s tha t a 3.5 to 4.5 wee k the delay fall s be low four weeks . A sam - f .2 E de lay ha s no effe ct, sho rten - pl ing of cus tome r contac ts may tel l us th at 0 ing the de la y ne ts us a ma xi - 410 0 there is not a wh ole lot of diff eren ce 20 (weeks) Delay Delivery mum gain of 5 pe rce nt (1 .05 bet ween 3. 5 and 4.5 weeks . On the oth er Agraphical function diagram conve ys amuch riche rdescrip - time s the sale s nu mber we h al so tel ls us that hand, past mar ket researc tion about the rela tionship between de livery delay and sa les wo uld hav e obt aine d if we gr eater than fi ve week s, if the delay grows (top) than ascatter chart based on hi storical data. Looking at we re in th e ne ut ral zon e) , sal es wi ll fall drama ti cally. Looki ng fur- the effect ofdelivery delay on sales (bottom) allows you to compare the relative effects of different variables, to see which and le ngt he ning the delay to ther , we rea lize th at even in the ext rem e is the domi nant driver. 20 wee ks can cho ke sa les by of sal es will case of a 20- wee k de lay, $200K 2 7 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

30 L B O O O X T R A L T H I N K I N G : S T R U C T U T H E W O R L D A C C O R D I N G T O A T O R S A N D F L O W S A C C U M U L vi ce pr es ide nt of a ma jo r U.S. man - count ry and ki lled al l th e cows. Wha t Syst ems tel ls thi s stor y: “W hi le per usi ng a A ufa ct ur er once que stion ed whe the r woul d the ab ove mo del pr edi ct fo r next well -kn own eco no mi c jour nal , I cam e tod ay’s ra pid pace of chang e me ans that all woul d yea r’s mi lk produ ct ion? The answer acro ss an arti cle wh ich descr ibe d a mode l ou r ol d to ols and ways of man agin g are now most likel y look a lot like the nu mber for th at ha d bee n cons tructe d to foreca st U.S . ina dequ at e. “Ar e we doomed to ke ep on last year ’s mi lk produ ct ion, wh ich is cl ear ly mi lk product io n. The model wa s of the g out ou r cu rre nt tools and prac tic es throwin inc or rec t. The mo del must be ab an do ned. Y=f( Xi ) form [Y = Y0 + a1X1 + a2X2 +... + as soon as the next wave of innov atio ns “U nf ai r,” you mi ght sa y. “It’s not th at anX n] , whe re the Xi ’s incl uded such thi ngs com es al ong ?” he aske d. the model is wr on g. It ’s just that the wo rld as: la st year ’s mi lk prod uct ion, int er est rates, Th e answ er is . . .“it depe nds.” It has cha nge d dra ma ti cal ly si nce the mo del sp end ing on catt le feed , GNP growt h, and dep end s on th e und erl yin g the ory on which was ori gi na lly bui lt and the chang es must other mac ro econ om ic fact or s. As the ar ti cle the cu rr en t tools an d me th ods are base d. If now be added. ” But wha t has real ly det ail ed, the mod el per for me d qui te wel l as on e’ s ma nag eme nt pr ac ti ces are based on chang ed? Yes, the co ws are now dea d, but a pr edi ctiv e devi ce—at least in ter ms of its tr an sien t or situation- spe cifi c ph eno men a, from cows, the ba sic fact that mi lk comes abi li ty to ‘track hi st ory. ’ The obv ious thi ng they are like ly to requ ire re vision whe ne ver and tha t wi th out co ws the re ca n be no mi lk , abo ut th is mod el , that woul d bother bot h the circ umstance s ch an ge . If, on the other is as true now as it was bef or e th e mass dec- da iry farm ers and peo pl e who we re par tia l ha nd , they ar e based on a stru ctu ral und er - imat ion. Fr om a st ruct ur al per spect iv e, the to ope rat io nal spe ci fi cati ons , is: ‘where ’s the , but the may change st andi ng , the situation nat ur e of the wo rld ha s not ch ang ed at al l. co ws?!’ Simp ly st ated, if you ’ve got no too ls wi ll still app ly. The mode l wa s inade quat e becau se it was co ws, you’ve go t no milk! Crude, but true .” ba sed on situat ion-sp eci fic data th at has Ho w doe s al l thi s tal k abou t cows rel ate WHER E ARE TH E COWS? now chan ged. to our vice pre si dent’s que sti on ? Wel l, d of Hig h Pe rform ance Ba rr y Richmon im agi ne th at an epi de mi c swep t over the ST RUCTU RA L THI NKING When we look at the wo rl d thr oug h a P R O D U C T I O N M O D E L M I L K st ruc tur al lens, we are int er est ed in und er - Maturation_Time st andi ng how thi ng s act ua lly wor k. We ar e less int eres ted in cor rel ational rel ati on ships and more inter est ed in the causal st ru ct ur es tha t produce the ob ser ved behavi or . Thi s is not to say th at no nst ruct ur al model s ar en’t Milk_Cows Calves va lua bl e. Regr essi on model s, for exa mple, Maturation Births MC_Deaths hav e man y applica tions and ar e usef ul for ident if ying cor rel at ion, ex pl ain ing sour ces of var ianc e, and extr apol ati ng from hist ori- cal da ta . Those model s ar e ina dequa te, Annual_Milk_Prodn. Milk_per_Cow how eve r, for ga ini ng insi ght int o how a sys - te m actual ly op er at es. If wewanted to create astructural representation of milk production, wewould begin with the cen - If we wer e to look at the milk pr od uc - tral acc umulator “milk cow s.” Milk production is determined by the number of cows and the amount of milk per cow. To create our hy pothetical scenario of anepidemic, we would simply ti on model fro m a st ructur al vi ewpoi nt , we enter zer ofor the number of milk cows. The resulting annual milk production would also be zero. woul d st ar t wi th the basic fact that mi lk 2 8 PE GA SU S COMMUNI CA TI ONS , INC . 781.398.9700 SY STE MS THI NKIN G TO OLS

31 ay s, man agi ng on the bas is of for ecas ts is a w em er ge s— the ability to tr ansfer insig ht. co mes fr om co ws. Th ere fo re, co ws ar e the lo t lik e tr yi ng to dri ve a ca r by looki ng This ability to see simila r st ruct ures occur - ce ntra l accumu la to r in the model—t he does it thr ou gh the rearv ie w mirr or . When ring in div erse sett ing s is refe rr ed to as lates ov er time, as num ber of cow s accumu work best? Wh en the roa d is str ai ght and ,” and the str uc tur es “g ene ric thinking co ws are bo rn , ma tu re, and becom e mi lk the re are no obstacl es in the wa y. Whe n the mse lves ar e ref err ed to as “ge neric co ws (s ee “Milk Produ cti on Mo del ”). do es it fa il? The re st of the time! When st ruc tur es. ” Dep en din g on the scop e of our st udy , usi ng a fo recas tin g mode l, you onl y rea lize For exa mp le, if we take the “Mi lk we may be in teres te d in rep rese nti ng the yo u ha ve missed a tu rn once you see the Produc ti on Mo del ” and subst itut e “hi res” lif ecycle of all cow s, or just mil k co ws. In cli ff’s edg e behi nd you and feel the sens a- fo r “b irt hs, ” “tr aine es” for “ca lves,” and thi s cas e, we will focus ou r att enti on on the ti on of fre e fal l hit yo ur stom ach. “sa les man age rs” for “mi lk cows, ” we can flow of cows fr om birth thro ugh maturi ty Fo rec asti ng pr ov ide s very littl e insi ght tr ansfo rm the milk co w model into a mo del int o the milk cow acc um ulat or . The an nual in to what actual ly pro duces the ob serv ed tha t ca n be used to explore the st ruct ur al milk pr od uction is the n deter min ed by the be ha vi or. Co nse quentl y, it al lows us to fo rc es tha t inf luence annual sal es (see “Sa les num ber of milk co ws at any on e ti me and anti cip ate and react to ch ang es onl y if they Gr owt h Model ”). The sam e gener ic the amou nt of milk per co w. Of co urse, do no t devi ate too much from pas t beha v- re sour ce devel opm ent st ructu re under lies the re ar e man y oth er fa ct ors that af fect io r. Mo de ls , on the ot her hand, cap tur e the bo th mo del s. Al though we may deba te milk pr od uction , such as foo d su pp li es, stru ctur al fo rces at wo rk and are the ref or e whet her it takes lon ger to produ ce a milk milk de man d, and dai ry farme rs . Thes e less sit ua ti on- depe nden t. To co me back to cow or a sa les ma nag er, we ca n both agr ee facto rs could also be added to our diag ram the vice pres id ent’s ques ti on, str uct ura l is fun - tha t the st ructur e of bo th processes in th e for m of add it iona l accu mul ato rs and thi nk in g prov ides a mor e sta bl e basi s of dam enta lly the sa me. flow s. • und ersta nd ing th at will last even thr ough The resul ting model can then be sim u- For further read ing aboutstruc tur alth ink ing and the ti mes of turbul ent ch ange. late d on a comp uter to see how an nu al oth ercr itical thinki ngskillsinc ludedunde rth esys tem s behav es over ti me . To cre - mil k production thi nki ngumbrella,seeBa rryRich mo nd’s The THINK ING GENERIC Thi nking in Sys tem s Thinki ng: Seve n Ess enti al ate our hy po th eti cal epi demi c sce nario, for (Pe gasusCo mm unicat ion s, 2000 ). Skill s SKI LLS exam ple , we woul d si mpl y put zero for the If we be gin to view the wo rld thr oug h a st oc k of cow s. In th at even t, th e an nu al st ructu ral pe rspectiv e, anot her benef it mil k production wou ld al so eq ual zero. Bec ause this mod el is tie d to the str uc tu re of the sy stem , not ju st hi st ori cal data, it woul d no t have to be thr own ou t even if all G R O W T H M O D E L S A L E S of the cows sud de nl y di ed . Training Time Maturation_Time OF EXPLANATI ON LE VELS We liv e in th e world of ev ent s. As a resul t, we encou nt er and navi gate thro ug h th e ra pid s of life on an even t-b y- even t bas is. But this does not mean th at we mu st act on Calves Milk_Cows each eve nt as if it we re an isol ated occur - Trainees Sales_Managers Maturation Births MC_Deaths re nce . We ca n lo ok at patte rns of behav io r Training_Rate Hires SM_Quits ov er time an d tr y to gl ean lesso ns fro m the pa st th at will improve our ab il ity to han dle Annual_Milk_Prodn. pr es ent situa tion s. Tha t is the purp os e of Annual_Sales Milk_per_Cow for eca stin g mod els. Sales_per_SM For eca st ing mod els, like the eco no mist ’s If we replace the names of the variables in the “Milk Production Model” with those listed above, milk pr od uction mod el descr ib ed ab ov e, we can create amodel that explores sales growth. The same gene ric resource dev elopment att empt to pr ovid e infor mati on ab out the st ructure can be used to de scribe both processes. futur e by lo oking at the past . Bu t in many 2 9 THI NK IN G TOOLS WWW.P EGAS USCOM. COM SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

32 O O O X L B T M U L A C C U A T O R S : B A T H T U B S , R E B S E V E R Y W H E B A T H T U D VS. ER HEA SHOW took he n’s the last time you actually ca n lea d to dis as tr ous result s. W BA THTUB THIN KIN G Whe n Jus t-in-T ime (JI T) ma nufa ctur - a re al, honest -t o-go odness bath? If yo u are like most people, it ha s pr obably ing fir st hit the U.S., for exa mp le, many Ta ki ng sho wers dis co nnect s us fro m exper i- be en qu ite a whi le. We live in the wo rld of comp anies im ple ment ed it us ing a showe r- enci ng one of nature ’s mo st ba si c str uc - . Yet, qui ck showers and instant breakfasts hea d pe rsp ectiv e. The basic conce pt of JIT and ponds are tu res— accu mul ato rs. Lakes it wasn ’t too long ago when tak ing baths is to ma na ge a stea dy flow of ma te ria ls accum ul ator s of vari ous wa te r flows. Glo ba l The wa s pa rt of our normal daily routine. thr oug h a fact or y with minim al accum ula - warmi ng has bee n attri bu ted to the cumula - sh ift from bat hs to show ers mark ed a far tions of inve nt or y at ea ch ste p. Ma ny com - ti ve eff ects of burni ng fo ssi l fue ls. Plants are in our think in g than mo re de eper change pa nies tha t ins tit ute d JIT tr ied to accum ul ator s of energy and nutr ition. me re ly a cha nge in personal hyg ie ne ize their own accumula tions by minim Di sp la cem ent, vel oci ty, and accel er at ion can tha t the ir supp lier s provide dem anding hab it s. be rep rese nted in ter ms of accum ul at or s. the m wit h ma te ria ls just when the y Wh en we ru n the bath water, we ca n Tha t is, di spl acem ent repr esent s the accu - in the vis ual ly see th e water accu mulating neede d the m and not any sooner. mu la ti on of past ve loci ty, and vel oci ty is an bs an d Acc umula tors”). We The pr ob lem with the above ap proa ch, tub (s ee “Bathtu accum ul atio n of past accel er at ion. kn ow we have to kee p an eye on the wate r of cour se, is tha t the flow of ma te ria ls ha s , we are If we us e showe rhe ad thinking level so it won ’t ove rflow . When we take to accum ulat e somew her e, and it was ac cu - les s con scio us of acc umula tions . Flow s of mula ting in the sup plier s’ war ehouse s. The sh owers , howe ver, the ac cumul ati on proc ess mat erial s suc h as wate r, fuel, or ene rg y JIT flo w wa s accomp lis he d by shifting is virtu all y el iminated . Water flows out of the si mpl y “go away ” so me wher e. But from a the sh ow erhe ad , ove r our bo die s, an d out tions to supp lier s, se ve rely str ain - ac cumula bat htu b—o r syste ms—p erspect ive, the re is the dra in . Wh ere doe s the water go? We ing the rela tionship be twe en sup plier s and no “away. ” Ever ything accum ula tes som e- would ma nuf act ur er s. Ba tht ub thinking ha rdly gi ve it any th ought. wh er e. Forge tting about tha t “som ewher e” hav e hig hlig ht ed the fact tha t unle ss the ent ire flow from raw ma te rials to fina l A C C U M U A N D B S B AT H T U L AT O R S cust ome r wor ked tog ethe r, the re would be able accumula tions for someb ody undesir in the sy stem. Faucet INV ISI BL E BAT HTUB S Time s the la st tim e you ac tua lly le t a When’ Accumulator (bathtub) ba thtub over flow? Pr ob ably not in a long tim e. Of cour se , we all know not to let the Time wa ter run indef init ely, be ca use the tub ha s Bathtub Flow a lim ite d ca pacity . The tub ’s dimens ions (faucet flow) ar e obv ious and so is the rising wa ter line. But sup pos e the bat htub is inv is ible , and so The core building blocks of dynamic thinking tools are reinforcin gand balancing loops. The analo - gous elements in the struc tural thinking set of tools are acc umu lators and flows. An accumulator is the wa te r once it lea ves the fa uce t. And (or stock) is represented by arectangular box, and the flow (or rate) is represented by apipe with sup po se you are not in the ba thr oom to adi rectional arrow, avalve, and acircle. The circle and the box each contain atimeline graph as ke ep an eye on the tub —you are off avi sual reminder that the dy namics of the two are intim ately connected. For example, the con - stant flowfrom the circle to the box as indicated in the diagram must produce the straight linear ans wer ing phone ca lls and dea ling with rise in the water level. No ot her behav ior is pos sible for that structure as it is drawn. the lat est cr isis at the of fic e. How will you 3 0 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

