behind the learning curve: linking learning activities … · 2008-06-03 · • investments have...

33
BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES TO WASTE REDUCTION by M. A. LAPRE* A. S. MUICHERJEE** and L N. VAN WASSENHOVEt 96/24/TM * PhD Candidate at INSEAD, Boulevard de Constance, Fontainebleau 77305 Cedex, France. ** Arthur D. Little, Boston, Massachusetts, USA. t Professor of Operations Management and Operations Research at INSEAD, Boulevard de Constance, Fontainebleau 77305 Cedex, France. A working paper in the INSEAD Working Paper Series is intended as a means whereby a faculty researcher's thoughts and findings may be communicated to interested readers. The paper should be considered preliminary in nature and may require revision. Printed at INSEAD, Fontainebleau, France.

Upload: others

Post on 02-Aug-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

BEHIND THE LEARNING CURVE:LINKING LEARNING ACTIVITIES

TO WASTE REDUCTION

byM. A. LAPRE*

A. S. MUICHERJEE**

andL N. VAN WASSENHOVEt

96/24/TM

* PhD Candidate at INSEAD, Boulevard de Constance, Fontainebleau 77305 Cedex, France.

** Arthur D. Little, Boston, Massachusetts, USA.

t Professor of Operations Management and Operations Research at INSEAD, Boulevard de Constance,Fontainebleau 77305 Cedex, France.

A working paper in the INSEAD Working Paper Series is intended as a means whereby a faculty researcher'sthoughts and findings may be communicated to interested readers. The paper should be consideredpreliminary in nature and may require revision.

Printed at INSEAD, Fontainebleau, France.

Page 2: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Behind the Learning Curve: Linking LearningActivities to Waste Reduction

Michael A. Lapre • Amit Shankar Mukherjee • Luk N. Van Wassenhove

INSEAD, Boulevard de Constance, 77305 Fontainebleau Cedex, France

Arthur D. Little, Boston, Massachusetts

INSEAD, Boulevard de Constance, 77305 Fontainebleau Cedex, France

May 24, 1996

Abstract

This exploratory research on a decade of Total Quality Control in one factory opens up the

black box of the learning curve. Based on the organizational learning literature we derive a

quality learning curve that links different types of learning in quality improvement projects

to the evolution of the factory's waste rate. Projects that acquired both know-why and

know-how accelerated learning, other projects either impeded or did not affect the learning

process. We explain these findings in the context of dynamic production environments.

(Learning Curve, Organizational Learning, Quality, Technological Knowledge, Learning by

Experimentation)

1 Introduction

Experts have advocated that, in order to compete succesfully, firms should (i) undertake

organizational learning efforts, and (ii) embark on quality improvement programs. The link

between the two, however, is ill-understood. In a previous paper (Mukherjee, Lapre & Van

Wassenhove 1995) we started to build this link. We analyzed 62 quality improvement projects

undertaken in one factory over a decade. We identified dimensions of the learning process

that took place in these projects: conceptual and operational learning. Conceptual learning

is developing an understanding of why a problem occurs, i.e. the acquisition of know-why.

Operational learning is developing a skill of how to fix a problem, i.e. the acquisition of know-

1

Page 3: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

how. We found that conceptual and operational learning played a crucial role in changing

factory personnel's attention outside the project context. For example, projects which ac-

quired both know-why and know-how were more likely to yield new Standard Operating

Procedures or changes in Statistical Process Control.

This paper extends the link between learning and quality from a cross-sectional, project

level analysis to a longitudinal, factory level analysis. In it, we open up the black box of the

learning curve, explicitly introducing the knowledge acquisition from the quality improve-

ment projects. Given the role of conceptual and operational learning in changing factory

personnel's attention outside the project context, we explore the cumulative impact of these

two dimensions on the evolution of quality measured at the factory level. More specifically,

we use conceptual and operational learning to construct four cumulative project variables:

firefighting, ad hoc experiments, unproven theories and empirically proven theories. We

then use these cumulative project variables to explain changes in the factory's waste rate

(measured by the ratio of wasted material to total material released to the process).

To our knowledge there are no learning curve studies that incorporate behavioral vari-

ables to explain quality improvements. Doing so, we build on, and contribute to the sparse

literature on the learning process behind the learning curve. We show that projects which

acquired both know-why and know-how -empirically proven theories- accelerated the learn-

ing process. Other projects either impeded the learning process or did not affect the learning

rate.

