Overcoming the improvement paradox

European Management Journal Vol. 17, No. 2, pp. 120–134, 1999 1999 Published by Elsevier Science Ltd. All rights reservedPergamon

Printed in Great Britain0263-2373/99 $19.00 1 0.00PII: S0263-2373(98)00072-3

MANAGEMENT FOCUS

Overcoming theImprovement ParadoxELIZABETH KEATING, MIT Sloan School of ManagementROGELIO OLIVA, Harvard Business SchoolNELSON REPENNING, MIT Sloan School of ManagementSCOTT ROCKART, MIT Sloan School of ManagementJOHN STERMAN1, MIT Sloan School of Management

Despite the demonstrated benefits of improvementprograms such as total quality management andreengineering, most improvement programs end infailure. Companies have found it extremely difficultto sustain even initially successful process improve-ment programs. Even more puzzling, successfulimprovement programs sometimes worsen businessperformance, triggering layoffs, low morale, andthe collapse of commitment to continuous improve-ment. We term this phenomenon the ‘ImprovementParadox.’ For the last four years, we have workedwith a variety of firms to understand the paradoxand design policies to overcome it. Our findingssuggest that the inability to manage an improve-ment program as a dynamic process is the maindeterminant of program failure. Improvement pro-grams are tightly coupled to other functions andprocesses in the firm, and to the firm’s customers,suppliers, competitors and capital markets. Failureto account for the feedbacks among these tightlycoupled activities leads to unanticipated and oftenharmful side effects. We describe these dynamicsand offer some guidance for managers seeking todesign sustainable process improvement programs. 1999 Published by Elsevier Science Ltd. Allrights reserved

Introduction

Process improvement has become an imperative forbusinesses seeking competitive advantage, yet it isdisturbing how few organizations make lasting andsuccessful use of process improvement tools such astotal quality management and reengineering. Thesetools should help to raise productivity, boost qualityand enhance competitiveness. However, quality pro-grams often struggle to gain initial acceptance and to

European Management Journal Vol 17 No 2 April 1999120

sustain continuous improvement (US GeneralAccounting Office, 1991; Young, 1991a, b). Despitethe demonstrated benefits of many improvementtechniques, most attempts by companies to use themhave ended in failure (Easton and Jarrell, 1998). Infact, companies have found it extremely difficult tosustain even initially successful process improvementprograms. Even more puzzling, successful improve-ment programs have sometimes led to declining busi-ness performance, causing layoffs, low morale, andthe collapse of commitment to continuous improve-ment. We term this phenomenon the ‘ImprovementParadox.’

If improvement tools were ineffective it would beeasy to explain their low use. The evidence, however,does not support that explanation. Firms that winquality awards have higher share-holder returns(Hendricks and Singhal, 1996). Easton and Jarrell(1998) found that among the top 1000 publicly-heldcompanies in the United States, firms with welldeveloped quality programs significantly outperformtheir counterparts in profitability, share price andreturn on assets.

These large sample results are consistent with ourown findings. In hundreds of hours of interviewswith our partner companies, discussing both success-ful and unsuccessful programs, we rarely heard ‘theprogram was just no good.’ Typical comments onstalled or abandoned programs were ‘I believe [aparticular program] is a good process. Some day I’dreally like to work on a project that actually followsit’ and ‘We’ve left a lot on the table by letting thisprogram go.’ Our findings suggest that the inabilityto manage an improvement program as a dynamicprocess — one tightly coupled to other processes inthe firm and to the firm’s customers, suppliers, com-petitors and capital markets — is the main determi-

OVERCOMING THE IMPROVEMENT PARADOX

nant of program failure. Failure to account for feed-back from these tightly coupled activities leads tounanticipated, and often harmful, side effects thatcan cause the premature collapse and abandonmentof otherwise successful improvement programs. Wedescribe these dynamics and offer some guidance formanagers seeking to design sustainable processimprovement programs.

For the last four years we have worked with man-agers at Ford Motor Company, Harley-Davidson,Lucent Technologies, and National SemiconductorCorporation to understand why improvement pro-grams often fail, and how practitioners can designsustainable improvement programs (Jones et al., 1996;Sterman et al., 1996). This work extends earlierresearch on the paradoxically poor financial perform-ance experienced by Analog Devices shortly after ahighly successful manufacturing improvement pro-gram (Sterman et al., 1997). Our research involveddetailed field studies with our partner organizations.We stressed multiple data sources including exten-sive interviews and archival data on the various met-rics of quality, product histories, internal companymaterials, and financial results. We used the systemdynamics method (Forrester, 1961) to understand themultiple feedback mechanisms that affect theimplementation of improvement programs, and toformulate integrative formal models to test ourhypotheses.

Our findings span both the internal dynamics of animprovement program and the interactions of a pro-gram with forces outside the intended area ofimprovement focus. We first describe the internaldynamics of an improvement program and themanagerial challenges they create. We then examinehow an improvement program interacts with otherimprovement initiatives, other organizational units,and with customers. Other improvement programs,organizational practices, and market response havea profound influence on whether programs can besustained and contribute to the improved perform-ance of the entire company.

Internal Dynamics of ImprovementPrograms

Well-functioning quality programs cannot be bought,like a machine tool. No one can go out and purchasea fully functioning 6-sigma quality program. A com-petence in improvement must be grown organically.To do so management must grapple with three cen-tral issues. First, managers need to address the funda-mental trade-off between current and future per-formance levels. Second, managers need to makesure that the source of commitment to ongoingimprovement effort shifts from managerial actions toemployee initiative. Finally, as a program succeeds,

European Management Journal Vol 17 No 2 April 1999 121

and so-called low-hanging fruit is harvested, man-agers need to adapt their improvement tools andmanage expectations for continued gains.

Fundamental Improvement Trade-Off

Process improvement theorists assert that theemployees doing a job are the best-informed expertsand should be responsible for identifying improve-ment opportunities and implementing changes(Deming, 1986; Ishikawa, 1985; Juran, 1969). Accord-ingly, most improvement initiatives rely on theemployees who perform the day-to-day work both toguide the improvement program and make the actualimprovements. The rationale behind this strategy istwo-fold. First, employees already understand theirprocess, reducing data collection and diagnosis time.Second, employees have a strong interest inimplementing changes when they develop the pro-posals themselves. Deming (1986) argued, in what hecalled the ‘productivity chain’, that resources freedup by productivity gains should be reinvested intothe search for still greater improvements, creatingself-reinforcing feedback stimulating continuousimprovement. Operationally, effort allocated toimprovement raises productivity, boosting processthroughput, thereby lowering production pressureand yielding still more time for improvement (loopR1 in Figure 1). An organization that reinvests earlyimprovement gains in further improvement effortcreates a powerful positive feedback that generatesever-greater gains in quality and productivity.