33 ful l or kn ow whe n the ba thtub is getting lea ds to st ill hig her sa vings (see which s you withdr aw mor e than drawal s—unles alre ad y ov erf low ing? tor s “Fr om Loop Dia gr ams to Accumula you ar e ear ning in inte rest, the account Flo ws ar e ea sy to kee p trac k of bec ause and Flows” ). If we start making with - bal an ce nev er go es down. Lik ewis e, if the the y in volv e ac ti on, and actio ns are easy to dr awa ls , the bala nce goes down and int er - - st ress “w ithdra wal” ra te (cop ing mecha me asure —ho w many pro ducts to ship, est pay me nts decr ea se, but sa ving s does not nisms) ar e no t ex ce eding the stress “int er - ho w man y peopl e to hire, for example. decr ea se . It st ill incr ea ses bu t at a es t” rat e (st re ss ful events ), then the best So me accum ula ti ons are also ver y visible , decr ea se d ra te.” So und confusing ? That ’s you can do is le arn to liv e wit h the highe r inve nto - su ch as or der bac klo gs or bulging wher e the accum ulat or and flow dia gr am st ress leve l. From the accum ulat or persp ec- ri es . The re are, how ever, many ac cumula - ca n he lp you act ually vis ual ize how that tiv e, the high- lev er age ac tion would be to tio ns tha t are not ta ngi ble but nonethele ss loop wor ks in te rm s of the flow of mone y or elimi - “c lo se the acc ount” by reducing possess the same behav - ve ry rea l. These into and out of the account. • nati ng the rea l so ur ce of str ess. iora l cha rac terist ic s as physica l ac cumula - RAMS VS . LOO P DIAG to rs and flo ws, but they are like invisible A C C U M U L AT O R T R E S S S AC CUMULAT OR S AN D ba tht ubs —we ca n never tell for sure FLOWS wh et her they are overflowing or not. Level Stress If cau sal lo op dia gr ams and sy stems YIN G ID ENTIF arc he type s ar e such power ful tools, why do UM UL ATIONS ACC Time we need to bothe r with accum ula tor s and flow s? Bo th tools hav e the ir unique le bath - So how can yo u lo cate the “invisib R st rength s. Tools like sys tems arche types ny? For ev ery tu bs” lur ki ng in yo ur compa o Time capt ure and com municat e dy na mic issue s flow (ac tio n, decis ion , polic y), try to fig ure Productivity in a conc ise wa y, but the y do not pr ovide a out wha t, if anything, is accumulating Events Stressful s det ai led repr ese ntatio n of the str uctur e tions of tho se and wha t are the implica Work Pressure prod uci ng the dy na mics. accum ul ati ons . Capacity s o Workload s to Handle Th ere are ca se s when tracing thr oug h outstr ips For exam pl e, as workload Workload a loop diagr am can be confus ing. Fo r beco me exc es - cap acit y and wo rk pressures Incr easing work pressure can lead toan and inte rest form a exam pl e: “S avings ), you siv ely hig h (see “St ress Accumulator” incr eased number of stressful ev en ts,wh ich re in for cing lo op wher e highe r sa ving s bal- sh ou ld quest io n whether those pr ess ure s adds to the accumulation of stress . an ce leads to hig her inte res t payments , sim pl y come and go or whether the ir effe ct s ar e accum ul ating somehow . For ex amp le , ex tra pressure may gene rate L O O P D I A G R A M S T O F R O M mo re str essful event s, which will ac cumu - A C C U M U L A T O R S A N D F L O W S High lat e in to inc rea sing lev els of stress. st re ss leve ls wi ll then lead to low er pr oduc - Savings Balance reduces wo rk cap acity tiv it y, which further This and lea ds to more stressful events. s Savings re in fo rci ng lo op of accumulating stre ss is Time int an gib le , yet all too re al for many people . R R If you loo k at th e sit ua tion from the acc umu la to r view po int and trace out the Interest Time s s Earned Interest s clear why typ - re in fo rci ng lo op , it becomes Rate Interest ica l stre ss reduc ti on effo rts do no t wor k ve ry well. Eac h round of stressful eve nts pr odu ce s mo re st ress, like compound The reinforc ing loop of savings and interest can be represented as acausal loop diagr am (left) or as an accumulator and flow diagram (right), where you can visualize the flow of int ere st int o And coping int ere st in a sa vi ngs account. savi ngs. with - me cha nisms are like savings 3 1 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

34 O X L B O O T M E N T : M U L A T I O N M A N A G E A C C U R A T ” N G A V O I D I T H E “ P A C K S Y N D R O M E NATO MY OF A ac cumula to r ma na ge men t. he re is a st ory abou t a tri via “pa ck T ACC UM UL AT OR ra t,” a man who had spent his RATS PACK AN D NOMA DS MANAG EM EN T He knew en ti re life mem oriz ing trivia. Li fe can in som e way s be viewe d as a ba seb all stati st ic s of every player in the his - A typ ica l acc umu la tor mana gement struc - neve r-en ding task of ma nag ing va rious tory of the maj or lea gue . He had me mo - tur e (AM S) has the follo wing elem ent s: the acc umu lato rs . Our pantr ies , ref rige rat or s, ri ze d th e tit les, direc tors, and ac to rs of it ions, Deplet ions, Ac cumul at io n, Acquis ch eck in g accounts , and clos et s are am ong hun dred s of movi es. He knew th e name of Des ir ed Accu mula tio n, and a Co rrect ive the man y acc um ulatio ns we ma nag e da ily. ev ery tele visi on show that had ever aire d. Ac tion (see “Acc umula tor Mana ge ment ”) . On one end of th e accum ula tio n man- But one day he found hi mse lf in an In addit ion , the re is alm os t alwa ys som e ag eme nt spec trum is the pa ck ra t who ho w awkwa rd pre dicame nt—no matter dela y be twee n the Co rr ect ive Act ion and the thr ow s no th ing awa y. On the ot her end is ha rd he tried , he cou ld not memo ri ze any Ac qu is it ion, beca use it takes tim e to act ua lly a virt ue of ow ning the “no mad” who makes more tri vi a. He had final ly taxed the limi ts me mor iz e dat a or cle ar out the clo se t once no mo re than what can be pa cke d int o one of his rot e me morization cap acity . Altho ugh we ha ve dec ide d to do so. suit cas e. In betwee n these tw o ext rem es lies he had worke d har d at acqu iring his sto ck The accu mu lat or ma nage ment st ruc - the ma jor it y of the po pula tion who is con - of trivia thr ou gh out his li fe, he had ne ve r tur e is a gen er ic st ruc tu re that ca n repr e- stant ly strugg ling to ma in tain the ri ght ba l- con sid ered how he might go ab ou t de ple ting sen t a wide range of bu si ness set ting s an ce bet wee n acq uis it ion s and de ple ti ons. it. He ha d not learn ed the fun dame ntal s of where accu mu lat ion mana gemen t is AT O R A C C U M U L I N S U R A N C E B U S I N E S S A S M A N A G E M E N T A C C U M U L A T I O N M A N A G E M E N T Retention Desired Accumulation Accumulation o s s Gap B y a l e D Corrective s Lost Policies Claims Adjusted Loss Events Policies New Action Depletions Acquisitions Settled Claims Claims Pending Policyholders Accumulation s y B a o l e D Desired Policies Lost Gap s Accumulation Corrective Investments Action s Cash Out Cash In can be viewed sim - Accum ulator management The insurance business can be mapped into arelatively simple diagram that high ligh ts the ma jor pl yas abalancing loop with delay (top). A accumulato rs and flows. If we assign numbers next to each accumulator or flow ind icat ingthe struc tura ldiagram (bottom )reveals that the resources devoted to it, the diagram can help reeva luat ethe organ i- per centage of organizational flow scontrolling the accumulation are acquisi - zati onʼs current emphasis. tions and deplet ions . 3 2 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

35 ING THE “BE ER MANAG im port ant. For exam pl e, th e ins uran ce na tions rathe r than co nnec ting the dyna mic s GA ME” ba ck to their ow n de cision -ma king. In fa ct , busi ness can be map ped into a rel at ivel y the wide osci llations in in ven tory are act ually si mple di ag ram by focu si ng on the basi c In a pr oduc ti on dis tribut ion syst em ga me ge ner ate d by the deci sions they ma ke . ac cumul at or s and fl ows (see “I nsu ran ce fond ly know n as the Beer Ga me, pa rt ici - en t”). Busi ness as Acc umu lation Managem pan ts ar e giv en the task of mana ging thei r AVO IDING THE “P ACK Insuranc e rev ol ve s aroun d managin g two ow n inve nt ory (a ccu mula tio n) of be er. Ea ch RAT ” SY ND RO ME mai n ac cum ul ators —pol icyhol ders and tea m is compos ed of fo ur play ers linke d If you wa nt to avoid the “p ac k rat ” sy n- invest ments . toge the r in a struc tur e simi lar to tha t repre - dr ome , you nee d to mana ge the whole If manag ers ass ig n a number ne xt to se nt ed in th e AMS dia gra m (s ee “Supp ly tor mana geme nt st ruct ure and ac cumula eac h acc umu lator and fl ow in th e diagr am Li ne and Del ay in the Bee r Gam e” ). With in not jus t focus on one piece of it . The obs er - to repr esent th e pe rcentage of organi za - that tea m, eac h par ticipa nt mus t ma ke va tions about the diff icult ies of mana ging ti onal resou rce s de voted to eac h, the dia- or de ring de cis ion s in order to main tain his the Bee r Ga me sug ges ts tha t you should areas recei ve the gram can highl ight which des ir ed lev el of in ve nt ory. think thr oug h the following quest ions larg est orga ni za ti onal foc us. Thi s exerci se Ac co rding to MIT pr ofe ssor John when co nf ront ing a typ ica l accum ul ator can point ou t any weak nesses in the cu rren t St er ma n, wh en par ticipa nt s try to manag e ma na gem ent sit uat ion: (1) Wher e are the or ganizat s—f or exam pl e, ional emphasi ac cu mulat io ns in the Beer Game they usu - sup ply line dela ys and how are they cha ng - sp ending to o li ttl e ti me tryin g to ret ain all y run in to thr ee co mmon prob lem s. Firs t, ing? (2 ) Wha t fact or s ar e de ter mining ways in curr ent pol icy hol de rs —an d reveal they typi cally unde restim ate the true le ng th wha t Desir ed Accumula be ? (3 ) tion should whi ch the comp an y can better ser ve its of the de lay fr om the time they orde r to How do cur rent policie s and decis ions fee d cust ome rs. whe n they re ceive the beer and then ove rad - ba ck into this sys tem to pr oduce the results jus t the ir orde rs —ev en whe n they are give n LINE S AND SU PPLY we ha ve obs er ved? The accu mula tion full info rma tio n abo ut the supply line dela ys . DE LAY S ma na gem ent st ruct ur e dia gr am is a use ful The y do no t app ea r to reco gnize tha t their If we ha d direct and immedia te contro l st ar ting point to beg in addr essing the se or de ring de cis ion s affect the leng th of the ov er all th e el em ent s in the AMS diagr am, quest ions. • su pply li ne de la y— tha t is, th e more they be simple: ma nagi ng accumul atio ns would or de r, the lo nge r it takes to rec eive the bee r. Furthe rReading:“Mod eling Mana ge rialBeh av ior: We wou ld cal cul at e the dep letion rate , set In addit io n, he found tha t whe n pe ople Mispercep tionsofFeedbackin aDyn am icDe cision ou r desir ed ac cum ula tions accordingly , find it diffic ul t to de termi ne th eir opti ma l Mak ing Exp erime nt, ”byJohnD.St erm an, Man ag eme nt Science , Vol .35, No. 3,Ma rch 198 9. and impl ement act ions that will immedi - inv ent ory le ve l, the y simp ly anc ho r the ir ate ly re su lt in acqui siti ons. In our ho me des ir ed inve nt ory on the ini tia l inve nt ory life we alre ady prett y much follow this and adjus t fr om the re. pa tt ern . Fo r ex am pl e, we plan our me als, This findi ng hig h- S U P P LY L I N E A N D D E L AY of food de ci de on an appro pria te amount light s the mo re gen- I N T H E B E E R G A M E to ha ve on hand, figure out how long it er al tende nc y peo ple wi ll be be fo re we run out of certain sta ples, hav e to an ch or on Delivery s Delay and go to the gro cery store as nee ded. pas t goa ls or st an- R Unfo rtu nat el y, thi ngs are not that str aight - dar ds ra the r tha n Orders Shipments fo rw ard when we mo ve into the or ganiz a- se arc h for be tte r one s. o tio na l cont ext . Th e th ird obs er - Inventory Supply Line One of th e most cha llen gin g as pe cts of vat ion is tha t peopl e Deliveries s B manag in g accumula tions with in or gan iza - gen er ally poi nt to fac - o Desired Corrective Action Gap ti ons is capt ured in one word—de lays . tor s ou ts ide the sys te m Inventory s in g an d ch aract erizing Identify the nat ur e as be ing re spon sib le and sou rc e of dela ys oft en plays a crit ica l role system in the Bee rGa meis The str ucture of the inventory management for the in st abi liti es the nature and so ur ceof si mi lar to the AMS diagram. Understanding tions effect ive ly. A big in ma nag ing accumula th ey obs er ve in th e as the supply line delay above— oftenplays del ays in asystems—such is that we usu ally ha ve part of the problem gam e. Th at is, pe opl e acr itical role in managing accumulations without overcorrec ting . ve ry litt le cont rol over the supp ly lin e dela y. of fe r open loop expl a- 3 3 THI NK IN G TOOLS SYSTEMS WWW.P EGAS USCOM. COM PEG ASU S COMMU NICA TIONS , INC.

36 O O O X L B T O R S D E L A Y S : U M U L A T A C C G U I S E I N D I S magin e a new man ag er at a beef tu re th at produ ces the de lays, it is difficul t tim e, and for prod uction it is the ma nuf ac- I pa ck ag ing plan t who knows to kn ow how lon g to wai t bef or e we sh ould tur ing cy cl e time. Tryi ng to short en the not hi ng about th e birt hing proce ss of calv es. On his ta ke furt her act ion . inh er en t tim e del ay by pu shing things thr ou gh the ac cu mula tors fast er can wre ak fir st da y, his worker s show him a newb or n Wh enever th er e are si gnifican t del ays havoc on the sy st em. in a sy ste m, you can bet that the re are accu - ca lf . Th e dollar sig ns go off in his hea d as The agi ng ch ain rep resent s a mult i- he cal cula tes: More calve s mean mo re beef; mul ator s invol ved . In som e ca ses the accu - mo re bee f me an s mor e sales; mo re sales mul ator s are less obvi ou s than in oth er s, st ag e proce ss wh er e “st uff ” (idea s, pr od - uct s, ca lves) movi ng th rou gh the syst em me an mor e profit s. He points to the mot her but the y alm ost always pl ay a rol e. In the ca se of th e pack agin g pl an t, the accu mu la- “I want you to get two more und er goe s var iou s st ag es of developm co w and barks, ent . ca lve s out of th at sucker by Monday mo rn - Each st age ca n be repre sen te d by an accu - Wh en you to r is th e cow (or st ock of cows). mu lat or , wh er e the st uff “a ges” bef or e ta ke a showe r, the del ay in get tin g hot in g, and th at ’s an ord er!” will find a way movi ng on to the next st age. The “ag ing ” wa te r is due to accu mu lat ion in the len gt h Of cou rse, the workers tim e at any stage can var y. to fulfill his re qu est , eith er by brin ging two of pip e from the hot water hea ter to the ca lve s from anot he r part of the pl an t or per - sh ower . Eve n though th e wat er “f lows” ha ps sla ught erin g th e mother cow to pro - STRUCTURE CHAIN AGING GENERIC du ce th e ext ra pou nds of beef. The Inflow Th e“aging chain” structure represents amulti-stage wor ker s will have su cces sfully ex ecut ed proce ss where “stuff” moving thro ugh the system r will co nti nu e th eir tas k, and the manage undergoes various stages of develo pment (indi - Advance Rate 1 cated by accumulators). contr ol the pro - to beli ev e that his orders du ctio n cy cle of calves. Advance Rate 2 1 Stage The stor y is ob viou sly far -fetched . No th rou gh th e a calf in one wou ld ex pect a co w to produce Outflow 2 Stage pip e, th e vol ume of one week en d. Bu t how do we kno w th at wa te r in th e pipes can be equa lly rid icu lou s deman ds are no t bei ng viewed as an ac cum ulat io n of cold ma de eve ry day on process es whe re we Stage n wa te r being pus he d throu gh the pi pes by ha ve le ss un de rst an ding of the tim e dynam - th e sub sequ en t hot wate r. ” equiv alen ts in ic s? Ar e there su ch “calving PR ODU CTION SY STE MS The “ag ing chai n” str uct ure is th e g, for examp le, where arb i- ma nufacturin One wa y to get a bet te r ide a of the dela ys name give n to a wh ole cla ss of pr ocesses tr ar y qua rterly sa le s tar gets given to inv est - involve d in a sys tem is to cr ea te a st ruc - th at in cl udes th e bi rt hin g cycl e of cows, me nt an aly sts tran slate into marc hi ng tur al ma p. A typica l pro duction sy st em is prod uc ti on- di st ribu tion syste ms, and the line? orde rs fo r th e pr od uction Chain” dia gram. in the “Pro duction shown sp re ad of inf ec ti ous di se ases (see “G en eri c AGING CH AIN In the dia gr am, dif fe re nt sta ges of produc - Ag ing Chai n Stru ct ur e”) . Th e key pr in ci - ST RUC TURE - tion pr oce ss are rep rese nted by accumula ple in the agi ng ch ai n is tha t th er e are Why is it tha t whe never we want som e- ency , this ional consist tor s. (F or dimens del ay s in the sy st em th at depen d on the thi ng right awa y, it seems to take foreve r? diag ram rep resent s account ing num ber s inh ere nt natu re of that sy st em, and th ose Then when we do get it, it oft en is more and no t the act ua l phys ica l stuff moving del ay s can not be sh ort ene d except wit hin than we eve r wante d? Ch an ces are , del ays tor and thr oug h the sy stem. ) The ac cumula so me narr ow bou nds . For cal f pr od uct ion pl ay ed a large par t in ou r mis -tim ing. We flow dia gram is ver y much like a proc ess th at de lay is the gest ati on pe ri od . For new bu t un less reali ze that “thin gs tak e time,” flow cha rt, showing ca n how production ide as it may be an in cu bat ion per iod , for we have a cle ar un ders tan din g of the st ruc - be ma pped out as a se rie s of st ag es . Each new prod ucts it may be th e devel opme nt 3 4 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