We explain these findings in the context of dynamic production environments (Jaikumar

& Bohn 1992). In a dynamic production environment, contingencies -unexpected events

which disrupt production- occur routinely. Causes for contingencies include heterogenous

inputs, constantly changing environmental variables, and incomplete technological knowledge

which is defined as incomplete understanding of the effects of the input variables of a process

on the output. Contingencies define problems. Hence, in a dynamic production environment

problem solving is a key task. Factory personnel continually have to create technological

2

Page 4: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

knowledge to adapt to new situations, and they deliberately try to enhance improvement

rates.

We would like to note that there are few longitudinal studies explaining quality improve-

ments. Ittner (1992) is a notable exception. He uses a time trend to explain conformance

quality. Studying improvements in waste rates, as we do in this paper, is not only interesting

from a quality perspective. Improvements in waste rates have led to dramatic improvements

in Total Factor Productivity, see e.g. Hayes & Clark (1986) and Ittner (1992).

The paper is organized as follows. In section 2, we review drawbacks of the traditional

learning curve. We introduce concepts from the organizational learning literature that will

help us understand the learning process behind the learning curve. In section 3, we derive

a learning curve model for waste reduction that addresses the drawbacks mentioned above.

Section 4 describes the context of our study: the research site, its quality improvement

efforts, and our previous study of these efforts. Section 5 describes our data, section 6 the

econometric results. In section 7, we discuss implications for scholars and managers, and in

section 8, questions for future research.

2 Behind the Learning Curve

The learning curve phenomenon has been observed frequently. Firms realize large cost

reductions as they gain experience in production (see reviews by Yelle 1979, Dutton &

Thomas 1984). The functional form which has traditionally been suggested for the learning

curve is the power form

c = coq-b ,

(1)

where c is the cost to produce the q-th unit; co the cost to produce the first unit; and b the

learning rate.

Despite its frequent use, scholars have established fundamental shortcomings of the power

form (1). It is an entirely empirical phenomenon (Levy 1965), and as Lieberman (1984) notes

3

Page 5: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

the appropriate functional form of the learning curve has never been rigorously tested. This

is peculiar in light of the following observations that do not support (1).

• The traditional power form does not accomodate two often observed patterns: initial

downward concavity, and the plateau effect: after some amount of production no

further improvements are made (Muth 1986).

• Estimated learning rates vary widely, within an industry, within a plant, and over time

(Levy 1965, Dutton & Thomas 1984, Garvin 1993). Therefore, Dutton & Thomas

(1984) have advocated that the learning rate be treated as a dependent variable as

opposed to a given constant. Jaikumar & Bohn (1992) provide a rationale for the dy-

namic nature of learning rates. In dynamic production environments with incomplete

technological knowledge, factory personnel deliberately try to enhance improvement

rates.

• Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982).

Mishina's (1992) study of Boeing's production of B-17 heavy bombers corroborates

the importance of investments. His findings indicate that the scale-up of production

triggers learning. His rationale is that learning occurs only if there is a challenge.

Scale-up of production provides such challenge. Mishina used proven effective capacity

to measure learning by new experiences. Findings by Epple et al. (1996) and von

Hippel & Tyre (1995) are consistent with Mishina's results.

Probably the most important reason why the traditional power form does not accomodate

these observations is that it lacks an underlying theory. It does not provide any insight into

the learning process behind the results. It assumes that cumulative volume is the only source

of learning and it ignores deliberately undertaken learning efforts. Bohn (1994) distinguished

production and deliberate learning activities as different sources for the learning curve. Dut-

ton & Thomas (1984) made a similar distinction between autonomous and induced learning.

Levy (1965) and Adler & Clark (1991) are the only two papers we are aware of that include

behavioral variables in a learning curve analysis. Yet, as the latter authors acknowledge,

4

Page 6: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

variables like training hours and engineering activity are still proxies of the actual learning

process.

Consequently, we believe that advancing our understanding of the learning curve implies

(i) proposing a theoretical foundation, (ii) making the distinction between autonomous and

induced learning, (iii) incorporating a dynamic learning rate, and (iv) dealing with the

impact of investments that change the production system.