However, reliance on operating employees to guideand implement improvement can limit the reinforc-ing process of the productivity chain. Imagine a qual-ity program designed to reduce defects and boostusable output. It takes time for improvement effortto bear fruit. Therefore, the first effect of an increasein improvement effort is a reduction in the timeemployees can devote to throughput. The short runeffect of improvement effort is a decline in output,exactly the opposite of the goal. As throughput falls,pressure to work harder builds. Employees facedwith high pressure to meet throughput goals will beforced to cut back the time devoted to improvement,boosting output but stalling productivity and qualitygrowth (the ‘effort squeeze’ loop B1 in Figure 1). Amanager in one plant we studied captured thedilemma clearly:

In the minds of the [operations team leaders] they had tohit their pack counts [daily quotas]. This meant if you werehaving a bad day and your yield had fallen... you had torun like crazy to hit your target. You could say ‘you aremaking 20% garbage, stop the line and fix the problem’,and they would say, ‘I can’t hit my pack count withoutrunning like crazy.’ They could never get ahead of thegame. (Repenning and Sterman, 1997)

To overcome the quandary, process improvementadvocates discouraging numerical throughput quotas


Figure 1 Improvement can be Self-reinforcing.Arrows Indicate the Direction of Causality: Signs (‘ 1 ’ and ‘ 2 ’) at Arrowheads Indicate the Polarity of Relation-ships: A ‘ 1 ’ Denotes that an Increase in the Independent Variable Causes the Dependent Variable to Increase,Ceteris Paribus (and a Decrease Causes a Decrease). Similarly, a ‘ 2 ’ Indicates that an Increase in the Inde-pendent Variable Causes the Dependent Variable to Decrease. Time Delays are Indicated in the Diagram by aDelay Box. Reinforcing Loop Polarity (Denoted by R in the Loop Identifier) Indicates a Self-reinforcing (Positive)Feedback Process. Balancing (B) Loop Polarity Indicates a Regulating (Negative) Feedback Process

and encouraging employees to allocate a portion oftheir normal workday to improvement effort(Deming, 1986). Managers can also reduce through-put pressure by adding more resources, therebyallowing sufficient time for both throughput andimprovement, or by lowering desired throughput(perhaps by increasing prices or reducing the numberof new projects undertaken). Managers must, how-ever, be prepared for a period when throughput willdrop or costs rise. While throughput drops immedi-ately at the start of an improvement effort, pro-ductivity only rises after the substantial delay inorganizing and deploying improvement efforts. Thedelay between allocating time to improvement andobtaining results, combined with the immediate dropin throughput, implies that performance will followa ‘worse-before-better’ pattern.

Du Pont’s efforts to improve maintenance and equip-ment availability (Carroll et al., 1998; Sterman et al.,1992) provide a clear example of this worse-before-better pattern (Figure 2). After a worldwide bench-marking study, Du Pont managers found that theirplants had maintenance expenditures considerablyhigher than best practices, while machine reliabilityand equipment availability were considerably lower.Further diagnosis showed that most maintenanceeffort was reactive, with insufficient effort devotedto preventive maintenance (PM), training, spare partquality, and design improvements. Over time, costcutting had slashed training and PM. Less preventivemaintenance increased the breakdown rate, pullingstill more resources out of PM and causing a stillgreater increase in breakdowns: the productivity


Figure 2 Cost Savings at a Typical Du Pont Plant AfterImplementing a Preventive Maintenance Program.

chain (loop R1) operated as a vicious cycle. The highbreakdown rate meant the cost of maintenance washigher, and equipment availability lower, than beforethe cost cutting began.

Escaping from the reactive maintenance trap requiresa large increase in PM. But the first impact of anincrease in PM is a decline in equipment availabilityand an increase in maintenance costs. Only after sometime will the benefits of PM start to show up inreduced breakdown rates. Many prior improvementprograms had failed because management could notunderstand or tolerate the initial drop in availabilityand rise in costs.

Stimulated by a system dynamics model, Du Pontcreated a training program designed to break out of


the reactive maintenance trap. A key part of the pro-gram was helping people understand the worse-before-better tradeoff. Figure 2 shows the cost sav-ings for a typical plant following introduction of theimprovement program. As expected, the initialimpact of the intervention was a rise in costs. Oncethe mean time between failures (MTBF) of equipmentbegan to rise, there were fewer breakdowns to repair,freeing up still more time for PM and boostingreliability. Equipment availability rose while mainte-nance costs fell: the productivity chain now operatedas a virtuous cycle. Plants adopting the programexperienced sharp increases in the rate of improve-ment. Mean time between failure for pumps roseabout 15 per cent with each doubling of cumulativeexperience, and costs ultimately fell by an average ofabout 20 per cent. Comparable plants pursuing tra-ditional approaches saw learning rates of only about5 per cent and a 7 per cent rise in maintenance costs(Carroll et al., 1998).

While some Du Pont facilities overcame the short-rundeterioration in performance caused by improvementprograms, many firms do not. Short-term perform-ance goals tempt managers to harvest productivitygains by cutting costly resources or raising through-put targets. These actions intensify throughput press-ure and shift effort away from improvement. Man-agers who fail to allocate enough resources findimprovement efforts stall as workers devote theirtime to short-term throughput goals.

The tendency to harvest initial productivity gains bydownsizing or increasing throughput objectives isstrong. Anticipated productivity gains are often fac-tored into equipment and labor planning. In severalof the firms we studied, expected productivity gainsfrom improvement efforts were assumed in pro-jecting future labor and capital requirements. Butgoals based on unrealized productivity gains will bea source of throughput pressure and ‘effort squeeze’(loop B1 in Figure 1). In one case, the staff allocatedto a critical activity was cut, based on projected pro-ductivity gains from the improvement program. Butthe reduction in resources ensured that throughputpressure remained high, preventing the workforcefrom devoting sufficient time to improvement. Theanticipated gains were never realized. Long leadtimes for new capacity and training delays for newemployees meant that even after management real-ized their error it was too late to compensate. Theimprovement program failed.