37 st ar ts and pus hing or de rs into produc tion acc umu la to r, in eff ect, adds a delay to the le ft wa it ing for th ei r goo ds . One re spons e to will exa ce rb ate the full ut iliza tion loop sys tem. ti on sys - such a ga p is to go int o the produc (R 2), le ading to fur ther dela ys and mor e If we wa nt ed to add grea ter de ta il, each tem an d expe dit e some of the more “impo r- pr essur e to ex pe dite (R4) . acc um ul at or cou ld be fu rthe r bro ken out tant ” orde rs and/o r push more produ ct ion , the int o sma ller stag es . For example PRODUCTION CHAIN Wo rk-i n-Pr oces s accu mulator may be Production Requests ach stage of the production chain flow dia - E broken in to vari ou s produc ti on sta ge s ram has its own time delay or cy cle time g su ch as as sembl y, paint, bak e, test, and associated with it. In many cases, the length Schedule Rate of time changes as afuncti on of the system— ins pect io n. Each stage has its own point that is usually not captured in a a tim e de lay or cyc le time associ ated with it. ypical process flo wdiagram. t Production Starts Production Request Th ese ti me del ay s have an important R1 s Backlog o as pe ct: They usua ll y do not stay con - Scheduling Production Production oad L d in a st an t— a po int that is not capture ueue Q typ ic al pr oc ess flo w chart. In the firs t s ork in W Shipments st age , for exa mp le, a rising ba cklog of pr o- requ es ts in to the sys - Process Full R2 inc rea ses the scheduling duc ti on requests tem. The int ent is to get Utilization o Finished Down dec reases the rate at whic h th e loa d, which mor e produc ts thr oug h th e Goods Time s This kee ps the re que sts ca n be scheduled. ch ain, bu t the se act io ns ar e the ba ckl og high, furt her exa cerbating like ly to pr odu ce the opp os it e In gene ra l, the aging cha in st ruc tur e sch edu le lo ad pro blem (R1). Onc e a sc hed - effe ct. te lls us tha t ther e are structur al lim its to can ac tu ally ule ge ts behi nd, the system Gi ven tha t the sy ste m wa s alr ea dy run - how fast you ca n for ce a sy st em to re sp ond. to fall fur the r re in fo rce the tendency ning at full tilt, the expedit ing act ions are ch ang e the Unle ss you can so mehow be hin d. reinf or cing likely to cr eate additional inher ent dela ys built into the va rious st ag es Si mil ar ly, at the work- in- pr oc es s (WI P) on product io n sou rc es of de lays . Pushing of the sy stem, the best exp ed iting ac tion st art s, the sta ge , if we load up on prod uction re qu est s will kick in the sche duling loa d one ca n take may be to simp ly do noth - WIP inve nt ory rises. As th e WIP ri ses, the loop (R1) , which will furt her dela y ship - ing— and wait . she er am ou nt of ext ra “st uff” in th e wor ks men ts and intens ify pr essur e to exp edit e • can slow thin gs dow n. Anot her cons eq uen ce (R3 ). Si milarly , re ar ra ngi ng the production of pumpin g up th e syst em with add it ion al WIP is th at it creat es pres sure to E X P E D I T I N G L O O P S run equipm en t at full ca pacit y, whi ch can lea d to increa sed down s Production Requests ti me (no time for main tenance R3 work) an d low er yield s (hi gher Schedule Rate s s sc ra p rates). All th is will ult imat ely Customer Shipment Expediting Orders Gap le ad to lower prod uct ion an d an s o Production ac cumu lati on of more WIP (R2 ). Production Starts R4 Request R1 s Backlog o JU ST DO DON’T Scheduling Production Production Load Queue G, ST AN D SO METHIN TH ERE! s in Work Shipments Process Full The “Prod uc ti on Ch ain” diag ram R2 Utilization o do es no t show the compl ete pi cture Finished Down Goods Time s of wh at goe s on in a typic al manu - fa ct uri ng se tti ng (see “Exp edi ti ng Efforts to expedite ord ers can backfire by actua lly lengthe ning the average production delay and reinforcing Loo ps”). Wh en shi pmen ts fall sho rt the nee dto exp edite (loops R3 and R4). of cus tome r orde rs , cu stomers are 3 5 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

38 O X L B O O T S - S H A P E D G R O W T H A N D T H E L A W I S H I N G R E T U R N S O F D I M I N ost of us ar e fam ili ar with th e story discov ery hits him— the La w of Suc ce ss” archet yp e. At the hea rt of bot h M of Sir Is aac Ne wton sitting unde r- Diminish ing Returns . st ruc tur es is a pai r of rei nf or ci ng and bal - neat h an appl e tr ee and “dis coveri ng” the anc ing lo op s tha t int er act to prod uce the S- HE LAW OF T law of gravi ty whe n he saw an appl e fall shap ed patt er n. DI MI NISHIN G RE TU RN S be en (some say on his he ad). Had Newton DIF FU SION DYN AMICS The phen om ena of dim inishi ng re tur ns— an ent reprene ur, he might have discover ed when more effo rt yi el ds fewer resul ts —i s The basic Bass Dif fusion Mode l is usua lly anot her law. With apple sales—an d pr of - ub iq ui to us. Oi l reco ve ry and min ing op er a- giv en as a set of equa tions : have shak en the it s— in mind, he might ti on s ex hi bit this be havi or . Co mp ani es expe - tree vig orou sly , cau sing more ap pl es to f(t) = dF (t)/dt = [p+qF( t)] [1 – F(t )] ri ence rap id ne w pr oduct sa les fol lowed by dro p to the grou nd . The harder he sh ook, ≥ 0 p, q decr ea sing dem and. At a per so nal level , we fal l, mean in g more the mor e app le s would p = coe fficie nt of adve rtis ing, see tha t wor ki ng longe r hour s, jo gg ing sal es and more profi ts. After a whi le, how - q = coef ficient of int era ction. mo re mile s, and eati ng le ss food lea d to ev er, eac h sh aki ng would fe wer produce di mi ni shi ng retur ns in pr oduct ivi ty gai ns , and fewe r ap ples . Int egrat ion yie ld sanS-s hap ed hea lth ben efi ts, and wei gh t loss. We can al most pic ture the scene : Sir gro wthcur ve ofdiffu sio n. The Law of Dim in ishi ng Ret ur ns can Is aac wip es the sw eat from his bro w, then be co nsid ere d the law of gr avi ty fo r the Al thoug h the equa tions may offer an down the tena - clim bs the tree to knock busi ness wor ld . Laun chin g a mar ket ing ele ga nt way to repr esent such dyna mic s, pe rc he d, cio us fe w appl es left . Precariously ca mp aig n, for exam pl e, is like th e tr aj ector y mo st of us don’t view the wo rld as a set of he strai ns to rea ch one of th e las t re main - of a ca nno nbal l— the retu rns cl imb hi gher equa tions . Fr om an accum ula tor and flow s ing ap ple s. The li mb gives way, and he and hig he r, unti l the “fo rce” of di mi ni shi ng per spe ctive , we see diffus ion dyna mics as a , anoth er fa ll s. As he la nds on the ground re turn s ki cks in and fr om a pool of pot ent ia l flow of people pul ls the rate of ret urn adopt er s to ado pt ers (s ee “B ass Dif fus io n B A S S U S I O N M O D E L D I F F — dow n. When tra ced out Mode l— A St ruc tur al View point ”) . R A L V I E W P O I A S T R U C T U N T ov er ti me, the cum ul a- Ins te ad of p’ s and q’ s, we talk about an tiv e ret urns of the mar - out h adve rtis ing effe ct and a wor d-of-m ke ting ef fort pro du ce an effe ct . This str uc tur al vie w mak es the S-s hap ed cur ve. dyna mi cs mor e explic it , close r to the wa y we The Law of act ua lly think about and exp erience the B R Time Time Di min ish ing Ret ur ns is wor ld. Adopters Potential Adopters esse nti ally ab out sat ur a- Fadd ish pr oduct s, such as hula ho ops tio n eff ec ts— reachi ng and Cab bag e Pa tch do lls, us ually exhib it Time the lim its of a par ticu lar cla ssic S-sh ape d growt h. Rem emb er the Rate Adoption sy ste m. The cha ra cte ri s- sudde n popula rit y of “pet rock s” in the la te Advertising Interaction tic S-sha pe d beha vi or of 1970 s? The y wer e just pl ain old roc ks this pr oc ess ca n be pro- repa ck age d in a pet -c arrie r st yle bo x. At The Bas sDiffusion Model is an example of S-shaped dynamics. duc ed by two di ffer ent fir st, dr ive n by strong adve rtising , the y we re Fr om astructural perspect ive, it can be represented by aflow ofpeo - struc ture s: the wel l- vie wed as a po pula r no velt y it em . Sa les pl efrom apool of Potentia lAdopters to Adopters. If we substitut e Potential Pet Rock Owners for Potential Adopters, Purchase Rate for kno wn Bas s Di ffusi on be gan to grow , incr ea sing the dem and for Adoption Rate, and Pet Rock Owners for Adopters, the sam estruc - Model and the mor e the rock s and spr eadi ng word-of- mout h tur al diagram descr ibes the dynamic sof the Pet Roc kfad. ge nera l “Li mi ts to endo rsement s. Soon sale s—and the rock s’ 3 6 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

39 BR EA KI NG THE LAW po pul ari ty— be gan to skyro cket. But ev en tu - decr eas es. Thi s lea ds to ca pa ci ty adequacy all y the po ol of pote ntial adop te rs (o r pote n- d lo wer quali ty and produ ces a do wnwar If yo u find your self “ca ug ht ” by the La w of tia l pet ro ck own er s, in thi s case) wa s on custo me r gro wth. That is, pr essure Dimi nishing Ret ur ns , using a st ruct ural dia - dr ained , and the re was no on e left to buy ca pa ci ty cons trai nts eve ntual ly di mini sh the gram may hel p yo u ide nt ify the crit ica l fa c- them . eff ecti venes s of the market ing ef for ts. tor s and find a wa y to bre ak out of it . In a In both the Bas s Di ffus ion Mod el and diffus ion dyna mic s case, fo r exa mple , qua n- CITY CAPA LI MITS th e “Limi ts to Su cce ss ” arch et ype , the S- tifying and me asur ing each of the di ffe rent In gen era l, dimi nis hin g returns oc cur whe n- sh ape d cu rve is pr od uced by a reinfo rcing pie ces of the diag ram may help de cide ev er we hit a cap acity limit. In th e Bass loop coup led wi th a ba la ncin g lo op. The whe ther yo u should try to exp and the poo l Mod el, ca pac ity is the numb er of pe opl e who re inf or cin g loop cre ate s th e init ia l gr ow th in of pot ent ia l adopt ers, se gm ent adopt ers into can ult ima tely become adopters of a pa rti cu - de man d, whi le the balan cin g loo p is gene r- diffe rent cat ego ries, bee f up the advert ising lar pro duct, tech nology, or id ea. The ado p- ally re spo ns ibl e for the diminish ing ret urns. budg et, or pus h on direc t sales effort s. tio n rat e fa lls to ze ro whe n the pote ntia l Th e balan ci ng loops do not sud denly In the mor e gener al cas e of ca pacity lim - ad op ters acc umu lator is depl eted, or, in the “a ppea r.” Th ey are almos t alwa ys pr ese nt it s, brea ki ng out of the dimin is hing ret urns co me s to their pet roc k cas e, whe n everyone from the ver y start . When th e dynam ic phe nome na requ ir es ident ify in g the acc u- sen ses (w hich ever come s fi rst!). ch an ges fr om on e of ris ing grow th to a slo w- mul at or (s) tha t are ope ra ting at or nea r full THE ” ar ch et ype ( The “L imit s to Success ing pac e, the forc e dr iving the syst em ha s ca pa cit y and ca lc ula ti ng the true work loa d SY STE MSTH INKER , December sim ply sh ifted from a re info rcin g to a ba la nc- dem and. Elimin ating any ga ps be tween 19 90 /J an uar y 199 1) is anot her wa y of des cr ib - ing loop. “Sat urat ion ” occ urs in bot h cas es— dem and and capa cit y is lik ely to produce ing the capa cit y limit dynamics that pr odu ce wh et her it is the satu rat io n of a giv en mark et ha rder on mo re result s than si mply pushing beh avior. In a “Limit S-shaped s to Suc ce ss ” ca pa cit y. or th e full ut ilizati on of a specific the reinfor cing loo ps in the sys tem. • str uc ture , a syst em’s pe rfor man ce impr oves ow ing to certa in ef for ts. Bet ter per fo rman ce resul ts in more ef forts , lead ing to fur th er L I M I T S T O S U C C E S S F R O M A R C H E T Y P E “ ” — to im prove men t (loop R1 in “‘ Limits A C C U M U L A T O R S T O Succ ess ’—From Arche type to Acc um ulat ors”) . Over time, howe ver , perfor - Constraint manc e be gin s to pla tea u des pite in cr eas ed o s s some limit eff orts— the syst em has reached Limiting or resist an ce tha t is preve ntin g fur the r Performance Efforts R1 B2 Action im prove men ts (lo op B2 ). s s If we loo k at the “L im its to Succ ess” ar ch ety pe fr om an acc umu lator and flow s Quality pers pec ti ve , we can se e more cl early the B5 B4 st ru ct ures prod ucin g th e un wante d be hav ior. Customers o In a se rv ice organiz ation, for exampl e, service o cap aci ty may be come th e limi ti ng fa cto r if it s Time Capacity do es no t ke ep up with inc re asin g deman d. s Adequacy Customers Lost Customers New in cust omer s will In the be gi nning , growth o Marketing R3 le ad to high er reven ues, in creased ma rket ing, s s Revenues and furthe r grow th in cust omers (R 3). Th is s rei nforcin g loop drive s the in itial gr owt h of Time Time the cust om ers accu mu lator. Service Load Service Capacity As the nu mber of cust om ers gro ws , how ev er , so does th e se rvice load on the The “Limits to Success” archetype (top) is one way of describing the capacity limitation dy na mi cs that pr oduce S-shaped behavior. Looking at the archetype from an accumulator and flowper- co mpa ny. If the serv ice cap ac ity do es not spective (bo ttom), we can see more clearly the structures producing the unwanted beh av ior. gr ow at le ast as fast as the ser vice load , 3 7 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

40

41 PARTIV COMPUTER-BASEDTOOLS COMPUTERMODEL MANAGEMENTFLIGHTSIMULATOR LEARNINGLABORATORY any of the systems we are charged with managing are so dynamically complex M that they are almost incomprehensible. Complex social systems frequently exhibit counterintuitive behavior, where actions that provide short-term relief often result in greater long-term pain. When actions and consequences are greatly separated in space and time, making effective decisions for the long-term well-being of a system becomes extremely difficult. Causal loop diagrams, archetypes, and structure-behavior pairs can help us build a better conceptual understanding of the key relevant structures of a system and perhaps even predict the general behavior of the system over time. “Modeling for What Purpose” (p. 40), however, describes times when we need even greater precision about the ramifications of certain actions at specific points in time. The rest of this section describes how we can translate our pen-and-paper represen- tations into computer-based models that can be simulated, converted into interactive decision-making games (management flight simulators), and embedded in a rich learn- ing environment (learning laboratories). 39 WWW.PEGASUSCOM.COM SYSTEMSTHINKINGTOOLS PEGASUSCOMMUNICATIONS,INC.