One impressive effort that addressed many of these issues was Levy's (1965) adaptation

curve. He assumed that for a new process a firm has a maximum rate of output P it would

like to achieve. The rate of output after q units have been produced is Q(q) < P. His

crucial assumption is that the rate of improvement in Q(q) is proportional to the amount

the process can improve. From this assumption follows the adaptation curve:

C 2 (q) = P[1 — e-(a+ti9)1,

where a represents the initial efficiency of the process, and 1.1 the process's rate of adaptation,

which can be affected by various factors y i , , yn like training and experience:

00+> /3=y= (2)

Equation (2) is used to explain differences in learning rates across workers.

The adaptation curve overcomes many of the problems associated with the power form

(1). However, Levy did not address the dynamic nature of the learning rate, and the impact

of changes in the production system. Furthermore, the adaptation curve raises new questions:

what theory underlies the assumption that the rate of improvement is proportional to the

performance gap, and how should the target level P be determined objectively?

One cannot obtain a value for P from estimation nor from hard data. Levy inferred

values for P from interviews. However, whatever value is inferred, it can be proven wrong.

At Chaparral Steel, for example, management sets targets for production rates "consider-

ably beyond current production capabilities." The company not only achieves these very

ambitious goals, but continues to improve performance unabated (Leonard-Barton 1992).

5

Page 7: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Likewise, Boeing more than quadrupled the target it set initially at the planning stage

(Mishina 1992).

We will propose a theory that addresses these problems. In particular, we feel that sig-

nificant value can be added by incorporating the research on organizational learning. Levy's

assumption that the rate of improvement is proportional to the amount the process can im-

prove is akin to what the organizational learning literature refers to as "performance gaps"

(Duncan & Weiss 1979). A performance gap is the discrepancy between actual performance

and aspired performance (managerial targets). This performance gap induces organizational

members to search for alternative actions that might reduce the performance gap. If the

gap cannot be attributed to factors outside the locus of control or to improper implemen-

tation, "... it must be considered a failure of organizational knowledge" (Duncan & Weiss

1979, p.52). Consequently, the organization should obtain better knowledge about action-

outcome relationships, i.e. it has to undertake organizational learning efforts to create better

technological knowledge (Bohn 1994).

Organizations learn if they process information about events in their environment so

as to change their range of potential behavior (Huber 1991). In its simplest form, this

process begins with individual learning: people experiencing and observing events. They

then reflect on their observations and conceptualize appropriate responses. Finally, they

test their concepts through implementation and thereby begin another learning cycle (Kim

1993).

Unstable environments can disrupt this simple cycle (Hedberg 1981). Such environments

possess characteristics that Senge (1990) called detail and dynamic complexity and March

& Olsen (1975) called ambiguity. Detail complexity arises when the presence of too many

variables makes it difficult to comprehend a problem in its entirety. Dynamic complexity

arises when distance and time make cause-and-effect difficult to establish. Ambiguity refers

to the simultaneous existence of equally plausible but mutually contradictory explanations of

a situation. These characteristics make it difficult for people to observe their environments,

6

Page 8: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

reflect on their observations and conceptualize and test possible solutions.

On completion of a learning cycle, individuals can change their beliefs. But how does indi-

vidual learning relate to organizational learning? In a perfect world, people act according to

their individual beliefs. Individual actions affect organizational actions, which evoke an en-

vironmental response. Finally, environmental responses affect individual beliefs. This cycle

of organizational learning can be disrupted as well. March & Olsen (1975) call a breakdown

between individual beliefs and individual actions "role-constrained experiential learning," a

breakdown between individual action and organizational action "audience experiential learn-

ing," a breakdown between organizational action and environmental response "superstitious

learning," and a breakdown between environmental response and individual beliefs "learning

under ambiguity." These breakdowns lead people to hold potentially inaccurate, subjective

beliefs called "myths" (see also Hedberg 1981).

In the next section we modify Levy's model. The resulting quality learning curve allows

us to investigate how learning affected quality improvement in a factory, where people have

to deal with detail and dynamic complexity, ambiguity and erroneous myths.

3 A Quality Learning Curve

In this section, we derive a learning curve model for quality improvement in a dynamic

production environment. We focus on waste reduction, a key driver of both quality and

productivity. We make four essential modifications to Levy's (1965) model: a theoretical

foundation; Mishina's experience variable; a natural, objective aspiration level; and a dy-

namic learning rate.