To sustain a program, managers must support thereinforcing nature of improvement by limiting theeffect of throughput pressure on effort allocation. Animprovement program is more likely to succeed ifmanagers facilitate a shift of employee time fromthroughput to improvement and limit opportunitiesfor employees to shift effort back toward throughput.For example, in Ford’s electronics division, managersprevented assembly workers from stealing time from


improvement to boost short-run output by limitingthe release of materials to the production floor. Oncethe materials allocated for the shift were used, oper-ators could not run their machines and turned theirattention to improvement opportunities.

Initiating and Sustaining Employee Commitmentto Improvement

Freeing employees to improve processes is essentialbut insufficient. Successful improvement requires theenthusiastic commitment of employees sinceimprovement activity is less structured and less eas-ily monitored than throughput. Shiba et al. (1993) dis-tinguish between two sources of commitment forimprovement programs: managerial push andemployee pull. Managerial push refers to efforts topromote improvement effort or mandate partici-pation. These actions range from inspirationalspeeches about the importance of improvement tomandatory participation in training and improve-ment teams to financial incentives and performancereview criteria based on improvement. Employee pullarises when workers come to understand the benefitsof improvement for themselves and commit them-selves to improvement effort independent of (andsometimes despite) management attitudes and sup-port (Schaffer and Thomson, 1992). Our fieldworksuggests that developing employee pull is essentialto sustaining improvement efforts.

Programs brought in by a high-level championrequire a certain amount of management push tobegin building commitment. Push techniques includeproviding training, demonstrating support, cham-pioning the value of the program, providing incen-tives, and clarifying the need to improve. Forexample, when Lucent Technologies launched the‘Achieving Process Excellence’ (APEX) initiative toreduce product development time, the efforts andpersistence of project leader Al Hofmann and othersserved as an initial push for participation. Lucent(while still part of AT&T) had attempted severaltimes to apply various quality management tools toits product development process. These initiativeshad failed, and many engineers in the organizationwere highly sceptical of the quality movement. Toovercome this initial resistance, Hofmann lobbied hispeers and his superiors to secure funding and releasetime from other responsibilities for the team mem-bers. He regularly sent notes of appreciation to parti-cipants and encouraged managers to recognize theirstaff members for APEX work. One team memberrecalled:

At first, people were ‘volunteered’ for teams by their man-agers, and people felt that APEX was a ‘flavor of themonth’ project... Hofmann put in a huge amount of per-sonal time. He was committed to spend one-third of histime, but he often spent more. He worked hard to showthat management cared. People then believed that manage-ment was serious. (Keating and Oliva, forthcoming)


Managerial push often creates temporary excitement,but must be replaced by other sources of motivationwhen that excitement begins to fade. Even the mostenthusiastic manager cannot personally contacteveryone in a large organization, so as improvementactivity spreads, the impact of individual leadersdeclines. Command-and-control structures aredependent on managerial supervision. They areunlikely to work in settings where employee partici-pation and contributions are difficult to monitor andassess. Participants in failed efforts commonly reportbeing unable or unwilling to continue after the pro-gram champion was promoted or reassigned.Employees accustomed to command-and-controlmanagement may never fully comprehend the pro-gram’s underlying logic or embrace its goals. Partici-pation becomes a matter of compliance to minimizeconflict with superiors. When the push to participateis removed, compliance fades.

All successful initiatives we observed were driven byanother self-reinforcing feedback. Initial commitmentto a program, perhaps stimulated by managementpush, motivates improvement effort. With somedelay, that effort leads to results. As employees seethat the improvement process actually works, theystart to believe it has some value, increasing commit-ment further in a self-reinforcing feedback (the‘employee pull’ loop R2 in Figure 3). At Lucent, thisfeedback soon took off:

As a result of the growing success of APEX, people startedto ‘self-identify’ themselves for teams. These folks werereally motivated; they generally came to the team withideas that they were eager to implement. They were theones who actively got things done.

No amount of management push can substitute forthe self-reinforcing feedback created as results mot-

Figure 3 Self-reinforcing Feedback Drives Employee Commitment to Improvement


ivate more people to participate, thus generatingmore results. Management reshuffling regularlystrips away program champions and replaces themwith managers who may not share the interests orthe skills of those who initiated existing programs. Inthese cases, it is the commitment of a stable set ofemployees that maintains the improvement effortsover time.

The employee pull feedback can function as a virtu-ous cycle (improvement boosts commitment, stimul-ating still more effort and improvement), or as avicious cycle (poor results lead to less effort, ensuringstill worse results).

As shown in Figure 3, a variety of factors can inter-fere with the employee pull feedback raising the oddsof a vicious cycle. First, complex processes are moredifficult to improve (a complex process has a long‘improvement half-life’ — the time required to cutdefects in half), intensifying the worse-before-bettertradeoff and slowing the growth of commitmentthrough employee pull. Second, the effectiveness ofany improvement effort depends on the scope of theinitiative and the adequacy of the chosen improve-ment methodology. Quality and reengineering toolsare more highly developed for manufacturing andoperations than for complex processes like productdevelopment, customer – vendor partnering, andsenior management functions. Third, inadequate sup-port infrastructure or training in improvement tech-niques limits the effectiveness of improvement effort.Finally, low job security can destroy commitment toimprovement — workers may shun improvementactivity if they believe productivity gains will lead tolayoffs. Each of these factors must be managedappropriately to generate the self-reinforcing com-mitment required for continuous improvement. Weconsider each constraint in turn.


Employee Perception of BenefitsEmployee commitment to improvement increaseswhen workers and line managers perceive thatimprovement effort works. In making the judgmentthat a program works, people compare the rate ofprogress they observe to their expectations. Commit-ment rises if progress is high relative to aspirationsand falls when progress is disappointing. The depen-dence of commitment on observed progress meansaggressive managerial push can adversely affect pro-gram success. For example, managers often setaggressive ‘stretch objectives’, or so-called ‘BHAGs’(Big Hairy Aggressive Goals) to encourage partici-pation in improvement programs (Hamel and Prah-alad, 1989; Collins and Porras, 1994). For example, asenior manager in a large manufacturing firm set anaggressive goal to reduce manufacturing cycle timefor the firm’s plants from 18 days to 1 day in fiveyears. He explained his rationale:

I didn’t know if we could get to one day, and, to be honest,I really didn’t care. If a facility was at 18 days and I set anobjective of 16.5 days, everybody would have just squeezeda little bit. This way everybody knew they had to make bigimprovements. If they only made it to two days or threedays that still would have been a lot better than 16.5.(Repenning and Sterman, 1997)

While aggressive objectives may be helpful in cre-ating initial push, they can undermine the develop-ment of long-term employee-pull effects. When objec-tives are set too high, expectations outstrip observedbenefits and commitment falls, weakening theemployee-pull feedback. As shown in Figure 4,stretch objectives can motivate greater effort byincreasing people’s aspirations as people ‘Rise to theChallenge’ (loop B2 in Figure 4), but create a ‘Credi-bility Gap’ that undermines effort when set too high(loop R3). As effort falls, performance suffers, con-firming people’s belief that the goal was infeasible.A vicious cycle of goal erosion and cynicism can setin. Few organizations undertaking ambitious changeprograms know what feasible rates of improvementin key metrics might be. It is all too easy for man-

Figure 4 Stretch Objectives and Commitment


agers, themselves under pressure to meet the expec-tations of Wall Street or catch up with competitors,to set objectives far beyond the feasible range, lead-ing to frustration and eroding their credibility.