42 O O O X L B T P O S E ? M O D E L I N G F O R W H A T P U R BY R RRESTE FO W. JAY al ly can agree) mere ly becau se they ini tia ll y ys tem dyn ami cs doe s not im pose beh avi or . In ge ner al, inf luent ia l sys te m S ic concl usi on s tha t mod els on peop le for the first ti me— dyn am ics pro je cts ar e thos e that cha nge di sag ree wi th the dynam mi ght fol low. mod els are alre ady pr ese nt in eve rything the way peo ple think abo ut a sy st em . Me re we knowle con firm atio n that cur rent belief s and poli - If we divide do . One does not have a famil y or corpora - dge of sy stem s into thr ee ca teg or ies , we ca n illu st ra te wher ein ci es ar e cor re ct may be satis fying but tio n or cit y or cou ntry in one’s hea d. In ste ad , tions about on e ha s obser vation s and assump hard ly ne ce ssar y, unle ss the re are dif fer - lie the st reng ths and we akness es of menta l en ces of op inio n to be resolve d. Chang ing tion mod els (se e “Thr ee mode ls and simula thos e sys tems . Such obs ervati ons an d an d unifying vie wp oints mea ns tha t the Ca te gor ie s of Inf or ma tion”). The top of as su mpt io ns cons ti tu te me ntal mo del s, wh ic h we the n use as a basis for ac tion . the fig ur e rep rese nts knowl edge about re levan t menta l mode ls are being alt er ed. st ruct ur e and policies; tha t is, about the Th e ult im at e suc cess of a system GE UN IF YIN G KNOWLED elem ent ar y par ts of a sy ste m. Th is is loca l dyn ami cs invest iga ti on depends on a clear Com pl ex syste ms def y int uitive solut ions. non-dy e. It des cr ibe s ini ti al id ent if icat io n of an important pur - na mic knowledg Eve n a thir d-o rde r, linea r diff ere ntia l a sys tem inf or mation ava ilab le at ea ch dec ision- po se and objec ti ve. Presumably eq uat ion is uns olv able by insp ect ion. Yet, e, clarify, and ma king point. It ident ifie s who cont rols dyn ami cs mo del wi ll organiz import ant situatio ns in ma na gem ent , eco - give should The model uni fy kn owledge. ea ch par t of a sys tem . It reve als how pr es - nom ic s, me dicine , and socia l behav ior usu - ng sur es and cr ise s inf luence . In decisions pe opl e a more ef fec tive understandi al ly los e rea lity if simp lif ie d to le ss tha n ant system ge ner al, inf ormation ab out str uctur e and that has pr evi - abo ut an import fift h-or der no nline ar dyn am ic sy st ems . rs ial ou sl y exh ibi ted puzz ling or controve po licies is far mo re relia ble , and is mor e Att emp ts to de al wit h nonline ar of te n see n in the same wa y by diffe rent sy st ems us ing ordina ry proc es ses of dy namic It is only pe op le, tha n is gener ally ass umed. I E S G O R T H R C A T E E E de scr ipt io n an d deb at e lead to int erna l nece ssa ry to dig out the inf or ma tion by I N F O R M A T O F I O N inc on si st enc ie s. Underly ing ass umpt ion s using syst em dy na mics ins ight s about how may have be en le ft unclear and cont ra dic - to org anize str uctur al inf or mat ion to tor y, and me nt al mo dels are of te n lo gic ally addr ess a particular se t of dy na mic is sues . Observed Structure inc omp le te . Resu lti ng beh av ior is lik ely to The mid dle of the figur e rep resent s and Policies a be con tra ry to tha t im plie d by the ass ump - ass ump ti ons abo ut ho w the syst em wi ll a tio ns be ing ma de ab out un derlying sys te m beha ve, ba sed on the ob serv ed st ru ct ur e an d Expectations About Behavior stru ct ur e and go ve rnin g polici es. bel iefs are, pol ic ies in the top secti on. These b System dy nami cs model ing ca n be in eff ec t, the ass um ed int ui tive sol ut ions to b ef fect iv e becaus e it bu il ds on the rel iabl e the dy nami c equa ti ons descri bed by the Actual Behavior par t of our un ders tandi ng of sys tems wh il e st ruc tur e and pol ici es in the top sect ion of co mp ensat ing for the unrel iab le par t. The the dia gr am. Th ey rep res ent th e solu ti ons, sy stem dy nam ics pro ce du re unt an gl es sev - ar rived at by int ro spe cti on and debat e and The re are three categ ories of info rmation era l threa ds that cause conf usi on in or di - comp romi se, to the hi gh- ord er nonl inear abou tasyste m: kno wledge about structu re na ry deb ate : unde rl yin g assump tio ns and po licies; assump tions ab out how the sys - sy st em descr ibed in the top pa rt of the fig- tem will beh ave based on the observed struc - (st ruct ure , pol ici es , and par amet ers) , and ur e. In the mi ddl e lie the presumpt ions that ture and policies; and the actual system im pli ed behav io r. By con si deri ng ass ump - lea d mana ger s to chang e poli ci es or lea d beha vior as it is observed in real life. The ti on s in dep ende ntl y fro m resul ting behav - usual di screpan cy is acro ss the boun dary a-a: gov er nmen ts to chang e la ws. Based on expecte dbeha vior is not consistent with the io r, there is less incl inati on for peop le to ass ump ti ons abo ut ho w behav ior is kn own stru cture and pol icies in the system. di ff er on assum pt io ns (o n whi ch they act u- exp ec ted to chang e, pol icie s and la ws in the 4 0 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

43 to p sec tion are alt ere d in an effor t to ab le way . A mo del shoul d link the pa st to men t waits for still highe r ba cklo gs bef ore achie ve assu med impr oved behav io r in th e the pr esen t by showi ng how pr esent cond i- ex pan di ng capac ity, or de rs are dr iven mid dle se ct ion . ti ons ar ose, and ex tend the pr esent int o per - down by unf avo rab le de liv er y dela y unt il sua si ve al terna tive fut ur es und er a var iet y Th e bo ttom of the figur e repr ese nts or de rs equa l cap acity (R 3). The await ed the act ua l sy stem beha vio r as it is obse rved of sc enar ios det er mi ned by poli cy al terna - si gn al for expans ion of capacit y nev er in re al li fe . Ve ry oft en, actual dif - ti ves. In othe r wor ds, a syst em dyn ami cs behavior sa les , is cont rolling co me s beca use ca pacity fe rs subs tant ial ly from exp ected be hav ior. mode l, if it is to be ef fect ive, must comm u- ra ther than pote ntial de ma nd contr olling exist acr oss In oth er wo rds, disc repancies ni cat e wi th and modi fy the prior ment al ca paci ty (B2) . mode ls . Onl y peop le’s bel ief s—th at is, th ei r th at the bo und ary b-b. The surprise Whe n sal es fai l to ri se beca us e of long ment al model s—wi ll det er mi ne acti on. obs erved struc ture and po licies do not lead t may then de li very del ays , managemen Com pu ter mode ls must rel ate to and to th e expec ted behav io r is usua ll y low er pri ce in an at tem pt to st im ul at e mor e ex pla ine d by assumi ng that info rmation imp ro ve ment al model s if the com put er sa le s (B4 ). Sal es incre ase br iefl y but onl y must hav e abo ut str uc ture and policies mode ls are to fill an ef fecti ve rol e. lon g en ou gh to bui ld up suf fici ent add i- • be en in corr ec t. Unj ustif iably bla ming ti ona l backl og and del iv er y del ay to com - ay W. Forr es ter , profe ssor emer it us at th e J ina de qua te kno wl edge about parts of the pensa te fo r the lower pri ces . In addi tion, Mas sa chu setts Instit ut e of Tech nolog y and for me r unc ounted sys tem ha s resul ted in devoting price redu ctions lower pr of it ma rg ins unt il ire ct or of th e MIT Sy stem Dyn ami cs Grou p, is the d fo unde r of the fie ld of syste m dyn ami cs. Si nce hi s mi ll io ns of hours to data gather ing, que s- th ere is no longer eco no mic jus ti ficatio n for re tire men t in 1989, he has be en wor king towa rd tio nn aire s, and interviews that have faile d ex pan sio n (R5) . In such a si tuat ion, ade - gh 12t h gr ad e rin gin g sy st em dyn amic s in to K throu b sc hoo ls as th e basis for a new kin d of ed uca tion . to si gn ifica ntl y imp rove the unde rs tanding qu ate info rm at io n about indi vidual rel a- of sys tem s. ti onshi ps in the syste m is al wa ys av ai lab le Thisarticl eisaselectionfr om“Sys temDynam icsand A sys te m dyn am ics inves tigation usu - for su ccess ful mod el ing, but mana ger s are the Less ons of35 Years,” inKenyonB. DeGr ee ne all y shows that the important disc re pancy no t awa re of how the dif fer ent act iv iti es of Sy ste ms-Ba sed Approa ch to Poli cymaki ng, (ed .) (KluwerAcadem icPubl ish ers ,199 3). is no t acr oss the bo unda ry b-b, but ac ros s th e co mp any are infl uen ci ng one ano the r. the bo und ary a-a. When a mode l is built Lack of ca pacity fr om the ob serv ed and agreed-upon str uc - may exi st in manufa c- U N D E R I N V E S T M E N T ture and pol ici es, the model usually turi ng, pr oduct serv ic e, I N C A P A C I T Y ex hi bit s the ac tua l behav ior of the re al sys - sk illed sales pe op le, or te m. Th e ex ist ing kno wled ge ab out th e ev en in pr om pt o o pa rts of th e sy st em is shown to exp la in the an sw eri ng of tele - Delivery Customer Delay in the dia - act ua l beh av io r. The dissidence ph on es . For ex am ple , B4 Orders B1 s gr am arises beca use the intuitively ai rl ines cut far es to Price sec ti on is ex pec te d beha vi or in the middle at trac t pas senge rs . But s s s inc on siste nt wi th the known structure and how of ten, be caus e of Backlog Standard o “Buffer” po li cie s in the top section. inad eq uate te le phone y a l e Backlog D s R5 Th ese disc rep an ci es can be fo und ca paci ty, ar e pote ntial Capacity R3 s B2 re peat ed ly in the co rpora te wo rld. A fr e- cu sto me rs put on Profit o s Perceived Need Margin y que ntl y recurri ng exa mp le in whic h “h ol d” until they hang to Invest a l e D cause a los s of kn own corp orat e policies r up in favo r of anothe Capacity s Investments s - of employ ma rke t sh are and instability ai rl ine? me nt ar ises from the wa y deliver y de lay System dy nami cs Rising backlog damp ens customer orders because of increasing (s ee affe ct s sa le s and ex pansio n of ca pacity mod els hav e littl e delivery delay (B1). However, if management is reluc tant to invest in . Rising “Un derin vest ment in Ca pacity”) imp act un le ss the y capacity expansions until the backlog reaches acertain level (Standard “Buffer” Backlog ), orders will be driven do wn until demand incre ase in ba ckl og (a nd the ac co mp anying cha nge the way pe op le equals capacity (R3). The awaited signal to exp and cap acity never de liv ery del ay ) disc ourage orders incoming percei ve a si tuat io n. A comes, beca use capacity is controlling sales rather than potential fo r a prod uct (B1 ) even while manage me nt mod el must he lp to tries lowering price demand controlling capacity (B2). If management to stimulate demand (B4), the resulting lower profi tmargi ns will fur - fa vors la rger ba ck logs as a safety buffer org an iz e infor mati on ther justify adelay in ca pacity investmen t(R5). aga in st bu si ness downturns. - As manage in a mo re unde rst and - 4 1 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

44 O O O X L B T M A N A G E F L I G H T M E N T S I M U L A T O R S : F L I G H T T R A I N I N G F O R A G E R S ( P A R T I ) M A N ope ration s and focus ing on the long -t erm ma gin e you’re leavi ng on a six- hour ing, in the tr aditio nal bu siness -scho ol I dyna mi cs of manag erial dec isions. fl ight fr om Bost on to Lo s Ange les. nt of ground-s cho ol se nse, is the equivale As th e pla ne pul ls aw ay from the gate, the for pi lot s. Ma nage rs -to -b e rea d te xtb ook s CRE ATI NG A FLI GHT ake r: “Hi, pil ot com es on ov er the loudspe an d sol ve alr eady for mula ted pro blem s, SI MUL AT OR I’m Ca pta in Bob, and I want to thank you bu t they don’t get much rea l ex pe rie nce Ther e are four st ag es invo lv ed in cr ea ting fo r ch oos ing to fl y with us today. . . . Just bef or e the y ha ve to per form on- line. a ma na gem ent flig ht simula tor : (1) sele ct - wa nted to let yo u kno w I’ve rece ntly com - MANAGEM EN T FLI GH T ing an iss ue focus, (2 ) de vel op ing a conce p- ple ted gr ound sc ho ol trainin g, and I have SIM ULATOR S a comp uter tua l mo del, (3) co nst ructing re ad all th e manual s, but this is my fir st mode l, and (4 ) trans la ting the comp ute r Mana ge me nt Flig ht Simu lat ors (MFSs ) pr o- time in the coc kp it. So sit back, re lax, and mode l into an int er ac tive si mula tor (see vi de a si mulate d env iro nm ent in which en jo y the fli ght, as we learn toge the r. . . . ” “Ma na ge ment Flig ht Sim ula tor man ager s can “lear n fro m exp er ience” in a Of co urse th is sce nari o is lud ic ro us—a De velop me nt Sta ges” ). The se four st ages The simula tor ca ptur es con tro ll ed setting. pil ot is allowe d into a cock pit on ly afte r integ rating ma ny of the tools of involve the inter conne ctions be twee n the dif fe re nt hun dred s of hours of exp eri en ce in a flight sy st ems think ing int o a sing le, powe rful par ts of the syst em under study and pro- sim ul at or. Th en he or sh e spen ds man y lea rning tool (see “A Pa le tt e of Sys tem s vi des a co mpute r inte rface tha t allows in the ad di tion al hour s as a co- pil ot, assisting Thinking Too ls,” p. 10, for a de sc rip tion of man ager s to inter ac t with the model op erat ion of an air craft. Th e resul t of this ea ch of the to ols). throu gh a familiar “lens ” (r ep or ts , graphs , an d tra inin g is carefu l sys tem of ed ucation 1. Sele ct issue focus . Th e fir st ste p in an d spr ea ds he ets ). an in du stry with th e highest safe ty record of desig ning a flig ht simula tor is to choose an Si milar to a pi lot ’s fli gh t sim ula tor co ck- an y mo de of tr an sp or tatio n. issue to exp lor e. To se lect a top ic, look for pit , an MFS put s ma na gers in cont rol of a TR AI NI NG FO R FLIGHT a pr ob lem sy mp tom tha t ha s bee n ar ound rea lis tic env ir onm en t wh ere they are in MANA GERS for a long tim e or a puz zling dy namic you char ge of maki ng ke y deci sions sim ila r to wa nt to inves tig ate (see “The Do’ s and Im agi ne if we trai ned pilots lik e we do the one s the y fa ce in their act ua l work set- ing on the Job, ” Don’ t’ s of Systems Think ma nage rs; how many peo ple would be tings . MFS’ s are pa rt icul arly use ful for ge t- p. 52, for guide lines on ide ntify ing good wil li ng to tak e a fli ght? Mana gerial tra in- ting aw ay fro m the det ails of da y-t o- da y H T S T A G E S S I M U L A T O R D E V E L O P M E N T T F L I G M E N M A N A G E Simulator Flight Model Conceptual Computer Model Select Issue Focus Morale COCKPIT Personnel DECISION INFO Service Quality STOCK R HIRING • capacity STOCK • quality HIRING Productivity Time T/O Hires • productivity Time B Pressure Pressure pressure • time Productivity • etc. Work Done Test Gather the Gather the Test Gather Test the Data Model Data Model Data Model Revise 4 2 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