Let xt be the production output in month t, zt = maxr<t x, the proven effective capacity

(Mishina's experience variable), W(z) the waste rate after z has been proven to be feasible

production output, and P the desired waste rate. W(z)— P is the performance gap. In our

7

Page 9: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

context Levy's assumption becomes

dW(z) — ii [W(z) — P]'dz

where p denotes the learning rate. However, instead of merely assuming (3), we can interpret

this equation in light of the organizational learning literature on performance gaps. The

performance gap W (z) — P induces the organization to search for alternatives to reduce this

gap. A larger discrepancy spurs the organization to exert more effort in searching for better

knowledge. The effectiveness of acquiring new knowledge depends on the learning rate p.

Consequently, we can model the rate of improvement as the product of the learning rate and

the performance gap.

Mishina's findings suggest that scale-up of production triggers learning. In a dynamic

production environment scale-up of production can be achieved by adding new machines or

increasing machine speeds. In both cases factory personnel need to acquire new technological

knowledge on how to control their process in the changed production environment. We

therefore use Mishina's experience variable.

In the TQC literature there are two natural choices for P: zero defects (Deming 1982)

and the optimal conformance level (Juran & Gryna 1993). As our research site aimed for

zero defects, we will —without loss of generality— employ P = 0 in the remainder of the

paper. Contrary to Levy's model, the TQC context of waste reduction provides a value for

P which is not subjective and which can never be overtaken by actual performance.

Our last modification to Levy's model concerns the learning rate p. In recognition of

Jaikumar & Bohn's work on dynamic production environments, we assert that the learning

rate is fundamentally dynamic in nature. Factory personnel in a dynamic production envi-

ronment have to create new knowledge unabated to adapt to new situations, and deliberately

try to enhance rates of improvement.

Formally, let yi , ... , y7, be managerial factors that affect the rate of improvement, like

the cumulative number of quality improvement projects. Let t be the time index. The most

(3)

8

Page 10: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

parsimonious formulation for the learning rate isn

th r---- /30 + E AYit. (4)

i=iSolving (3) with condition W(z) > P gives (recall P = 0)

W(z) = ea+Pz, (5)

Substituting (4) in (5) and taking logarithms, we obtainn

In W (zt ) = a + ( fici + E PiYit )zt. (6)

i=1

Equation (6) can be estimated if we have data for waste, production, and the managerial

factors yi . Estimation of (6) is essentially an investigation of Dutton & Thomas's (1984)

autonomous/induced learning dimension like Adler & Clark (1991) did. Induced learning

involves efforts which are deliberately undertaken to acquire new knowledge. In contrast,

autonomous learning does not involve deliberately undertaken learning efforts; it is a much

less cognitive process that occurs naturally "on-the-job". In (6) #0 measures the autonomous

part of the learning rate, E7_ i Ayit the induced part.

In sum, we derived a learning curve model with (i) a theoretical foundation, (ii) an au-

tonomous/induced learning distinction, (iii) a dynamic learning rate, and (iv) an experience

variable that accounts for investments that change the production system. We now turn to

the context in which we estimate equation (6).

4 The Context

This section summarizes our previous work, aimed at building a link between organizational

learning and quality improvement. It is largely based on Mukherjee, Lapre & Van Wassen-

hove (1995).

4.1 The Research Site

Several considerations prevailed in choosing a research site. First, the site should provide

access to detailed data about the systems used to improve quality. Second, a study on

9

Page 11: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

knowledge acquisition should obviously focus on a site with incomplete knowledge. Third,

for our findings to be generalizable, the site should have implemented TQC methods adopted

by most quality-minded firms and recommended in the quality literature. Fourth, the site

should have an established record on successful quality improvement.

These considerations convinced us to choose N.V. Bekaert, S.A., a Belgian multinational

corporation. Bekaert is the world's largest independent producer of steel wire. In particular,

its Steel Cord Division, which hosted our research, produces about one-third of the world's

output of the steel wire (called "tire cord") used in the production of steel belted radial tires.

Our project received the enthusiastic backing of then CEO Karel Vinck and current CEO

Rafael Decaluwe, providing us with unlimited access to people and documents.

Bekaert's basic process flow is deceptively simple: Thick wire are pulled ("drawn")

through dies which progressively reduce their diameter. Very thin wire ("filaments") are

wrapped around each other to form tire cord. The simplest cord has two filaments; the most

complex, hundreds.