In the case above, the 1-day cycle time stretch objec-tive was not well received, as the program leaderrecalled: ‘When I first announced [the one day objec-tive], a lot of people just doubled over in pain. Theyreally thought I was crazy.’ One member of the affec-ted staff commented ‘ At the time I thought to myself,“I really hope I get transferred within the next fiveyears because this is never going to happen. This guyjust does not understand [our business].”’

To avoid frustration and abandonment of theimprovement program, the senior manager allowedthe improvement teams to define cycle time metricsand set their own goals at the operational level.Employees and lower level managers workedtogether to measure the ‘touch time’ for each product(the time actually spent working on a product) as afraction of overall cycle time. They were shocked tofind that touch time was less than one per cent oftotal cycle time. Reducing cycle time from eighteendays to one would still leave touch time less thantwenty per cent. Involving workers in the definitionand measurement of cycle time and touch timehelped everyone realize the goal was feasible, weak-ening the credibility gap loop. The program was amajor success as initial results led to greater commit-ment and still lower cycle time throughout thedivision (Repenning and Sterman, 1997).

Complexity and the Improvement Half-LifeObjective setting requires an answer to the questionof ‘when’ the goal will be achieved as well as to thequestion of ‘what’ the goal is in absolute terms. Care-ful analysis of a firm’s processes and benchmarkingof other firm’s processes can help to set aggressiveyet realistic ‘what’ goals for performance. These tech-niques do not, however, provide a sense of how longit will take to move from current to desired perform-


ance. People commonly underestimate the scope anddifficulty of tasks (Kahneman and Lovallo, 1993) andexpect benefits sooner than is reasonable. Thus evenwhen the absolute goal (e.g. a one day cycle time) isset appropriately, the time provided to reach the goalis often far too short.

Reasonable estimates of the time required to improveare needed to allocate sufficient resources toimprovement. An organization that continuallyunderestimates the time required to realize resultsdevelops a demoralized and sceptical workforce thatdiscounts managerial promises and will rely moreheavily on tangible benefits. In one firm we studied,the workers had been through so many quality pro-grams that they described each new one by the acro-nym ‘AFP’ — Another ‘Fine’ Program.

One helpful concept in developing program objec-tives is the improvement half-life. Schneiderman(1988) found that in a wide variety of firms, ‘anydefect level, subjected to legitimate QIP [qualityimprovement processes], decreases at a constant[fractional] rate.’ The result is an exponential declinein defects characterized by the ‘improvement half-life’ — the time required for defects to fall by 50 percent (Figure 5).

The basis for the half-life dynamic is the Plan-Do-Check-Act learning loop or ‘PDCA cycle’ at the heartof TQM and other improvement techniques (AnalogDevices, 1991; Shewhart, 1939). Improvement teamscycle around the PDCA learning loop. With eachcycle a team identifies and eliminates the largestremaining root cause of defects, then moves on to theremaining sources of defects. The fractional rate ofdefect reduction each month (R) is the product of thefractional improvement per cycle (I) and the numberof cycles per month (L).

R 5 I·L (1)

Consequently, the rate at which defects are generated(D) falls exponentially towards the theoretical mini-mum level Dmin at the fractional rate R(Schneiderman, 1988; Sterman et al., 1997; Zangwilland Kantor, 1998):

D(t) 5 Dmin 1 (D0 2 Dmin)exp( 2 Rt) (2)

Figure 5 Semiconductor Fabrication Defects After theStart of a TQM Program. The Improvement Half-life isAbout 9 Months


The faster the rate of improvement R, the shorter theimprovement half-life will be (the half-life is givenby ln(2)/R). Improvement half-lives vary across pro-cesses and functions. Simple processes like a singlepiece of equipment in a manufacturing cell werefound to have half-lives on the order of a fewmonths. Complex processes, like product develop-ment, have half lives on the order of several years.Schneiderman (1988) found that improvement half-lives grow with the technical and organizationalcomplexity of a process (Figure 6). Greater technicalcomplexity slows the improvement rate due to diffi-culties in designing, conducting, and interpretingexperiments that might reveal approaches to defectreduction. Organizational complexity refers to thenumber and type of people, from different organiza-tional functions, required to carry out an effectiveimprovement effort for the process. Coordinating themarketers, product architects, designers, suppliers,finance people, and others involved in product devel-opment takes longer than getting the operators andmechanics responsible for a particular machine tomeet. Greater complexity reduces both the learningper cycle and the number of improvement cycles theorganization can carry out per month, slowingimprovement.

Several characteristics can be used to determine thetechnical and organizational complexity of a processor function. Technical complexity is higher when (a)the process cycle time is long, (b) it is difficult to per-form experiments, or (c) a high degree of technicalknow-how is required. Organizational complexity ishigher when (a) many functions are involved, (b) sep-arate organizations need to implement changes, (c)many people, from different backgrounds and withdiffering loyalties, are required to operate the pro-cess, or (d) scheduling meetings is difficult.

The half-life concept provides a reality check on pro-gram goals. For example, a product engineeringgroup in a research partner company was chargedwith cutting warranty costs in half by the year 2000.The group met in October 1997, and working back-wards from the target date, realized that the neces-sary changes would have to be completed by thesummer of 1998, leaving less than one year to rede-sign a wide range of parts and associated tooling anddevelop new procedures and training materials.After locating their process in the upper right-handcorner of technical and organizational complexity,suggesting a half-life of several years, the grouprevised their goals. With less pressure, the engineers’morale and commitment to the effort improved, andtheir faith in their leadership rose.