45 sys tems pro bl em s). The goal at this stage is a com put er mod el. It pr ovi de s a frame - ou ld requir e mana ge rs to ma ke dec is ions c to ga th er rel ev ant da ta thro ugh inter vie ws, work for peop le to dist ill th ei r exper ie nc e abo ut hir ing /f ir ing, mont hly product ion in to exp li ci t st at em en ts th at ca n be repr e- comp any rec ords, and the experie nc e base num be rs, and qua lit y st anda rd s. By imp le - sen te d in a com put er mod el . Just as the of tho se inv ol ved in dev el op ing th e MFS. me nt ing a “Qua lit y First ” po lic y, fo r ex am - d For exam pl e, let’ s sa y we ar e puzzle pil ot’ s fli gh t sim ulator is crea ted ba sed on ple , we may dis co ver tha t if we ra ise quality by a pa tt er n of osc il lating quality leve ls in th e laws of phys ic s an d aerod yn am ics, the st anda rds but don’t adj ust cap acit y, we actu- the cus tomer co mpu ter mod el in g pr oc ess use s a set of serv ic e dep art ment. ally end up wit h lowe r qua lity in the lo ng fu ndam en tal bu ild in g bl ock s (e. g. , accu mu - Int erv iews wi th peop le in the de partme nt run. Quali ty im pr ove s in the short run, but time pr es- la tors an d flows) to repr es ent a coh er en t set re vea l a pat tern of tremendous as tim e pr ess ur e pe rsis ts, mor ale decre ases, cyc le. su re that repea ts in a regular of th eori es ab ou t th e inter con nect ion s in an tur nov er inc re ases , whic h in turn inc rea ses org ani zat ion . Co mp any dat a pro vide a record of steadily time pr essur e, resu lti ng in more turnov er. In the cu st ome r serv ice qua lit y exa mple, lev els of customer ri si ng sa le s and irregular re inforcing The ove ral l dyna mic is a vicious the pro - sa ti sfa cti on . Thi s process grounds we cou ld mode l the numb er of “pe rso nnel” cy cle in wh ich ca pa cit y cont inua lly erodes as an ac cu mula to r and “h ires” and je ct in real dat a fro m whic h to build a and qua lit y suffe rs. “t ur no ve r” as inflo ws an d outf lows , re spe c- cau sal th eo ry. als o pr ovide The simula to r should tive ly. The ef fe ct of time pre ssu re on After sel ect - 2. Bu ild con cep tua l model. ma na ger s with the same ty pes of rep ort s, tur nov er ma y be modeled using a gra phica l in g an issu e fo cus , yo u can begin to bui ld a sp rea dsheet s, and graphs th ey use to ma ke func tion dia gr am represe nt ing a nonl inea r th e dat a co ncept ual model that organizes decis ions. Ther e are ma ny issues involving ic theory. in to a coh er ent dynam link be twe en th e two va ri able s. Tha t is , Syst em s the desig n of the sim ulat or ’s manag eme nt ther e may be li tt le or no ne ga tive ef fect s at arc het ypes and caus al lo op diagram s (C LDs) with inf or mation sys tem tha t ar e entwined low lev els of pr essure, but be yond a ce rta in can be very helpful for trying to unde rstan d the inte nded use of the MFS as a whole . what is goi ng on (s ee “Sys tems Ar chet ype s at thr es ho ld, the re ma y be a sudde n dra matic Thes e and othe r is sues ar e cover ed in Pa rt li nes for a Glan ce, ” p. 20, and “Guide inc rea se in tu rn ov er. II (p. 44) . • Pilot s ,” p. 18). 4. Trans late to fli ght sim ulat or . g Ca usa l Loop Diagrams Drawin For help onco nve rt ing conc ep tualma psto com put er fir st learn abo ut the pr incip les and con - exa m- In th e cus tom er servic e quality model s,see“Accum ulato rs: Bathtub s,Bath tub s pl e, we can st art bui lding caus al structures ce pt s of aviatio n in sc hoo l and then us e the Eve rywhe re, ”p.30;“S tru ctural Thi nki ng: TheWorl d Acc ord ing toAccum ulato rsan dFlows,” p. 28;an d tha t provid e pla usi ble exp lanations for the si mulat or to gain a be tter under st anding of “Gr aph ica lFu nct ion s:‘See ing ’th eFul lSto ry,” p. 26. how thos e princip les ac tu ally pla y out in obs erved dat a. When cust omer demand Tolear nmor eab out cons tru ctingcomp utermod els, Intr oducti on to Sys tem Dyna mi cs Mo delin g see re al lif e. Like wis e, a man ag em ent flig ht inc re ase s, we know our service pe ople fe el (Pe gasusCo mmuni ca tions) ,an d Aca dem ic User’s si mulat or is cr eated by tr ansla ting the add ed pr essure. If the increa se in time Guide to STEL LA (Han ove r, NH:Hi gh tends to quality pr ess ure is no t addressed, “pr inc iples ” capt ure d in the com put er Per for man ce Syst ems ). dr ift do wn wa rd and ev entually dampens mode l into a fo rm tha t allows mana ge rs to Loop s”) . interac t with it in a re al - de man d (B 1 in “Ti me Pressure ist ic way. ini - Ou r ser vi ce peop le tell us people T I M E P R E S S U R E L O O P S A go od simul at or e by tia ll y re sp ond to the tim e pressur o Morale int erfa ce shoul d pr ov ide pr o- increasing wo rking ha rder, thereby D e l R3 a mana ge rs wit h a se t of duc ti vi ty and get ting mo re wo rk done y s s (B 2). If the ti me pressure persists, how ev er, de cis io ns tha t eit he r th ey Customer s Demand mo ra le decli nes and begins to hur t pro duc - co nt rol, or tha t di re ctly Productivity Time s affe ct the m. The main tiv it y even though people contin ue to work B1 B2 Pressure escalates hard er. As tim e pressure cr it eria shou ld be th at the , morale o Work Quality sp ira ls downwa rd (R3). By adding de cis io ns ar e dir ec tly rel - loops , s Done o we ca n cont inue bui lding a dynamic ev an t or eas ily tr an sf er - theor y abl e fr om the simu lat or setting. abo ut our custo mer service Customer demand puts pressure on service personnel. Initially, people may work harder, thu sincreasing productivity and reducing Th e 3. Con st ru ct co mp ut er model. to the wo rkplac e. the pressure (B2). Over time, however, morale can su ffer, hurting theor y de vel oped in the conc ep - In the se rvic e qu alit y dy namic productivity and increasing the time pressu re (R3). tual model help s guide the con st ruc ti on of ex amp le, the simula tor 4 3 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

46 O X L B O O T M E N T F L I G H T M A N A G E R S : F L I G H T T R A I N I N G S I M U L A T O F O R M A N A G E R S ( P A R T I I ) AGING MAN VS. LEA RN IN G • A clea r, real -wo rl d con text pr ovi des a man agem en t fl ight simulator, A re al op er at ional focus tha t enga ges line al ong wi th causal loop diagr ams Th ere are two funda ment ally dif fer ent ma nag ers in lea rn ing mor e about th ei r own and sys te ms arc het ypes, allows you to see and sim ula - us es for a com pute r model issues. mo re cl ea rly the connecti ons betwee n your and le arning . Simula tor s tor —m anaging • Fa ce val idi ty: Ma ke th e MF S real es. As sim - de ci sio ns and future co nsequenc an d mod els des igne d to sup por t de cision- enoug h so the si mul at or grou nds peopl e in of mi nutes ula te d mont hs pa ss in a matter mak ing in a re al ope ratio na l se tting mus t the ir own rea l- li fe ex per iences. and the co nsequenc es of your actions foc us on ca ptur ing the op er ationa l reality • A st ron g concep tual fra mewor k hel ps for unfo ld , an MFS pro vide s a means prec isel y be caus e ope ra tiona l or strat eg ic ma ke sy st emi c sense out of the com plex and long- ma kin g sense of the sho rt-term de ci si on s will be ba se d on those num ber s. dy nam ic s (e.g ., syst ems arch etypes). te rm effe cts of your decisions. Si mula tor s tha t ar e de signe d for lea rn- • Co nvent iona l and unconv ent ion al rs can be Ma nage ment flight si mulato in g, on the othe r hand, are much mor e info rm at io n syste ms provi de a fami lia r in situations mo st usefu l fo r understanding con ce rn ed with sur fa cing the ta cit ment al , as well as an info rm at io n envi ronment wh ich ca usal ity is di stant in ti me and mod els that drive mana ge rs’ de cision- op port uni ty to ex pl or e and exper im en t wi th sp ac e. When the inherent time lag is par - mak ing. Acc urac y of specif ic num be rs is new ones . tic ula rly long (o n the order of months or not as impo rtant as the rele vanc y of the • Sur fac e an d cha lle nge ment al models ye ars) and orga ni za tional comp le xity is is su es and co ncep ts captu re d in the sim ula - to brea k thr oug h in dividua l ment al st raig ht - hig h (s ee “Org aniz atio nal Comple xity” ), tor ; in ot he r wor ds, simula to rs for lea rning ja ck ets and cor por ate sa cred cow s and lea rni ng from ex perience can be fraught ar e idea-r ich versus data- ric h. adva nc e tea m lea rning . you to leve r- wit h pitfa ll s. An MFS allows The re ar e se ve ra l dif fer ent simula tor age yo ur abi li ty to lea rn from exp erie nc e in de sign cr ite ria to ke ep in mind whe n DES IGNING MFS ’S AS a co mp le x envi ronment. de sign in g an MFS: TR ANS IT IO NAL OB JE CTS Desi gni ng an MFS for le arnin g re qu ires an T I O N A L O R G A N I Z A C O M P L E X I T Y int er face tha t ma int ains a ca refu l ba lan ce be tween real ism and com pr eh ens ibilit y. It ” Cross- National Standard S nee ds to be real eno ug h to serv e as a tra ns i- S Industry of Living E M “ tiona l obj ect , whi ch Seym our Pap ert say s A Cross- Supplier-Distribution F al lows man ag er s to “pla y ou t” a scen ario , O Networks Organizational Y T y I lea rn abou t the syst em , and ex plo re how the y t i X x e l E p L Product m Cross- P int er act wi th that sys tem. Howev er, it also o C Development Cycle M c i Functional m O a C nee ds to be ma na gea ble— if the mo de l tries n y D L g n A i s Manufacturing to capt ur e eve ry little detail of realit y, it can Multiple- a N e r O c Time Cycle I Intrafunctional n I T be com e just as com pl ex and inc om pre hen si - A Z I and Scrap N bl e as the rea l syst em. A Unifunctional Costs Repair G R In a lea rni ng se tting, it is also im port an t O Decades Months Days Years Hours no t to po si tion the mod el as an an swe r- of) order the (on PROCESS OF LAGS TIME ge ner at or, but rat her as an explo ra tory tool Th egreate rthe complexity or the lon ger the time lags, the mo re an MFS is beneficial for help - for ga in ing a bett er und er st an ding ab out ing peopl elearn from experience. on e’ s envi ro nme nt . The MFS acts li ke a 4 4 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

47 mi rror th at ref lects ment al models in a wa y ing , qua lit y and cos t, most of them act ually ma na ge di ff er entl y whe n provi ded wi th that he lp s us un derst an d curren t re alit y be tter. cho se to inv est ver y lit tle in coordinat ion. such inf or mati on. elemen The re are three major ts of an Inst ea d, the y foc use d on trying to get the OUT COMES FRO M TH E and a manage men t MF S: de cision s, reports, tas ks don e in ea ch area . SIMULATOR in fo rm ation syst em (se e belo w for a samp le A simul ato r ca n be ev en Te am lea rning . Micro Wor ld in ter fa ce produ ced with mor e usef ul if us ed in gr oups. Th e inter pla y On ce parti cipants wor k wit h the si mul at or Cr eato r). bet ween the par tici pants, as they pr opose and under stand th e theo ry behi nd it, the y De cis ion s. Th e kin ds of de cisi on s made new st rategi es and exp lai n thei r rea son in g, ca n mak e conn ecti ons be tween the si mul a- in the simulat or should be thos e the parti ci - hel ps the m to sur face and cl ar if y thei r to r and the ir real wo rk si tua tio n mor e eas- pants wou ld eit her make themsel ve s in rea l ass ump ti ons . The si mul ator can be st ruc - il y. Pa rti cip ants can al so exp lor e what li fe or tho se so meon e els e in the org aniz a- tur ed to requi re pa rti ci pa tio n an d coor di na - in terv ent io ns the y might make in order to ti on wou ld make that affect the part ici - ti on amon g a gro up of peop le to en co urage be tte r ma nage th e pr oces s: Wha t ki nds of be direct ly pants . The decision s should te am lea rni ng . For exa mp le, in a pro duct adju stm ents need to be made? What con - rele va nt or easily tr ansferr ab le fro m the eng ine eri ng case, the tea m cou ld be mad e tro ls do par tici pan ts nee d to mo ni tor ? sim ula tor to the wor kpla ce. If the decisi ons up of a pro duct and a proce ss eng inee r. The sim ul at or ca n be Th eor ies -i n- us e. are too fa r remov ed , the simul ato r beco me s be res ponsi ble for sta ffing Each one would a po werfu l to ol for surf aci ng taci t assum p- mo re of an acad emic exercis e or a gam e, and wor kwee k de cisions for the ir part icu - ti on s, fo r it refl ects part ic ipant s’ under - is bui lt eve n when a mean in gful context lar funct ion, but tog et her the y would sta nd in g of the sys tem . When some one arou nd it. decide on a schedule comp let ion dat e and ma ke s a decis io n and the n exp lai ns it wit h Altho ugh the partici pan t mi gh t no t be ma na ge the coor dinat ion betwee n the two da ta or in fo rmat io n that ar e no t in the th e one who makes hir ing and staf fi ng deci - funct ions . The use of the simula tor can mo del , the y exp licate their own theor ies le, these sion s at his /h er le vel, for examp the re for e be des igned to provide a ric her and under standi ng of wha t’ s going on. For de cis ion s can be in clude d in the MFS pr act ice field for a te am to mana ge. ex am ple, in a managem en t flig ht simul at or be caus e they are still par t of the real en vi- crea ted fo r insur ance clai ms mana ger s, par- RY SU MMA ronm ent in which the partic ipant s manag e. ti cip ants assum ed that sett lement do llar s MFS’ s pr ovide ma na gers wit h a sim ulat ed In fact, putt in g that manager or su perv isor were ri sing becau se of inf lat ion. Howev er , ex per ienc e of wor king thro ugh issues or role can be an il lu - in to the decision -making when they di sco ve red th at inf lat ion was no t impl eme nt ing a strateg y. The pra ct ice fie ld mina ting expe rien ce: He or she wil l learn in clu ded as part of the sim ul ator , they ha d ele me nt also ena bles the sim ulat or to pro vide and real - what role the y play in the system to reth ink the ir own unde rst andi ng of wha t an ex per ie nt ial “fe el” for the dyna mic s of fro m the iz e the ch alle nge of managing ca uses se ttl em ent do ll ars to rise. dec isio n-m aking . Pa rt ic ipa nt s gain pra ctice le ve l ab ove . Th e simula to r can also reve al the ga p in the art of dec is ion -m aking : re fl ect ing on As far as th e actual phy sical Repor ts. that oft en exis ts betw een espo us ed theorie s the caus al con - the cons eque nce s, ex ploring or interface, de si gn of the simulat there are (wha t we con ce ptua lly believ e is the rig ht nec tions, and under standin g the underlying The repo rts som e gene ral guid elines. co ur se of act io n) an d our the ories -in- use st ruct ur e pr oduc ing the beha viors. • to wh at peo pl e ty picall y shou ld loo k similar (wha t we cho os e to do, giv en the surro und - rece iv e— they sh ould not pr ovide add it io nal ing cir cums tanc es). For da ta th at is normally not acc essib le, su ch as ex amp le, in a ses sio n wit h a ti me pr essur e, or perce ived quali ty by the pr odu ct de ve lo pme nt flig ht cu sto mer . If th ese variables ne ed to be simula tor, the par tic ipa nt s in clude d, the y sh ou ld not be as pro mi nen t all agr ee d that inv es ting in as mor e ty pical day-t o-day data. co or di nat io n and commu - In form at io n sy st em s. De signin g the nic at io n be tw een upst ream a lot mo re in fo rm ation syst em provides and do wns trea m ac tivi tie s fle xibil ity in repor tin g variable s that are is im por tant . But when le, a cri ti - no rm ally inacces sible. Fo r examp they were pla ce d in the ca l va ria ble like “time pres sur e” can help simula tor and gi ve n the you ex pe rimen t to se e how pe op le may ob je ctive of me eting tim - 4 5 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

48 L B O O O X T L E A R N I N G L A B O R A T O R I E S : P R A C T I C I N G B E T W E E N P E R F O R M A N C E S magi ne you are wa lking acr oss a enco unt er in the ac tua l set ting . It doe s no t mean yo u agre e or disa gree with I tig htro pe stret ched betwee n the A man ag er ia l pr ac tice fie ld should als o tha t pe rson’ s view ; you simply ac know ledg e wo rl d trade towers in New York Cit y. Th e ha ve it s ow n se ts of equipm ent and tools for the right of tha t pe rson to ho ld that vie w. A wi nd is blo wing and the rope is sh akin g as mak ing the pr ac tic e sessio ns mea ning ful. se cond rule is to sus pend one ’s own ass ump - you inch you r way forwa rd. One of your The pur po se of a “l ea rning la bora tory” is to tions and the ot her pers on’ s and hold them te amm ates is sit ting in the wheelba rr ow yo u in whic h mana ge rs pr ov ide an envir onment equa lly in our minds, wi tho ut judg ing ours ar e bala nc ing in fron t of you , wh ile an ot he r can ex pe ri men t wi th alt erna tive pol ic ie s, tes t to be sup erior or “rig ht .” Crea ting suc h a pe rc hes on your sh ould ers. There are no as sumpt ions , and pract ice worki ng throug h lea rning spa ce als o me ans enga ging in dia - sa fe ty ne ts, no harn esses. You are th in king to comp le x is sue s pr oduct iv ely. It should allo w log ue rat he r than dis cus sion— op era ting in a yourse lf , “One false move and the thr ee of us man age rs to pr act ic e wo rking toge the r as a spir it of inq uir y rather tha n advo cac y. will be tak ing an express eleva tor str aigh t tea m on is sue s of re al signific ance to the m. MAPP ING ME NT AL ly your tr ain er dow n to the stre et.” Sudden To be effe ctiv e, the le arn ing la b must pro - MODELS ye lls fro m the othe r side, “Try a new move! conduciv e to le arn - vi de (1) an envir onment Along nt, we wit h the pr ope r env ironme Exp erime nt! Tak e so me risks! Re mem ber , ing , (2 ) a way of sur facin g de ep- root ed nee d too ls fo r he lping pe ople surfac e and yo u are a learning tea m!” as sumpt ions that affe ct the wa y we thin k sha re the ir assum pti ons . For ex am ple, the Soun d ludi crous ? No on e would be an d act, (3) tools for und ers tanding our re al- “L adder of Inf er enc e, ” develo ped by Chris ng ne w in a si t- cra zy eno ug h to tr y somethi it y in a wa y that hig hlig ht s the int erco nnec - Ar gyris, dis ting uishe s bet wee n direc tly ua ti on like that. An d yet that is prec isel y tio ns and the sys temi c co ns eque nc es of ou r ob se rvable da ta, sha red cult ura l mea nin gs, wh at man y companie s expe ct man age me nt ac tio ns , and (4 ) a mana gem ent fli ght sim ula - jud gment s, conc lusi ons, and va lues and tea ms to do— exp er imen t and learn in an tor that al lo ws us to speed up or slow do wn assum pt ions . Ar gyr is use s the la dder to en viro nme nt that is risky , tu rb ul en t, and time , ex pe rime nt wit h diffe re nt stra teg ie s, illus tr ate the “lea ps of infe renc e” tha t occur act or unp red ict abl e. Un like a hig h-wire an d see the long -t erm co nsequ ence s of our whe n peo ple take a lit tl e bit of obs erved da ta r, man agemen t tea ms sp ort s tea m, howeve ac tio ns (s ee “A Sample Le arning Lab ora tory (a pe rson wa lk s int o a 2:00 me et ing at 2: 15) fie ld in whic h to do not have a practice De sign” ). and go straig ht up the la dde r to the lev el of th ey are alway son the perform ance field. learn ; ATIN G A SAF E CRE va lues and assum pt ions (“ He’s la te and NG SP AC E LEARNI DE SIG NING MEA NIN GFUL doe sn’t ca re about the proj ec t or the other PR AC TICE FIE LDS Lea rning us ua lly invo lves making mis takes pla yer s”) wit ho ut ev en being consc ious of it. A lea rni ng lab can be viewed as a man age r’s be ca us e we are trying thi ng s we ha ve ne ver The la dder pr ovide s a usefu l fram ew ork fo r eq ui va lent of a sports te am’s practic e fie ld. us to appro ac h do ne be for e. It requires “wa lk back do wn the la d- helpin g people The goal of a lear nin g lab de si gn is to pro - thing s fr om a plac e of “not know ing .” It der ” to under st and wha t is re ally ha ppe ning vid e a “real” enou gh pr ac ti ce fi eld so that inv ol ve s ris k. How , then, ca n we crea te a and beg in mana ging by fa cts, not op inions. l, but safe enou gh the les son s ar e me aningfu safe spa ce whe re pe opl e fe el fre e to le arn? Sy st em s archet ype s als o prov ide a pow - to enc ou ra ge exp er imen tation an d lea rn ing. Ther e ar e some grou nd rule s tha t can er ful se t of tools for mapping out a pe rson’s e, a practic e fie ld In th e ti gh trope exampl he lp cr eat e suc h safe spac es. One gro und unde rstand ing of a prob lem or issue in a cou ld be a rope stre tche d across tw o pill ars rule is to ho ld eac h perso n’ s viewpo int as fo rm that invit es others to inquire and cla r- six feet off the gr ound . The re may be mats va lid. That req uir es taki ng the po sit ion tha t ify the pict ur e toge ther. Ha ving one’s belo w to cus hion th e fall , but also a large fa n “If I co uld sta nd in the oth er pe rson’ s sho es , assum pt ions capt ur ed in te rms of archet ype s to simu la te the kind s of win ds yo u wou ld I too could se e wh at the other pe rson se es .” nal - and ca us al loo p dia gram s helps deperso 4 6 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

49 fo rce s that produce a give n set of out com es. ize th e issue an d focus es eve ryon e’ s ene rg y of time at each de cis io n point . Wit h a si mu - The se dia - on th e diagr am, not the person. lat or, a man ag er can test out new st rat egi es FORM ING ON THE PER th e assu mptio ns be hind gram s al so explicate and po lic ie s, ref le ct on the out co mes, and TI GH TR OPE , an d cl ar ify the po ints of the co nnections dis cuss pe rtin en t is sues wit h ot hers in th e In order to fost er lea rning among tea ms of t or con te ntion. agreemen tea m. ma na gers, we must look fo r alte rna te wa ys to By pr ov iding quic k and acc urat e fe ed - GEM ENT FLIGHT MANA hel p th em dea l with th e in cr ea sing co mp lex - bac k, the comput er simul at or ca n fac ilit ate ATO RS SI MUL ity of today’ s busi ne ss wo rld. Prov idin g safe , le ar ning by shor tening th e dela y betw een ng a conce rto, an orc he stra ha s Wh en practici yet mea ni ng ful lea rni ng env iro nm en ts where ac tion and out come . Man ag ers can cha rt a the abil it y to slow down or spee d up ti me to the y can co nt inua lly alte rn at e be tween prac - str ategy and imple ment it ov er a sim ulat ed pr act ice cer tain se ctions. Th rou gh ma na ge - tice and pe rfor ma nc e is one ap proa ch . r of ye ar s in a ma tt er of minut es . numbe ter mo de ls ment fl igh t simul ators (compu Whe th er we ar e wal kin g on a tig ht ro pe The y can tr y scenar io s that ba nk rup t the tha t have be en turn ed into in te rac tive de ci- st ret che d acr oss two build ings or across two co mp any or lo se ma rket sha re wit hout ri sk - sio n-m aki ng games ), man age rs can also co mpet ing pr od uct stra teg ie s, prac tic e is ing a singl e do lla r or jo b. As they explo re the acc elera te time to see the long -term conse - bo un d to impr ov e our per fo rm an ce. sys temic re asons for their re sult s, mana gers • quen ces of decision s, or sl ow down the flo w ca n beg in to unde rsta nd the underly ing A S A M N I N G L A B O R A T O R Y L E A R P L E D E S I G N 3. Intro duc ing th e Tools In des ign ing aLe arn ing Lab (LL) ,the goal isto create an Causalloop di agr ams (CLD s) and sys tems arc hety pes envir onm en tthatis of operati onal relevance. The lab areintroduced in a“stor ytell ing” for mat in whic hpar tici- should help managers step out of day-to-day demands to: pants begi nto tel lsyst emic stor ies about thei rissues. • ref lec ton their de cisi on -maki ng Facil ita tor sthen des cr ibe asmall por tion of the CLD sthat • deve lop aco mmon lan guag e were in the game model to connect the tool to the iss ues • lea rnnew too ls for thinki ng syst em ica lly athand.The underl ying pur pos eisto get people to imme - • dis cu ss op er ational ob ject ive sand stra tegi es inan diatelybegi nto connect each str uctur etocor res pondi ng open forum patte rnsof behavi or over time. • tes top er atin gassum ption s • expe rim en twith new pol ici es andstra teg iesfor 4. Us in g the Tools —Co nc eptua lizi ng man ag ing In sma ll gr oups, the par ticipants are ask edto focus on a • have fun. particulariss ue, such as one ofthe dec ision vari ables in thema nagem ent flight simul ator ,and (1) determi ne the 1. The Fi rst Crucial Hou r— Bu y-In ke yfactor sthat affect that var iable, (2) sket chpat ter ns of Explaining the context of the LL to par ticipants (the hist or y behavio r,(3) pr ov ide str uctur al explanati on (using CLD s) , of its de ve lop me nt, the or iginal int en tor pur po se) is cr itical and(4)identi fy inter venti on points .Byhavi ng the gr oup for esta blis hin gacom mon under standi ng: The LL is not explore these vari ables, the par ticipants can repl icate part meant to pr ovid e“the answer s, ”but to serve as ausef ul ofthemodel -buil ding pr oces sand ac cept the pr edevel - vehic le for illu min atin gand communi cati ng issues of opedmo del .The over al lobjec tive in this sec tion isto importa nc e. The facilitator sar epositioned as enablers, have thegr oup cover all the major issues contained in the not au tho rityfigu res, and the part ici pants are encouraged mo deland have achanc eto chall enge and test the inter- to que stio ntheas sumpti ons behi nd the LL design. relationsthat di ffer ent peopl ewithi nthe group may pr o- Reality—W 2. Current he re Ar e We? pose . Thi sexe rcisehe lps the part ici pants construct agr oup pic - 5. Intro duc ing th e Co mp ute r Simu la tor tur eof cu rrent proble ms and iss ues they face intheir jobs. Thefacili tat or sbegi nby show ing asimpl ifi ed CLD that Working in sma ll grou ps, they are then asked to br ain - co ntainsal lthe maj or var iables inthe model .They trac e storm an dcome up with alist of oper ati onal objectives, throughthe major loops and ex pl ain the dy namic conse - str ategies requ ired toachieve them ,and obstacl es that quence sof apar ticul ar acti on or inc ident, and then sketc h need tobe ov er come in or der toreach their goals (e. g. , acorresponding patter nof behav ior and connect it back to redu ce settle me nt exp enses by 10 percent). The idea is to thestr uctur e. This is fol low ed by ahands- on intr oduc tion get ev er yon ethinking in ter ms of their own oper ational to thecomputer and simulator . real ity. Co nt inue donthe next pa ge 4 7 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

50 Co nt inued from theprevious page L E A R N P L E A S A M I N G L A B O R A T O R Y D E S I G N 7. Fr ee Pla ys—Cut ti ng th e Re in s 6. Pl an ned Scenari os— Ho ldi ng the Rei ns This time, the parti cipants ar efree tochoos ethei row n In this sec tio n, it is best if people work in groups of two at objective sand time tabl es. Agai n, eac hteam st rat egi zes , each com pu ter. The teams are instr ucted to pursue a andthenexpl ai ns tothe res tof the gr oup how they plan to si ngle-min de dstra tegy where they are accountab lefor achievethei rgoals. For the des igners of the LL, thi ssec - meeting one par tic ul ar goal (e.g. ,hir ing freeze). Each two- tionprovides the oppor tunity to chall enge deep-r ooted per son team is resp onsi ble for doing the foll owing: normsand assumpt ions, addr es sspecif ic “hot topi cs,” or (1) plan astrate gy and com mit to it on paper, (2) predict re-cr eate var ious histor ical behav ior modes for fur ther the con se qu en ce sof execut ing the strategy by sketching exploration. in behav iorov ertime of some key vari abl es, (3) play the gam e, an d(4) debrief game resul ts and explain tothe res t of the group .Th is stage allow spar ticipants tobegi nto addr ess pa rticu lar or gani zat ional issues withi nacarefully contr olled se tting . 4 8 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

51 P A R T V R E F E R E N C E G U I D E 4 9 PEG ASU S COMMU NICA TIONS , INC. WWW.P EGAS USCOM. COM SYSTEMS THI NK IN G TOOLS

52 L B O O O X T O F S Y S T E M S T H E V O C A B U L A R Y T H I N K I N G : A P O C K E T G U I D E yste ms thinking can serv e as a la n- form of a clos ed loop th at dep icts cau se- the tw o ov er the en tire ra ng e of re lev ant S gua ge for communicati ng ab ou t com - and-effe ct lin kag es. val ues . The res ulti ng diag ram is a conc ise encie s. To be fu lly ple xi ty and in terdepend —A sy ste ms arc hety pe. In a Dr ifti ng Goa ls hypoth es is of ho w the tw o vari ables in ter - co nversant in any lang ua ge , you mu st ga in “Dri ft ing Go als ” sc en ari o, a gra du al relate . Also call ed . Ta bl e Func tio n som e maste ry of the vocab ul ary, es pe cia ll y dow nw ard slid e in perfo rm an ce goal s Growth an d Under inv es tm ent —A sys tem s the phrases and idioms un iq ue to th at la n- goe s unnoti ced , th reate ni ng the long-term arch ety pe. In thi s si tu ati on, res ourc e guag e. This glossary lists man y te rms th at futu re of th e sys tem or org ani zati on. investme nts in a gro wi ng area are no t may come in handy wh en you ’re face d with a Exa mpl e: len gthe ni ng del ivery de lay s. made , owin g to short- term pre ssures . As . sys tem s problem —A sys tem s arc he typ e. In th e Es cal ati on grow th be gin s to stal l bec au se of lac k of “Es ca lation ” arche typ e, tw o parti es co m- resource s, th ere is les s incen ti ve for —An yth ing th at bu il ds up or Accum ulator pete for su pe riori ty in an are na. As on e addi ng cap ac ity , an d grow th slows eve n dw indl es; for exampl e, wa ter in a bath - party’s action s pu t it ah ea d, th e oth er furth er. in a bank accou nt , inve ntory tub, savings party “re tal iat es ” by incre as ing its acti ons. Lea rn ing Lab orato ry —O ne of th e 10 to ols of in a wareh ouse. In mo de li ng softw are , a The resu lt is a con ti nual ra tch eti ng up of system s th inki ng. A le arn in g la bo rato ry stock is of ten used as a gen eric sym bol for acti vit y on bo th sid es . Ex am pl es : pri ce bat- embed s a man ag em ent fl igh t simulator in or Also know n as accum ulators. Stock tle s, the Col d Wa r. a learn ing en vi ron men t. Gro ups of man - Le vel . Feedb ack —Th e re turn of info rm ati on abou t agers use a com bin ati on of sy ste ms th ink- Proc ess/L oop Bal ancing —Com bi ne d wi th the statu s of a proce ss. ing too ls to exp lore th e dyn ami cs of a rei nfo rcing loop s, bal an cin g proces se s Exa mpl e: ann ual pe rform an ce rev ie ws parti cula r sys te m an d in qui re into th ei r blocks of dyna mic sys - fo rm th e building ation to an employ ee about retu rn inform own un ders tan di ng of tha t syste m. tem s. Balancing process es seek eq ui li b- the qua li ty of his or her work. Learni ng la bs serve as a man age r’ s pra c- ri um: They try to bring th in gs to a —A sy stem s arc hety pe. In a Fixe s Th at Fa il tice fi el d. des ired state and keep the m th ere . The y “Fixe s Tha t Fail ” si tu ati on , a fi x is Accum ulator . See Lev el — also lim it and constra in ch an ge ge ne rat ed applie d to a prob le m and has immed iate Lev er ag e Point —A n area where small proces se s. A bala ncin g loop by reinforcing pos itiv e re sul ts. How ever, the fi x al so has chan ge ca n yield larg e im pro vem en ts in a loop diagram de pi cts a bal anc- in a causal unfore see n lon g- te rm co ns eq uence s th at system . ing process. eve ntu al ly worse n th e prob lem . Al so —A sys tem s arch ety pe . In a Limits to Succ ess Bal ancing Proc ess with Del ay —A com monly know n as “Fi xes Th at Backfi re. ” “Lim its to Suc ce ss ” sce na rio , a co mpan y struc ture. Wh en a ba la nci ng occ urring —Th e am oun t of ch an ge som eth ing Flow or prod uct line gro ws ra pidly at firs t, bu t pro cess has a long dela y, th e usu al unde rgoe s duri ng a parti cul ar unit of eventu al ly beg ins to sl ow or ev en dec li ne. res po nse is to overcorrect . Ove rcorrecti on tim e. Ex am pl e: th e am ou nt of wate r th at The rea so n is th at th e sys tem has hit some Exa mpl e: in beh avi or. leads to wild swings flow s out of a bath tu b each min ute, or the ac ity con st rai nts , re so urce lim- limit—cap real estate cycles. amoun t of inte res t ea rn ed in a savi ngs its , market sa turati on, etc. —th at is Over Tim e (BO T) Di ag ra m Be havior —One accou nt each mont h. Als o cal le d a Rate . inhib iti ng fu rth er gro wth . Als o cal led s thi nkin g. BO T of the 10 tools of system Gen er ic Stru cture s —Stru cture s th at ca n be “Lim its to Grow th .” di agram s capture the his tory or tre nd of gene rali ze d across man y differen t setti ng s —On e Mana geme nt Flig ht Simu la tor (MFS) one or more variables over tim e. By becau se the unde rly in g rel ati on shi ps are of th e 10 to ols of syste ms th inki ng . variab le s on one grap h, sketch in g several fun da men tall y th e sam e. Syste ms Sim ila r to a pilo t’s fl ight simula tor, an you can gain an expli cit un derst an di ng of arch et yp es are a cla ss of gen eri c stru c- MFS al low s mana ge rs to tes t the outc ome how the y interact over tim e. Als o ca ll ed ture s. of diffe ren t po lic ies and dec isio ns witho ut Refere nce Mode. Gra ph ica l Fu ncti on Dia gra m (GF D) —On e “cras hing an d burn ing ” re al com pan ies . (C LD) —On e of the Caus al Loop Diagram of th e 10 tools of sy ste ms th in ki ng. GFD s An MF S is bas ed on a sy stem dynam ics thin kin g. Caus al loop 10 tools of systems show how on e va ri ab le, su ch as del iv ery compu te r mode l th at has bee n cha nged how va ria bl es in a sys - di agram s capture dela ys, int era cts wi th ano th er, such as into an inte rac tiv e dec ision -ma kin g si mu- tem are interrelated. A CL D take s the sales , by pl ottin g th e rel ati on shi p betw een lator th rou gh th e use of a user interfac e. 5 0 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