Success and Increasing ComplexityThe half-life concept has also proven helpful in revis-ing objectives over time. Early efforts tend to focuson relatively simple problems, low in technical andorganizational complexity, for which known tech-niques can be easily applied — the famous ‘low


Figure 6 Improvement Half-life.Source: Adapted from Schneiderman (1988)

hanging fruit’ discussed in the quality literature. Asthe simpler problems are solved, the program mustfocus on problems with greater technical and organi-zational complexity, causing the improvement half-life to rise (the ‘tougher challenges’ loop B3 in Figure7). As the rate of improvement slows, the self-rein-

Figure 7 Increasing Technical and Organizational Complexity


forcing employee-pull process (loop R2) weakens,and programs can falter.

The APEX initiative at Lucent Technologies faced thischallenge (Keating and Oliva, forthcoming). Initially,Lucent decomposed the product development pro-


cess into smaller sub-processes. Individual improve-ment teams tackled each activity in isolation. Focus-ing on the pieces of the process reducedorganizational complexity and led to a shortimprovement half-life — at Lucent this process wasdescribed as ‘catching the fat rabbits.’ With a shorthalf-life, product development time fell rapidly atfirst, demonstrating the value of the improvementprogram and boosting participation through theemployee pull loop R2. Product development timesfell from about 40 months to 20 in just two years.

However, the improvement teams soon exhaustedopportunities within the sub-processes, threateningprogress. The APEX leadership refocused theirefforts on the interactions of the sub-processes. Theseissues — for example, the coupling between market-ing, product definition, and engineering, or the link-ages among different technology and product plat-forms — created tougher challenges involvinggreater technical complexity and much greaterorganizational complexity, slowing the rate ofimprovement (loop B3 in Figure 7). Improvementteams also found that the improvement method-ologies that had worked so well at first became lessand less adequate for the new challenges arising fromthe couplings among phases and activities (loop B4).

The APEX leadership responded to these challengesby investing in the development of new improve-ment tools more appropriate for the interactionsacross organizational units (requiring that additionaltime be diverted from design work to training inimprovement techniques). They revised their aspir-ations for future development time reduction to bemore consistent with the growing complexity anddifficulty of the task. Over the next two years, theycut product development times another 40 per cent.

Eventually, as the organizational scope and breadthof the improvement effort grew, it became harder toattribute results to particular individuals or teams.The rewards of participation fell, and people wereless willing to tackle the problems (the ‘diffuse bene-fits’ loop B5 in Figure 7). An APEX leader com-mented:

The integration effort was not an easy task to sell to man-agement. The project was really tough to keep going sincethe benefits were for so many people. If you looked at it,the near term benefit to any one person was zero, so noone would work on it or pay for it. It is a common good,like clean water. It is too much to everyone’s benefit.

Lucent’s strategy in the successful APEX initiativehighlights important lessons for the management ofimprovement programs. First, the rate of improve-ment of any process depends on its technical andorganizational complexity. Objectives, and timeframes for achieving them, need to be ambitious butconsistent with the improvement half-lives of theprocess. However, the improvement half-life is not


an immutable constant. Lucent’s decision to focusinitially on individual pieces of the product develop-ment process reduced organizational and technicalcomplexity, lowering the half-life and speedingresults. Given the history of failed improvementefforts, demonstrating that the new program workedwas important to overcome initial resistance from theengineers. Lucent also recognized that initialimprovement rates were high while the easy prob-lems were solved, and adapted both their aspirationsand their improvement methodology to keep pacewith increasing technical and organizational com-plexity.

Skill DilutionGenerating enthusiasm among employees and reduc-ing throughput pressure so they have time forimprovement is far from sufficient. The employeesinvolved require training and support. Training, likeimprovement itself, cannot be bought and deliveredinstantly. It takes considerable time to develop train-ing and support infrastructure (the ‘capacity build-ing’ loop R4, Figure 8). Training capacity includesqualified instructors, well versed in the specific set oftechniques and the specifics of the organization, andcustomized materials. Active improvement teamsalso require support in the form of experiencedpeople they can turn to for help, libraries of past pro-jects and tools, and resources to implement theirimprovement ideas. Improvement programs canfalter as aggressive push and pull effects that expandthe demand for training and support grow far fasterthan capacity and support infrastructure (the ‘skilldilution’ loop B6 in Figure 8). One organization westudied aggressively promoted a new improvementprogram, overwhelming the training organization.The initially enthusiastic participants then stumbledin applying the tools, causing them to question themethodology and, ultimately, abandon the program.

Scope CreepA successful program can attract too much attentionand lose focus. Initially successful programs attractthe interest of more senior people in the organizationand in other functions and departments. The pro-gram is then applied to problems far outside thescope of issues for which it was designed. Such‘scope creep’ (loop B7 in Figure 8) causes a successfulprogram to be expanded to all sorts of problems forwhich it is ill suited, reducing the benefits obtainedand further diluting worker skills and support infra-structure. In several organizations we studied, over-zealous application of quality tools led to decliningeffectiveness and a backlash that damaged even theeffective programs.

External Interactions

So far we have treated improvement programs in iso-lation. But improvement programs affect, and areaffected by, other initiatives and other functions and


Figure 8 Scope Creep and Skill Dilution

organizations in the firm. One of our partner compa-nies initiated an average of two new improvementprogram initiatives each year for the past 15 years.Many programs were carried out simultaneously byoverlapping groups in overlapping areas of focus(Oliva et al., 1998). In other firms, improvementinitiatives in manufacturing had profound effects onproduct development, pricing, human resources,inventory management, and even the financial mar-kets. Many of these interactions were unexpectedand harmful.

Interactions with Other Initiatives

Few organizations rely on only a single improvementeffort. More often a stream of programs isimplemented concurrently. Even when they addressdifferent issues, these programs are linked throughshared resources including human effort, funding,information, and senior management attention. Theseinterconnections can create substantial synergiesacross programs as well as damaging competition.

Multiple programs can lead to synergies. Successfulprograms help focus organizational awareness on thepotential for improvement and the availability ofimprovement tools. Successful programs also gener-ate commitment to improvement that can be trans-ferred from one program to another. The techniqueslearned to support one program (e.g. processmapping) often carry over to other programs, short-ening the time required to build competence and ach-ieve results in subsequent initiatives. In many cases,the substantive knowledge built in one programhelps to identify the specific organizational elementsmost in need of improvement and the skills neededfor future improvement.


An example of these synergistic interactions is theuse of Theory of Constraints (TOC) (Goldratt, 1990)and Total Productive Maintenance (TPM) (Nakajima,1988) at one of the wafer fabrication plants ofNational Semiconductor Corporation. TOC is amethod for improving throughput, loweringexpenses, and managing material flow. The TOC pro-gram required managers to determine the capacity ofeach process step, then focus improvement on thoseprocesses that constrained throughput — the bottle-necks in the production flow. TPM focuses onincreasing the capacity of individual process steps byencouraging machine operators to focus on preven-tive and predictive maintenance rather than reactiverepairs. Armed with the results of the TOC analysis,TPM effort was focused on the bottleneck processeswhere increased capacity would boost factorythroughput the most.