53 co mpl ex who le . Alm ost always de fin ed —One of th e 10 Diagram Pol icy St ructure imp orta nt to a pr ob lem or an issue and thinkin wit h re spect to a speci fic pu rpo se wi th in a g. Policy stru ctu re tools of systems then sim ul ate th e int era cti on of those vari - lar ger sys te m. Ex ample : An R&D depart - dia gram s ar e used to crea te a conce ptu al bles over time . a “ma p” of the dec ision-m men t is a syst em tha t ha s a purpo se in the aki ng proces s tha t See Accu mul ator . Stock— co nt ex t of th e la rg er org ani zat io n. is embedded in an orga ni zation . It hig h- Str uctur al Di ag ra m —Dra ws ou t the ac cumu - System s Arch ety pes —O ne of th e 10 to ols of that are weigh ed at ea ch lights th e factors lat ors and fl ows in a sys tem, gi ving an decis ion point. sys tem s thi nki ng . Syste ms arch etyp es are ov ervie w of the major st ructu ral ele ment s th e “c lass ic stori es ” in sy st ems th in ki ng— Flow. See Rate — that prod uce the sy st em’ s beh avi or. Also See Mod e— Re ference Behavior Ove r Tim e com mon pattern s and struc tu res th at call ed fl ow diag ra m or accumu la tor/f low Diagram. occ ur re peated ly in differen t setti ng s. dia gra m. Re inf orcing Proc ess/L oop —Al ong with bal - —A sch oo l of th ought th at System s Thin kin g —O ne of the 10 to ol s Str uctur e- Be havi or Pai r ancing loop s, reinforci ng loops form the fo cu ses on reco gni zin g th e interc on nec - of syste ms th in king . A stru cture -b eh av io r buil din g blocks of dy nam ic sys tem s. ti ons betw een th e parts of a system and pair cons ists of a st ruc tura l rep re sentati on Rei nforcing syn th es izi ng th em into a uni fi ed vi ew of processes comp oun d cha ng e of a busine ss issu e, using acc umulato rs in one direction with even more ch ange in th e whol e. and flow s, an d the corre sp ond ing beh av - Graphi cal Func ti on See As su ch, the y ge ner- that same direction. Tabl e Func ti on — ior over tim e (BO T) diagram fo r th e issu e Dia gr am. ate both growth and col la pse . A rei nforc - being stu di ed. Templ ate —A to ol use d to id en ti fy sy ste ms ing loop in a cau sal loop diagram de pi cts mann er in whic h a syste m’s Str uctur e —The a re inf orc in g proce ss . Als o kn own as arc hetyp es. To use a te mplate , you fil l in ele me nts are org ani zed or inte rrel ated . th e blank vari ab le s in causa l loop di a- vic iou s cycles or virtu ous cycl es. Th e structu re of an org an izati on, for gram s. arch etype. —A systems the Burde n Shifting exam ple , cou ld in cl ud e not only the Trag ed y of the Co mmon s In a “S hifting the Burde n” si tu ati on, a —A syst ems org aniza tion al ch art but al so inc en ti ve arch etype. In a “T ra ge dy of the Commo ns” sho rt-t erm solut io n is tri ed th at succe ss - syste ms, inform ation fl ows, an d inter - proble m. As the sce na rio , a sh ared reso urce be com es ove r- full y solves an ongoing person al inte ractio ns . sol ution is used over an d ove r aga in , it burd ene d as ea ch pe rso n in th e syst em use s —A sys tems Succe ss to th e Succe ssfu l takes attention away from more fu nda - more and mor e of the re so urce fo r in div id - arche type . In a “S ucce ss to the Succe ssf ul” ual ga in. Eve nt ua lly, the resou rce dwin dle s sol uti ons . Ove r ti me, men tal, enduring sit uat ion , two acti vi tie s comp et e for a com - or is wipe d out , re sult in g in low er gains fo r the ab ility to apply a fun da me nta l sol u- mon but li mi ted resource . Th e acti vit y th at tion may decrease, the re su lti ng in more an d everyo ne involv ed. Exa mpl e: is in it ial ly more succe ssfu l is consi ste ntly Gre en ho us e Effect . more re liance on the sym ptom atic solu - giv en more res ou rces , all owing it to suc - • Examples: drug and al cohol dep en - tion. ceed even more . At the sa me tim e, th e dency. activi ty that is ini tia ll y less succe ssfu l Theabov egl ossary isaco mpi lationof defi nitionsfrom the Burde n to the Inte rve ne r Shifting —A s and event u- bec om es starv ed for resource manysources, inc luding : special case of the “Shi ftin g the Bu rd en ” la y- Exa mpl e: all y die s out. the QW ERTY an dGKA’s Introduction •In nova tion Associates’ to e that occurs sys tem s archetyp whe n an ou t of typ ewrite r keyboa rds. Syste msThinking coursebooks in to he lp solve an inte rve ner is brought —A fi el d of stud y th at Syste m Dyna mi cs The Fift h Di scip li ne: The Ar t an d Pra cti ce of the • ongoing . Over tim e, as th e in te r- problem incl ud es a method ol og y fo r co ns tru cti ng ,byPe ter Sen ge Lea rn ing Orga niz ati on vener succ essf ully han dl es th e prob le m, com pute r sim ul atio n mode ls to ach ie ve Hig h Per fo rmance Systems’ Ac ad emic User ’s • the peo ple with in the sys tem becom e le ss bet ter un de rstan di ng of soci al an d corp o- Guide to STEL LA capable of solving th e prob le m the m- ra te syste ms . It dra ws on org an izati on al • The Ame ric an He ri tag e Di cti ona ry an d Th e even more dep en - selv es. They become stud ie s, behaviora l dec is ion the ory , an d Ran dom House Dic tio nar y. dent on the intervener . Exa mpl e: on goi ng eng in eeri ng to provi de a th eo reti cal an d con su lta nt s. use of outside emp irical ba se for struc tu rin g the re la- Simul ation Model —One of the 10 tool s of ti ons hi ps in com ple x sys tem s. sys tems thinking. A co mpu ter mod el th at Syste m —A gro up of in te ra ct ing , int err elat ed , lets you map th e relations hi ps th at are or inte rd epe nd ent eleme nt s for ming a 5 1 WWW.P EGAS USCOM. COM SYSTEMS THI NK IN G TOOLS PEG ASU S COMMU NICA TIONS , INC.

54 O O L B O X T T H E D O ’ S A N D D O N ’ T ’ S O F S Y S N K I N G T H E J O B O N T H I T E M S G M N A ODM O BY HAEL IC GENERAL GUIDELI NE S Cue s tha t no n-sys temic th ink ing is go ing o, yo u’ ve taken a sys te ms thin king S on : Phra ses suc h as “W e need to hav e im me - rea d a few co ur se— or may be you’ve DON’T us e sy st ems thi nki ng to furt he r di ate resul ts, ” “We just ha ve to do more of —a nd now you iss ues of The Syste ms Thin ker yo ur own ag enda . Sy stem s think ing is mos t wh at we di d la st time,” or “It’s ju st a matter wa nt to sta rt usi ng systems thi nk ing on the effe ctive whe n it is used to loo k at a pr oble m of tryi ng harder .” job. How do you be gin? Your best be t is to in a ne w way , no t to advo cate a prede ter - ap proa ch thi s ende avor in th e spi rit of mine d so lut io n. St rong advoca cy wil l cre ate GET TI NG STA RTED “l earn in g to walk befor e you ru n. ” He re are res is tan ce —b ot h to you r idea s, and to sy s- so me sugges ti on s: DON ’T attem pt to solv e a prob lem im me - tems thinkin g. It shou ld be use d in the spirit , muc h less dia tely. Do n’t expec t to re present of inqu ir y, no t inquis it io n. E OVE RAL L GU IDELIN unde rstand , persis tent and com plex syst emic DO use sy st ems thi nking to sift out The too ls of sys te ms th inki ng ar e best use d pr ob lem s ov ern ight . Th e tim e and co ncen - maj or iss ue s an d fact ors . as vehi cl es to promote team le arni ng in the tration requi re d should be pro port io nal to : Sy ste ms thin ki ng ca n hel p you Ben efi t or ga ni za tion . Wh ethe r you are doi ng “pa per the difficul ty and scope of the issue s br ea k th rough the cl utter of ev eryda y an d pen ci l” model s or creatin g ful l-fled ged invol ved. l pa tt er ns of ev ents to recogni ze genera microw orl ds, the pr oces s of con struc ting a fuller and More re alist ic goal: to achieve be ha vi or and the st ructur es that ar e pro duc - an d us in g mod els is primaril y ab out ex plo r- wide r unde rs tanding of th e pro blem . in g them . It al so he lps in sep ar ati ng sol u- ing and ex amini ng ou r “men tal mod els” — DO sta rt wi th sma ller -sca le pro blem s. ti on s fro m un derl yi ng prob lems. Too oft en the dee pl y he ld ass ump ti ons th at influ enc e attem pt to dia gra m the whole DO N’ T we id ent ify probl em s in ter ms of th ei r so lu - the way we th ink and act. sys tem —ot he rwise yo u’ll quick ly becom e ti on ; for ex amp le , “t he pro bl em is that we ove rwhelm ed. have to o ma ny ___ _____ Bette r : Try to foc us on a prob lem iss ue (fill in the bla nk: pe ople, and dr aw the mini mu m varia ble s and loops init iat iv es, steps in our pro - T E M S I D E N T I F Y I N G A S Y S P R O B L E M you’ ll need to capt ur e the prob le m. ce ss),” or “the prob lem is DO N’ T wor k wit h syst em s think ing tha t we ha ve too lit tle Th e pro blem of the following ALL shoul d have tech nique s “on lin e” under pressure, or in ___ __ ___ __ (r esour ces, ch ara cter ist ics: red for or fr ont of a group tha t is unprepa info rma tion, budg et . . .). ” 1. Th eissue is important to me and my busi - ness . int ole ra nt of the le ar ning proces s. use sys te ms DON’ T Th eproblem is chronic ,rather than aone- 2. Addit ionaldan ger: If the audience is no t thi nki ng to blame indi vi du - timeevent . fa milia r wit h the con cep ts and met hods of als . Chroni c, un res ol ved 3. Th eproblem has aknow nhistor ythat Ican sys tem s think ing , they mig ht not under - pro blems ar e mor e of ten desc rib e. st and tha t the pr oc ess re vea ls ment al mo d- the re sul t of syst emi c brea k- Prof its wer esteady for 2years, Exa mp le: butha vebeen decli ning for the last 6months. els , ca n be cont rov er si al, and is highly dow ns than indi vi dual mi s- Or:Pro duc tivit yrose rapidly until about ayear it er ativ e in na tur e. It is fa r more be neficial take s. Solut ions to thes e ago, wh en it leveled off. to ha ve the group enga ge in their own loo p pro blems lie at the syst emi c, Peo ple hav etried tosolve this problem 4. bu ilding aft er app ropria te ins truc tion and not th e in di vidual , lev el . before,wit hlit tle or no success. fo unda tion ha ve be en given. us e sy ste ms thi nk- DO char not hav e all of these does If you r pr oblem - DO deve lo p yo ur dia gra ms gradua lly ing to pr omote inqui ry and the fir st three), acteristics not it may (esp ecially and inf orma lly, in or de r to build confidence ved ch all eng e preconcei is. be approp riat e for a syste ms thi nk ing analys ap pro ach. Tr y redef ining it for a diffe rent in us ing sy st em s th in king . ide as. 5 2 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

55 uninte nded side -ef fect s of an inter vention. Goo dpra cti ce: Look at ne wspa per arti - persp ecti ve on the proble m. cles and try to draw a fe w loop s that ca pture “Toda y’s proble ms Ge ne ralprin cip le : Exa mple : Wi th a man ufa ct uri ng del ay of te n come from yester da y’ s solut ions .” the dy na mics of a problem bei ng de scribe d. prob le m, you mi gh t chec k wi th fi nanc e to Even bet ter : Tr y match ing a te mp la te to the Any solut ion is bound to have tr ade -offs, see if the re ar e an y dyn am ics in th e finan ce ar ti cl e. so us e sy stems think ing to exp lor e the aren a that are af fect in g the ma nu factu ring DON’ T imp lica tions of any propo sed solut ion worry abou t dr aw ing loops de la ys (cap ital inve stm en ts and pu rc ha ses, rig ht away . On e of the stron gest be ne fit s of be for e tr ying to im ple ment it. et c. ). Th e same can be don e for ma rk et ing, be su rpri se d if som e sit uat ions is that it can the system s th ink ing pers pective DO N’ T sa les, etc. he lp you lea rn to ask the righ t que stion s. def y sol uti on, esp eci ally if they are chr on ic wo rk ite rat iv ely. There is no “fina l” DO an t fir st st ep towa rd under - Thi s is an import prob lems. Rushi ng to act ion ca n thwar t mode l (s et of lo ops ). Loo pi ng is a lea rning sta nding a prob lem. lea rni ng and ul ti mat el y underm in e eff or ts pr oc ess tha t sho uld con ti nue to evolv e with to id enti fy hi gher lever ag e int erven tions. ne w dat a and pe rs pect iv es . DRAW DI AGRAMS ING Resi st the tende ncy to “s ol ve” the issue and DON’ T pre sent “fi nal” loo p dia gra ms DO sta rt with the proce ss of de finin g vari - fo cus on ga ini ng a deep er under sta ndi ng of as fin is he d pr odu ct s. abl es. DO ns. encou rage airi ng of as sumptio the str uc tures pr oduci ng the pr ob lem. Be Bette r : Pr ese nt as a tent ative and evolv - : be tte r share d unde rsta ndin g of a Benefi t war y of a symp toma ti c fix di sgui sed as a ing pic tur e of how you are se ein g th ing s. To is a ve ry effe ctiv e pr oblem. Diag ramming long- ter m, high-l ever age int er ven tion . ge t buy -in and ma ximi ze le arning , the audi - • to ol for promoting group in qui ry into a en ce nee ds to par tic ipa te in th e mode ling ic hael Go odman is a prin cip al of Inn ova tion M pr oblem or iss ue. pr oc ess . Ass oci ates, Fra min gh am, MA, an Arth ur D. Li ttle DO star t wi th a cen tra l loo p or pro ce ss . le arn fr om his to ry. Whe n po ss ib le , DO co mpan y. The mat eria l in this articl e wa s drawn from his 20 yea rs of expe rie nce in the field, as well as Then add loops to “fill in” de tai l. ch ec k dat a to see if your dia gra m correct ly bus in es s cou rs es deve lop ed by Innova ti on : Th e cent ral loo p ma y show Exam pl e de sc ribe s pas t behav io r. Ass oci ates. ho w the syste m is sup pose d to work, and the INT ERVEN TI ONS ad di ti on al loops can explore what is pu sh in g it out of whac k. DO ge t all stake holde rs inv olved in the DO N’T get bog ged dow n in de tai ls. proc es s. This will help ensur e that all St art simpl y, at a high le vel of genera liz a- vi ewp oints hav e bee n co n- tio n, bu t with enoug h detail to su m up the si der ed , and will im pr ov e obs erv ed be havior. the ac cep tance rate fo r th e G U I D E L I N E S F O R I N T E R V E N T I O N S Exam pl e : If you are explo ri ng the ca uses interve ntio n. of mis sed deli very dates in a factory , lump go for va gue, DON ’T Tobeeff ective ,aninterve nt ionmus tbeself- 1. to geth er th e ty pe s of produ cts th at are ex pe - gen er al , or op en- ended su stai ni ng ,self-cor recting and lon g-las ting .It ri enci ng sim ilar delays. sol utio ns such as “Imp ro ve mus tma ke long- termchan ge sin theper for- man cetren d. DO beg in by loo kin g fo r tem pl ates or com munic ations .” Ty pesofinter ven tio ns inaca usa lloop 2. clarif y the ge ne ral stru ct ures that might Bet ter : “Red uc e the dia gr am: pro blem . in form ati on del ay bet we en •Addalin k. : Sys tems archet ypes pro vi de a Adva nta ge sa les an d man uf act ur in g by •Breakalink. fo cal point or a st ory line to beg in th e pr o- cre at ing a new in for mat ion •Sho rtenade lay. ce ss of un der stan din g a prob lem. sy stem .” •Ma keagoalex plic it. wo rk with on e or mor e par tn ers . DO make an inter ven - DO •Slowdow nagr owt hproc ess ;relieve a Adv anta ge : Mu lt iple vie wpo ints add tio n spe cific , me as ur able , • limit ingpr oc ess . ri ch nes s an d detail to the und erstan ding of a an d ver ifiab le. Th ebe st interve nt ionislikelyto be aco mbi - 3. nat ionofinterve nt ion sapp lied ge nt ly an d pr oblem. Examp le : “Cut the inf or - pat ient ly. DO che ck with othe rs to se e if the y can mat io n de lay betw een sa le s 4. Avoid pus hi ng onast ruc tur efromthe ou t- add som e in sight or im prove upon your and ma nufa ct ur ing dow n to si de . di ag ram— es pe cial ly people in ot he r fun c- 24 ho ur s.” Lo okfor varianc ebe tweenlong -an dshor t- 5. have a dif feren t ti onal areas who might lo ok for pote ntial DO ter mimpact s,to ant icipat eune xpe cted effec ts. 5 3 THI NK IN G TOOLS WWW.P EGAS USCOM. COM SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