Interactions across programs can be detrimental. Forexample, a cycle time reduction effort in the elec-tronics division of Ford led to dramatic improve-ments in manufacturing productivity. Initially, theimprovement relieved throughput pressure andallowed line workers to dedicate even more time toimprovement — the self-reinforcing productivitychain loop operated as a virtuous cycle. The programwas so successful, however, that soon the plantsdeveloped excess capacity. The leader of the effortrealized that if the excess capacity could not be used,a layoff would be inevitable. ‘Empty plants,’ he said,‘meant unaffordable plants.’

To fill its underutilized manufacturing facilities thedivision launched a new initiative aimed at improv-ing the throughput of the product development pro-cess. Getting new products into production fasterwould generate more demand and solve the excess


capacity problem. The effort led to a new andsuperior product development process, involvingbetter design and management tools (CAD/CAM,project management, a library of reusable, docu-mented designs, and so on). While millions werespent, the program did not achieve its objectives. Theinitiative failed because product development engin-eers also face the fundamental trade-off — they caneither spend their time on improvement or spendtheir time developing new products. Excess capacityin manufacturing led the organization to initiate thedevelopment of many new products. These new pro-jects intensified the pressure on the developmentengineers at the same time they were being asked toimplement the new process. A senior manager said,

There was tremendous pressure to grow....We would getourselves in situations where we would have a suc-cess...which translated into a resource problem for theengineers. We typically never said no (Repenning and Ster-man, 1997).

Under intense pressure to get new products into pro-duction, the engineers had no choice but to cut backthe time they spent learning to use the tools andmethods of the improved product development pro-cess. One engineer noted ‘the only thing they shootyou for is missing product launch. Everything else isnegotiable.’ Many never learned to use the newmulti-million dollar CAD/CAM system, andskimped on documentation, so the library of reusabledesigns never materialized. Because product devel-opment did not improve significantly, new productscould not be launched fast enough to utilize theexcess manufacturing capacity, and the division hadto reduce headcount.

Interactions with Other Organizational Units

Improvement programs interact with one anotherand with existing decision rules and organizationalroutines. Though subtle, these interactions can havedramatic effects, as the experience of Analog Devicesshows. Over a period of three years, a spectacularlysuccessful quality program at Analog Devicesroughly doubled the manufacturing yield andslashed the defect rate of outgoing products. Withlower production costs, higher yield, and better pro-ducts, Analog should have realized dramaticimprovement in financial results. Instead, the com-pany’s profits and stock price sank. The company’sperformance dropped not only in absolute terms, butrelative to competitors who had not adopted TQMpractices as quickly.

Dramatic improvements in yield, cycle time, andquality effectively doubled production capacity, out-stripping improvement in other areas. Decision rulesfor production starts did not adjust as fast as yieldimproved, leading to excess inventory. Capacitygrew much faster than demand, leading to fear of


layoffs that undermined commitment to furtherimprovement (the ‘fear of layoffs’ loop B8 in Figure9). Excess capacity could be absorbed if higher qual-ity and lower costs led to market share gains (the‘better products’ loop R5 in Figure 9), or if new pro-ducts could be brought to market faster (the ‘moreproducts’ loop R6). But Analog’s market share didnot rise appreciably because it was already the domi-nant firm in many markets and because competitorssought to preserve their market share by cuttingprices. New product introduction could not rise fastenough due to its greater complexity and longer half-lives compared to manufacturing.

Most important, rapid improvement in manufactur-ing caused a large decrease in unit direct costs. How-ever, indirect costs per unit, driven by R&D and gen-eral, administrative, and selling expenses, did notimprove as quickly due to their greater organiza-tional and technical complexity. Thus, while loweringcosts overall, the TQM initiative also changed the tra-ditional relationship between direct and indirectcosts, a relationship embedded in organizationalnorms for pricing. Analog, like many firms withextensive product lines, used markup pricing. Unitdirect costs were marked up by a standard ratio toyield a base price level, which was then adjusted onthe margin to respond to market conditions. The tra-ditional markup ratio, a little over 200 per cent, wasinitially sufficient to cover indirect costs and providea reasonable return (Table 1). The traditional markupchanged only slowly. Between 1985 and 1989, thesuccess of the TQM program led to a drop of about16 per cent in unit direct costs, and average sellingprices fell by about the same ratio. However, indirectcosts per unit fell less than 9 per cent. Analog’s tra-ditional gross margins were no longer sufficient, andoperating income fell by 45 per cent (Table 1). Theprecipitous drop in profit lowered stock prices, andAnalog responded by laying off 10 per cent of theworkforce, the first layoff in its history. Commitmentto improvement plummeted as the ‘fear of layoffs’loop dominated the system (Sterman et al., 1997).Thus Analog’s success in improving operations trig-gered unanticipated side effects that fed back to harmthe firm and undercut continued improvement.

While Analog eventually rebounded, and a new cropof quality efforts eventually grew, the unanticipatedside effects of rapid and unbalanced improvementsdamaged morale, disrupted continuous improve-ment, and threatened the survival of the firm.

The Iron Law of Layoffs

As the previous examples show, successful qualityprograms, by increasing yield and slashing scrap,defects, and cycle time, can lead to rapid growth incapacity. Unless demand grows rapidly as well, theresult is excess capacity and pressure for layoffs.Excess capacity is common since processes with low


Figure 9 Interactions With the Market and Job Security

Table 1 Changes in Cost Structure Caused by TQM Interacted With Pricing Policy to Yield Lower Profit (AllData Expressed per Unit Sold)

Historical data, Analog Devices

$/unit 1985 1989 %D

Average selling price 16.32 13.51 2 17.22 Cost of goods sold 7.61 6.41 2 15.8

5 Gross profit 8.71 7.10 2 18.52 Indirect costs 6.35 5.80 2 8.7

5 Operating income 2.36 1.30 2 44.7

Markup ratio (%)5 100*(ASP/COGS) 214 211 2 1.7

Source: Sterman et al. (1997).

complexity and short improvement half-lives (e.g.scrap and cycle time reduction) tend to be capacity-augmenting, while demand-generating activities (e.g.new product development, customer needs assess-ment, and supply chain integration) have longimprovement half-lives and involve long delays(Figure 9).