56 1 0 T O O L S T H E O N D I N G R R E A F U R T H E 1. Dou ble-Q Di ag ram 6. Struc tur e-Be havi or Pa irs Bas ed on TQC too l “C ause- and- Effect Diagr am.” See Aca demic User ’s Gu id e Ref er red to as “At oms of Str uct ur e” in (1 982) Guide to Quality Control , Ann Arbo r, Ish ika wa, Kaoru to STE LLA by Ba rry Richmo nd, publis he d (a s par t of softwa re MI: UNI PUB. NH . enta tion) by Hig h Per fo rm anc e Syste ms, Ha nover, docum Also , see Goodma Stu dyNo tes in Sy ste m n, Micha el (1 974 ) 2. Beh avior Ov er Time Di agram Dyn amic s , Waltha m, MA: Pega sus Com munica tions. Bas ed on dia gram s refer re d to as “r efer ence modes ” in sys - te m dyna mic s li terat ure. See Richardso n, Geor ge and Alexander 7. Polic y Structur e Diag ra m Pugh (19 81) Intro du ct ion toSyste mDy namics Mode ling , Cont act Pr of es sor John Mor ecr of t at the London Busines s Wa lt ha m, MA: Pega sus Communications. School. 3. Ca us al Lo op Di agra m 8. Com put er Model Pugh (1 981 ) See Rich ard son , Geor ge and Ale xander One of the be st sof tw are for building sys te m dy nam ics com - Intr odu ction to Sys tem Dyna mics Mode li ng , Waltham, MA: puter models (Ma cinto sh) is ithin k™ STELL A™ and by High Peg as us Com muni ca tions. Per for ma nce Syste ms , Hanove r, NH. For IBM -comp atib les, the re is PowerS im St ud io by Ve nta na Syst ems , and Vensim 4. Sys te ms Archet ypes En te rp ris e2000 by Powe rS im Cor p. , Ne w York: The Fifth Discipline See Sen ge, Peter (1 99 0) s appli - Dou bl eda y. Also cov ered regula rly with curr ent busines 9. Ma nag em ent Flig ht Si mula tors THE SYSTEMS t newsletter, cat io ns in a managemen Cont act Pr of es sor John Ster man at the M. I.T. Sloa n School TH INKE R catio ns, Inc., Communi , publi shed by Pegasus of Ma na gem ent (6 17 -2 53-19 51) for copie s of com put er simula - Wa lt ha m, MA. to rs on Peop le Exp res s, mana ging product life cy cles, rea l-e sta te ma na ge ment , and sup er tank er ma na ge ment . 5. Grap hica l Funct io n Di agram in system Bas ed on dia gram s refer re d to as “table functions” 10. Lea rning Labor at ory dyn am ics litera ture. See Richardson, Ge orge and Ale xande r Kim, Daniel (1 989) “Le ar ning Lab or ator ies : Desig ning a Intro du ct ion toSyste mDy namics Mode ling Pugh (19 81) , s of the 198 9 ,” Pro ceeding Ref lect ive Lea rning Env iro nment Wa lt ha m, MA: Pega sus Communications. Int er na tiona l Sys te m Dy na mics Conf er ence , Stut tg ar t, Ger ma ny: Spr ing er- Ver la g. 5 4 PEGA SU S COMMUN ICA TI ON S, INC. 78 1.398.9700 SY STE MS THI NKIN G TOO LS

57 I N D E X S Y S T E M S T H I N K E R T H E T O PA RT I: AN OV ERVI EW 6 Sys te ms Thi nk ing as a Lan gua ge byMich ae lR.Goo dma n V2 N3, Apri l 1991 ng: “F ire-Fi ghting” at Multiple Levels V4 N5, June/July 1993 8 Lev el s of Und erstandi V1 N3, August 1990 10 A Pale tte of Sy ste ms Thi nk in g Tools 13 PA RT II: DYNAMIC THIN KIN G TOOLS 14 Rei nfo rc in g an d Bala nci ng Loo ps : Building of Dynamic Sy stem s V1 N1, Apr il/ May 1990 Blocks 16 Balan cin g Lo ops with De la ys: Teet er-T ottering on Seesaws V1 N2, June/July 1990 18 Gui de li ne s for Dr awin g Cau sa l Loop Diagrams V3 N1, Febru ary 1992 20 Sys te ms Arc he types at a Gl an ce V3N4 , May 1992 23 PA RT III: ST RUCTUR AL THIN KING TOOLS V2N 8, Octo ber 1991 24 Fro m Ca usal Loops to Gr aph ic al Functions: Articulating Chaos 26 Graphi 1991 cal Fu ncti ons: “Se ei ng” the Ful l Story V2 N7, Sept ember 28 Stru ct ural Thi nking: to Accumulat ors and Fl ows V2 N2, March 1991 The Wo rl d According 30 Ac cu mu la to rs : Ba thtubs , Ba thtu bs, Everywhere V2 N1, Febru ary 1991 32 Ac cu mu la ti on Manage men t: Avo iding the “Pack Rat ” Syndrome V2N4 , May 1991 34 Del ays: Accum ulators in Disg ui se V2 N5, June/July 1991 36 S-Sh ape d Gro wth and the La w of Diminishing Returns V2 N3, April 1991 39 PA RT IV: COMPUTER TOOLS -B ASED 40 Mod el in g for Wha t Pur pos e? by Jay W. Fo rres ter V4N4 , May 1993 42 Manage men t Flight Sim ul at or s: V3 N9, November 1992 Flig ht Tra in in g for Man ag ers —Pa rt I 44 Manage men t Flight Sim ul at or s: V3N 10, Dec 199 2/Jan 1993 Flig ht Tra in in g for Man ag ers —Pa rt II 46 Learn in g Lab oratories : Pr ac ti ci ng Between Performances V3N 8, Octo ber 1992 49 PA RT V: REFERENCE GUI DE 50 Th e Voc ab ular y of Sy st ems Thi nk ing: A Pocket Guide V2 N10, Dec 199 1/Jan 1992 1992 on the Job of Sy st ems Thinking V3 N6, August 52 Th e Do’s an d Don’t’s by Mic hae lR.Good man 5 5 WWW.P EGAS USCOM. COM THI NK IN G TOOLS SYSTEMS PEG ASU S COMMU NICA TIONS , INC.

58 A B O U T T H E X R E P R I N T S E R I E S T O O L B O is a volume in the Tool box Repri nt Series. volumes Sy stems Thinki ng Tools :AUser’s Ref erence Guid e Other Sy stems Archetyp es I: Diagn osin gSystemic Issues an dDesign ingHigh-Leverage inc lu de Intervent ions, Sys tems Archety pes II: Usi ng System sArchetypes to Ta ke Effective Action ,Sys tems Archet yp es II I: Unders tandi ngPat terns an d of Beha vi orand Delay, All volumes are avail - The “Thin kin g” in System sThin king: Seven Essen ti al Sk il ls. ab le for $16. 95 each . As th ese book lets are often used in training and introduct ory cours es, volume di scou nts are avai lab le. See th e copyri ght page or call 1-8 00 -27 2-09 45 for det ails. ® The Toolbox Reprin t Series has been com piled from The Systems Thin ker New sl ett er, whic h present s a sy ste ms pe rspect ive on current issu es an d provides tools for framing probl ems in new and insightf ul systems The Sy stems Thi nk er wa ys. systems think ers, case st udies of syst ems thinking inc ludes artic les by leading impl emen ta tion, sof tware and book reviews, and num erous ot her colum ns geared to dif ferent level s of sy ste ms thin kin g abi lity . Indi vidual su bscription rates to The Systems Thi nk er are as fol lows: A speci al rate of 20 issue s for $169, or 10 issues (one year) for $10 9. Library sub script ions are avai lable for $1 89 for 10 issues (on e ye ar). Visit www. peg asuscom .c om for more information. 5 6 SYS TE MS THINKI NG TOO LS PEGASUS COMMUNICATIONS , IN C. 781.398.9700

59 P E G A S U S P U B L I C A T I O N S Ant ho lo gie s Ma na gin gthe Ra pids: Stor ie sfr om the For efro nt of the Lear ning Organization Ref lect io ns onCrea ti ng Le arn ing Or ganiz ati ons The New Workpla ce: Tran sfor mi ng the Cha ra ct er an dCulture of Our Organ izations Org an iz at io na lLea rning at Wor k: Embrac ing the Challen ges of the New Work place Ma kin gItHap pen: Stori es fr om In side the New Work place The Pega sus Work boo k Se ries Sys tems Arc he type Basics: Fr om Stor yto Str ucture Systems Thinking Basics :From Con cepts toCausal Loops The Lea rn er’s Pa th: Practic es for Rec over ing Kno wers The “Bill ibonk” Series Frankl’s “Thorn Patch” Field book Bi llib onk &the Thor nPatch Bi llib onk &the Big Itc h Frank l’s “Big Itch” Fieldbook Lear ni ng Fabl es Out lea rni ng the Wol ve s: Sur vi vi ng and Thri vi ng in aLearning Organization Illum inatin gthe Beliefs That Limit Our Or gan iz ations Shad ow softhe Ne anderthal: Dil emma: Living with Pur pos e, Lead ing with Vision The Lemming ion The Ti pof the Ice berg: Man agin gthe Hid den Fo rces That Can Make or Break Your Organizat Lis ten in gtothe Vo lc ano: Con versations That Op enOur Min ds to New Poss ibilities Oth er Tit les Huma nDy nami cs: ANe wFrame work for Und erstanding People and Realizing the Potential in Our Organizations in Our World Through Wh en aButt erfl ySneeze s: AGui de for He lpi ng Kids Explore Inter connections Stories Favorite The Inn ova ti ons in Manageme nt Ser ie s From Mecha nist ic toSoc ial Syste mic Thin ki ng :ADigest of aTalk by Russell L. Ackoff Appl yin gSys tems Archetyp es Tow ar dLe arning Organiz ation s: In tegrati ng Total Quality Control and Syst ems Thin king Des ign in gaSy st ems Thin kin gInte rve ntion: ASt ra tegy for Leveraging Chan ge The Nat ura lSt ep: AFrame wor kfor Ac hie ving Sustainability in Our Organizations Anxi ety in the Wo rkplace: Usin gSy ste msThi nki ng toDee pen Un derstanding The Soul ofCorpor at eLead er ship: Guid el in es forValue s-Cente red Governan ce an dSocial Challenges Crea ti ng Sus ta ina ble Organ iz ation s: Mee ting the Economic ,Ecological, of the 21st Century Crea ti ng Va lue: Linking the Inte re sts of Cus tomers, Employees, and In ve stors Rel in kin gLife and Work :Toward aBe tt erFut ure Fac in gth eCo mpe ti ti on: AnOr ganiz ati on Mobi li zes for Large-Scale Change Journey Org an iz at io na lCha nge at Phil ips Display Com pon ents :Ref lections on aLearning Intro duct io nto Sy st ems Thin kin g Reb oun ding ,Re buildi ng, Re ne win gat She ll Oil: AFormer CEO Reflects on Large-Scale Change Rei nv en ti ng Human Resourc es at L.L.Be an: Lessons for Learnin gand Change The Ess en tia ls ofSe rvant-L ead er ship: Pr inc ip les inPractice Dia lo gue atWork: Skil ls for Le veraging Col lecti ve Intelligence The Tool box Repri nt Se rie s Sys tems Arc he types I: Diagnosin gSy ste mic Issues and Designing High-Leverage Interventions Sys tems Arc he types II: Usin gSy ste ms Arche typ es toTake Effective Action Sys tems Arc he types III: Und er stan din gPatterns ofBehavior and Delay Sys tems Think ing To ol s: AUse r’ sRefe re nc eGuid e The “Th in king” inSystem sThink ing: Se ven Es sential Skills E-N ewsle tte rs ® ® THE SY STEMS THI NK ER EPOINTS LEVERAG for aNe wWor kp lace ,Ne wWor ld 5 7 PEG ASU S COMMU NICA TIONS , INC. WWW.P EGAS USCOM. COM SYSTEMS THI NK IN G TOOLS

60 Pega sus Com muni cations, In c. resources that help people explore, underst and, artic ulate, and is ded icat ed to providing ing the complexities of a changing world. Since 19 89, Pegasus has worked to addr ess the cha lle ng es the y fac e in manag buil d a commu nity of pr ac tit ione rs thr ough newslett ers, book s, audio and video tapes, and it s annual Sy ste msThinking ® in Actio n Conf er ence and othe r eve nts . For mor e information, contact us at: Peg asus Com mu nic at ions , Inc . • One Moody Street • Waltham, MA 024 53-5339 USA • www.pegasuscom. com Orde r Ph one: 800- 27 2-09 45 / 781 -3 98 -97 00 • Fax: 781 -894 -7 175 5 8 SY STEM S THI NKIN G TOO LS PEG AS US COMMUN IC ATI ON S, INC. 78 1.398.9700

Related documents

DoD7045.7H

DoD7045.7H

DoD 7045.7-H EPARTMENT OF D EFENSE D F UTURE Y EARS D EFENSE P ROGRAM (FYDP) S TRUCTURE Codes and Definitions for All DoD Components Office of the Director, Program Analysis and Evaluation A pril 2004

More info »
Controller Tool Help

Controller Tool Help

Controller Tool Help No. LIT-1201 1147 Code 13.0 Release Software 2018 December Issued for the most up-to-date Exchange of this document. Refer to the Knowledge website version 21 Getting ... Started ...

More info »
Dell EMC Unisphere for PowerMax Online Help

Dell EMC Unisphere for PowerMax Online Help

™ ™ Dell EMC Unisphere for PowerMax Version 9.0.0 Online Help (PDF version)

More info »
UL White Book

UL White Book

GUIDE INFORMATION FOR ELECTRICAL EQUIPMENT THE WHITE BOOK 2015-16 UL PRODUCT CATEGORIES CORRELATED TO THE 2011 AND 2014 NATIONAL ELECTRICAL CODE® UL’s General Guide Information is updated daily. To co...

More info »
GAO 09 232G Federal Information System Controls Audit Manual (FISCAM)

GAO 09 232G Federal Information System Controls Audit Manual (FISCAM)

nt Accountability Office United States Governme GA O Februar 2009 y FEDERAL INFORMATION SYSTEM CONTROLS AUDIT MANUAL (FISCAM) GAO-09-232G

More info »
Security and Privacy Controls for Federal Information Systems and Organizations

Security and Privacy Controls for Federal Information Systems and Organizations

-53 NIST Special Publication 800 Revision 4 Security and Privacy Controls for Federal Information Systems and Organizations JOINT TASK FORCE TRANSFORMATION INITIATIVE This publication is available fre...

More info »
Nastran Dmap Error Message List

Nastran Dmap Error Message List

Overview of Error Messages NX Nastran displays User Information, Warning, and Error messages in the printed output. The amount of information reported in a message is controlled by system cell 319. Wh...

More info »
Thriving on Our Changing Planet: A Decadal Strategy for Earth Observation from Space

Thriving on Our Changing Planet: A Decadal Strategy for Earth Observation from Space

TIONAL ACADEMIES PRESS THE NA This PDF is available at http://nap.edu/24938 SHARE     Thriving on Our Changing Planet: A Decadal Strategy for Earth Observation from Space DET AILS 700 pages | 8.5 ...

More info »
2017 NAICS Manual

2017 NAICS Manual

ORTH N A MERICAN I NDUSTRY LASSIFICATION C YSTEM S United States, 2017 EXECUTIVE OFFICE OF THE PRESIDENT AND BUDGET OFFICE OF MANAGEMENT

More info »
am08.dvi

am08.dvi

Feedback Systems An Introduction for Scientists and Engineers ̊ Karl Johan Aström Richard M. Murray Version v2.10b (February 22, 2009) This is the electronic edition of Feedback Systems and is availab...

More info »
Diagnosis Related Groups in Europe

Diagnosis Related Groups in Europe

Diagnosis Related Groups...pb24.5_Diagnosis Related Groups...pb24.5 23/10/2011 11:32 Page 1 European on Health Systems and Policies Series Observatory Diagnosis-Related Groups in Diagnosis-Related Gro...

More info »
MVS Diagnosis: Tools and Service Aids

MVS Diagnosis: Tools and Service Aids

z/OS Version 2 Release 3 MVS Diagnosis: Tools and Service Aids IBM GA32-0905-30

More info »
ASAALT Weapon Systems Handbook 2018

ASAALT Weapon Systems Handbook 2018

AMERICA’S ARMY: TODAY’S MODERNIZATION IS TOMORROW’S READINESS

More info »
Joint Software Systems Safety Handbook

Joint Software Systems Safety Handbook

DEPARTMENT OF DEFENS E JOINT SOFTWARE SYSTEMS SAFETY ENGINEERING HANDBOOK ----------------------------------------- THE JOINT SOFTWARE S YSTEMS SAFETY DEVELOPED BY ENGINEERING WORKGROUP Original publi...

More info »
book.dvi

book.dvi

Fuzzy Control Kevin M. Passino Department of Electrical Engineering The Ohio State University Stephen Yurkovich Department of Electrical Engineering The Ohio State University ddison-Wesley Longman, In...

More info »
Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation

Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation

MANAGING THE RISKS OF EXTREME EVENTS AND DISASTERS TO ADVANCE CLIMATE CHANGE ADAPTATION SPECIAL REPORT OF THE INTERGOVERNMENTAL PANEL ON CLIMATE CHANGE

More info »
quality cancer care

quality cancer care

This PDF is available from The National Academies Press at http://www.nap.edu/catalog.php?record_id=18359 Delivering High-Quality Cancer Care: Charting a New Course for a System in Crisis Laura Levit,...

More info »
MCO 1200.17E MILITARY OCCUPATIONAL SPECIALTIES MANUAL (SHORT TITLE: MOS MANUAL)

MCO 1200.17E MILITARY OCCUPATIONAL SPECIALTIES MANUAL (SHORT TITLE: MOS MANUAL)

DEPAR T MENT THE NAVY OF ADQ UARTE UNI T ED ST ATE S MAR INE CORPS HE RS RINE COR N PS PENT 3000 MA AGO 20350-3000 NGTON, HI D.C. W AS 7E 00 .1 12 MCO 465 c AUG 0 8 013 2 ORDER 1200.17E MARINE CORPS C...

More info »
Justification Book

Justification Book

UNCLASSIFIED Department of Defense Fiscal Year (FY) 2019 Budget Estimates February 2018 Office of the Secretary Of Defense Defense-Wide Justification Book Volume 3B of 5 Research, Development, Test & ...

More info »