A simple calculation reveals how fast productivitycan grow before creating excess labor and pressurefor layoffs. The labor requirements of any firm aregiven by sales divided by labor productivity. Thefractional rate of change of labor requirements, l*, isthus equal to the fractional growth in sales, s, less thefractional rate of productivity growth, p:

l* 5 s 2 p (3)


Given the fractional attrition rate of the labor force(denoted ‘a’), the maximum rate of productivitygrowth consistent with a no-layoff policy is thus:

p # s 1 a (4)

This is the ‘Iron Law of Layoffs’: Productivityimprovement greater than the rate of sales growthplus the labor attrition rate necessarily creates excesscapacity. The more successfully an organizationimproves its manufacturing operations, the moreintense the pressure for layoffs. In Analog’s case,sales growth averaged 27 per cent/yr from its found-ing through 1985 and labor turnover was also high(10–20 per cent/yr) as employees readily found newopportunities in the expanding electronics industry.Analog could have absorbed productivity growth of


40 to 50 per cent/yr. However, as growth falteredand unemployment rose in the late 1980s and early1990s, voluntary quits fell below 5 per cent/yr. Withnegligible attrition and sales growth less than 10 percent/yr, even small rates of improvement led toexcess capacity.

The Iron Law of Layoffs provides several policyinsights. Deming (1986) exhorts management to‘drive out fear’ by guaranteeing job security to work-ers who participate in improvement programs. Butthe Iron Law of Layoffs means that such commit-ments are often not credible. In mature, slow growthindustries, or times of recession when voluntaryattrition is low, it can be difficult to sustain commit-ment to improvement. Yet slow demand growth andweak economic conditions motivate firms to under-take ambitious improvement programs. Many firmslaunch improvement initiatives precisely when theyare least able to absorb productivity gains withoutdownsizing.

There are several policies a firm can use to resolvethis dilemma. First, firms can sometimes convinceworkers that while improvement may cost some jobs,failure to improve will cost all jobs. This strategy,reversing Deming to ‘Drive in Fear’, enables firms tocredibly demonstrate that participation in improve-ment programs is in the employees’ best interestsdespite the threat of job losses (Repenning, 1998).Second, improvement efforts can be directed at theslow-improving processes first, so that the rate ofimprovement in demand and capacity is more bal-anced.

Interactions with the Market

Successful improvement can also create the oppositeproblem: too much demand. Improvement initiativesat one of our partner companies improved productquality dramatically. Prior to the quality program,the firm’s products often had multiple defects requir-ing substantial maintenance and frequent repairs.Through an aggressive improvement program, qual-ity rose dramatically, attracting large numbers ofnew customers.

The resulting market growth has been profound.Over a recent five-year period revenues grew atnearly 30 per cent/yr, much faster than productioncapacity. Huge backlogs meant many customerswaited more than 18 months for delivery. The com-pany struggled to meet the booming demand byreengineering manufacturing lines and aggressivelyhiring new workers. Production capacity doubled,but quality began to suffer:

In [one year], we moved four hundred machines [to doublecapacity in an existing plant]...ripping [them] up, moving[them] to a new location, setting them up and going. Qual-ity is at risk when you are doing all of that.


The effect of growth on quality is that you get your newbodies for making parts from the assembly line, [and] theyare the less experienced workers.

The high workload threatened commitment to con-tinued improvement. Workers throughout theorganization, from the assembly line to productdevelopment, felt strong pressure to reduce the timedevoted to improvement to boost near-term pro-duction (the ‘effort squeeze’ loop in Figure 1). Inter-views with line workers and engineers showed howthe pressure to ‘get the iron out’ — deliver productto customers — eroded their commitment to quality:

Sometimes we use junk because management doesn’t wantus to shut the line down.

Sometimes you will take a part that is cosmetically defec-tive — [with] a slight nick or scratch. They’ll take that piece[and] say it is a reject [but when] the next day comes upand they are really short of those pieces, they’ll take a fewof those pieces back, and say, well it is not that bad.

Thus initial success in boosting quality fed backthrough the market to increase demand making itdifficult to maintain the quality levels that led totheir success.

Conclusion

The failure of promising programs is a symptom ofthe organizational and economic challenges involvedin making them work. Managers are often unpre-pared for the interactions of improvement programswith processes outside the programs’ apparent focus.The improvement paradox arises because it is diffi-cult to anticipate the wide-ranging effects ofimprovement, especially when the intended changesare so clearly beneficial and the unintended adverseeffects are delayed or occur in other functions ororganizations.

Companies can strengthen the self-reinforcing pro-cesses that can lead to sustained improvement byactively managing the feedbacks that limit programsuccess. Managers must carefully plan the roll-out ofa new program to ensure demand for participationdoes not outstrip training and support infrastructure.Staffing, resources, and goals must be consistent withthe improvement half-life of the process to preventeffort squeeze. If employees are free to allocate timeto improvement, are adequately trained, and pro-gram scope remains focused, initial results will buildcommitment. By activating the virtuous cycle ofemployee pull early in the process, rapid pro-ductivity gains will follow, sustaining the programwithout command-and-control management.

However, managers should anticipate a slowdown inimprovement results as the complexity of the prob-


lems addressed increases. Managers may need toadopt new process improvement techniques toreduce complexity. Management must also recognizethe feedbacks arising from other improvement pro-grams, organizational units and the market. Decisionrules and procedures throughout the organizationshould be reviewed even if they do not appear tobe affected by the improvement program. In short,managers must become adept in understanding theirorganization as a dynamic system.

At our partner companies this process is currentlyunder way. We have developed a management flightsimulator and learning environment addressing theissues discussed here. These tools allow managers tomanage simulated improvement programs, experi-ence the long-term and distant side effects of theiractions, and design new strategies for the improve-ment programs they are leading. Our currentresearch focuses on evaluating and improving thelearning laboratory and assessing its impact as a cata-lyst for the change in organizations.

Acknowledgements

This work has been supported by the National Science Foun-dation (Grant SBR-9422228), Analog Devices, AT&T, FordMotor Company, Harley-Davidson, Lucent Technologies Inc.,and National Semiconductor Corporation. We are grateful tothe people of these firms for their outstanding help in thisresearch. They generously provided their data and consider-able amounts of their most precious resource — their time.

Note

1. Please direct correspondence to John Sterman, MIT SloanSchool of Management, Cambridge MA 02142, USA, [email protected]. Further information is availablethrough http://web.mit.edu/jsterman/www.

References

Analog Devices (1991) TQM at ADI. Analog Devices, Nor-wood, MA.

Carroll, J., Sterman, J. and Marcus, A. (1998) Playing the main-tenance game: how mental models drive organizationaldecisions. In Debating Rationality: Nonrational Elements ofOrganizational Decision Making, eds J. Halpern and R.Stern, pp. 99–121. Cornell University Press, Ithaca, NY.

Collins, J. and Porras, J. (1994) Build to Last: Successful Habitsof Visionary Companies. Harper Business, New York.

Deming, W.E. (1986) Out of the Crisis. MIT Press, Cam-bridge, MA.

Easton, G. and Jarrell, S. (1998) The effects of total qualitymanagement on corporate performance: an empiricalinvestigation. Journal of Business 71, 253–307.

Forrester, J.W. (1961) Industrial Dynamics. MIT Press, Cam-bridge, MA.

Goldratt, E.M. (1990) What is this Thing Called Theory of Con-straints and How should it be Implemented? North RiverPress, Croton on Hudson, NY.

Hamel, G. and Prahalad, C. (1989) Strategic intent. HarvardBusiness Review 67, 63–77.

Hendricks, K. and Singhal, V. (1996) Quality awards and the


market value of the firm: an empirical investigation.Management Science 43, 415–436.

Ishikawa, K. (1985) What is Total Quality Control? PrenticeHall, Englewood Cliffs, NJ.

Jones, A., Krahmer, E., Oliva, R., Repenning, N., Rockart, S.and Sterman, J. (1996) Comparing improvement pro-grams for product development and manufacturing:results from field studies. Paper presented at the Pro-ceedings of the 1996 International System DynamicsConference, Boston, MA.

Juran, J. (1969) Managerial Breakthrough: A New Concept of theManager’s Job. McGraw Hill, New York.

Kahneman, D. and Lovallo, D. (1993) Timid choices and boldforecasts: a cognitive perspective on risk-taking. Manage-ment Science 39, 17–31.

Keating, E. and Oliva, R. (forthcoming) A dynamic theory forsustaining process improvement teams in product devel-opment. In Advances in interdisciplinary studies of teams:Product development teams, eds M. Beyerlein and D. John-son, Vol. 5. JAI Press, Greenwich, CT.

Nakajima, S. (1988) Introduction to TPM: Total Productive Main-tenance. Productivity Press, Cambridge, MA.

Oliva, R., Rockart, S. and Sterman, J. (1998) Managing mul-tiple improvement efforts: lessons from a semiconductormanufacturing site. In Advances in the Management ofOrganizational Quality, eds D. Fedor and S. Ghosh, Vol.3, pp. 1–55. JAI Press, Greenwich, CT.

Repenning, N. (1998) Drive out fear (unless you can drive itin): the role of agency and job security in processimprovement. Working paper, available from author.

Repenning, N. and Sterman, J. (1997) Getting quality the old-fashioned way: self-confirming attributions in thedynamics of process improvement. Forthcoming in TheQuality Movement in America: Lessons for Theory andResearch, eds R. Scott and R. Cole. (Working paper avail-able from authors.)

Schaffer, R. and Thomson, H. (1992) Successful change pro-grams begin with results. Harvard Business Review 70,80–89.

Schneiderman, A. (1988) Setting quality goals. Quality ProgressApril, 55–57.

Shewhart, W. (1939) Statistical Method from the Viewpoint ofQuality Control. US Department of Agriculture, Wash-ington, DC.

Shiba, S., Graham, A. and Walden, D. (1993) A New AmericanTQM: Four Practical Revolutions in Management. Pro-ductivity Press and the Center for Quality Management,Cambridge, MA.

Sterman, J., Banaghan, E. and Gorman, E. (1992) Learning toStitch in Time: Building a Proactive Maintenance Culture atE.I. Du Pont de Nemours and Co. (Case study availablefrom author). MIT, Sloan School of Management, Cam-bridge, MA.

Sterman, J., Repenning, N., Oliva, R., Krahmer, E., Rockart, S.and Jones, A. (1996) The improvement paradox: design-ing sustainable quality improvement programs. Paperpresented at the Proceedings of the 1996 InternationalSystem Dynamics Conference, Boston, MA.

Sterman, J., Repenning, N. and Kofman, A. (1997) Unantici-pated side effects of successful quality programs: explor-ing a paradox of organizational improvement. Manage-ment Science 43, 503–521.

US General Accounting Office (1991) US Companies ImprovePerformance through Quality Efforts 9–190. GAO/NSIAD.

Young, E. (1991a) International Quality Study — Best PracticesReport. Ernst and Young/American Quality Foundation,Milwaukee, WI.

Young, E. (1991b) International Quality Study — Top Line Find-ings. Ernst and Young/American Quality Foundation,Milwaukee, WI.

Zangwill, W. and Kantor, P. (1998) Toward a theory of con-tinuous improvement and the learning curve. Manage-ment Science 44, 910–920.


ELIZABETH K. KEAT- ROGELIO OLIVA, Harv-ING, MIT Sloan School, ard Business School, Boston,Cambridge, MA 02142, MA 02163, USA.USA.

Rogelio Oliva is AssistantElizabeth K. Keating is an Professor of Management ataccounting doctoral candi- the Harvard Businessdate at the Sloan School of School. His current researchManagement at MIT. She is explores how the structuralCPA and holds an MBA in and operational elements ofFinance from New York service delivery processesUniversity. She has worked interact with human

extensively in the banking and non-profit industries. resource and marketing policies to determine servicequality and the sustainability of improvement pro-grams.

NELSON REPENNING, SCOTT ROCKART, MITMIT Sloan School, Cam- Sloan School, Cambridge,bridge, MA 02142, USA. MA 02142, USA.

Nelson Repenning is a Rob- Scott Rockart is a strategyert N. Noyce Career Devel- and system dynamics doc-opment Professor of Man- toral candidate at the Sloanagement at the MIT Sloan School of Management atSchool. His current research MIT. His research exploresexamines why process how the system nature ofimprovement techniques, firms creates multiple stablesuch as total quality man- levels of performance, and

agement and business process re-engineering succeed how managerial actions can affect the performance levelin some organizations but fail in others. of a firm.

JOHN D. STERMAN,MIT Sloan School, Cam-bridge, MA 02142, USA.

John Sterman is the Stand-ish Professor of Manage-ment Science and Directorof the System DynamicsGroup at the MIT SloanSchool of Management.


Date post:	16-Sep-2016
Category:	Documents
Upload:	elizabeth-keating
View:	215 times
Download:	1 times

Overcoming the improvement paradox

Documents