Of course, reality is much more intricate. First, each plant has a handful of huge drawing

machines upstream and hundreds of small drawing and filament wrapping machines down-

stream. The upstream machines handle continuous inputs of thick wire while the down-

stream machines individually process small spools. Second, swimming pool sized pits supply

lubricants (for the dies) to the drawing machines downstream. Thus, while these machines

are run independently, the soap circuits link them to each other. Third, the wire is heat

treated at two intermediate points to make it ductile. At one of these points, a chemical

process also coats the wires with brass. Despite the use of sophisticated controls, wires which

are heat treated and coated together do not necessarily have identical properties. Fourth,

even for similar cord, different customers demand customized product properties. Fifth,

Bekaert's suppliers, which included some of the best known steel companies in Europe and

Japan, cannot guarantee homogeneity of properties across the thousands of tons of wire they

deliver.

10

Page 12: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

As a result, plant personnel have to contend with very high levels of detail and dynamic

complexity (Senge 1990). The former arises from the need to coordinate the large numbers of

machines and the different technical skills required at the different stages. The latter arises

from the ease with which effects of problems experienced at any machine or production

stage could be transmitted to other machines and stages. In other words, Bekaert's factories

readily fit Jaikumar & Bohn's (1992) definition of dynamic production environments.

During the 1980s, Bekaert's customers —tire manufacturers— experienced traumatic change,

and responded by making simultaneous demands of high quality, short lead times, low cost

and product line flexibility. Bekaert responded by initiating a pilot quality improvement

program at its flagship Aalter plant (Belgium) in 1981. Little scientific knowledge existed

concerning the production process of tire cord. Through its joint venture with the Bridge-

stone Company of Japan, Bekaert had ready access to knowledge about Japanese quality

control techniques. Over the next ten years, it introduced and institutionalized among oth-

ers a structured approach to problem solving, a functional TQC organization, Statistical

Process Control (SPC), Standard Operating Procedures (SOPs), TQC project teams, infor-

mation systems providing standardized daily, weekly and monthly production and quality

data, process capability measures, and quality control circles. Researchers on quality have

long prescribed these methods (see e.g. Juran Gryna 1993, Deming 1982, Ishikawa Lu

1985, Imai 1986, Wadsworth et al. 1986).

Its diligent efforts seem to have borne fruit, for in 1990, CEO Karel Vinck won the

first European Forum for Quality Leadership Award. Many people believed that the jury

was favorably impressed by his vision for quality exemplified by the innovative practices at

Aalter. Bekaert's second major recognition for quality came in 1992, when its Burgos plant

(Spain) won a European Quality Prize.

In sum, Bekaert provided access, was a dynamic production environment with incomplete

technological knowledge, used common TQC methods, and achieved considerable success.

For further information on Bekaert and its TQC record we refer to the case "Bekaert: Beyond

11

Page 13: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

the Quality Prizes" (Mukherjee, Lapre & Van Wassenhove 1994).

4.2 Quality Improvement Projects at Aalter

The literature on quality distinguishes two types of projects (e.g. Deming 1982, Imai 1986,

Juran Gryna 1993). "Breakthrough" or "common cause" projects deal with issues of

potentially wide-ranging impact, which may be able to improve performance dramatically.

Bekaert calls these projects "key projects". They are often interdepartmental in scope, and

managed by engineers. Kaizen or "special cause" projects are undertaken in the spirit of con-

tinuous improvement. While they do not have significant impact individually, cumulatively

they can propel firms to very high levels of performance. Bekaert describes these projects

as IKZ (the Flemish acronym for TQC) projects. They are typically intradepartmental in

scope, and managed by foremen or supervisors.

We had unlimited access to all of Aalter's records on improvement projects undertaken

throughout the 1980s. From these, we selected projects undertaken between 1982 and 1991

that (i) sought to improve product attributes or process control (as opposed to say, house-

keeping), (ii) had progressed (at least) past the testing stage, and (iii) had been adequately

documented. This selection left us with 62 projects: 55 IKZ projects and 7 key projects. Five

of these key projects were undertaken on Aalter's "model line". In the late 1980s, Bekaert

realized that central R&D laboratories lacked the characteristics of the dynamic production

environment encountered in the factory. It therefore re-located process optimization to the

factory (Mukherjee et al. 1994). In 1988, Bekaert established a model line at Aalter for an

important, representative product, and asked its personnel to create fundamental process

control knowledge without sacrificing the production of saleable wire. The model line rou-

tinely used natural and controlled experiments to solve problems. It is essentially a learning

laboratory in the factory (Leonard-Barton 1992).

12

Page 14: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

4.3 The Learning Process in Quality Improvement Projects

We coded the projects on questions which dealt with their learning process and with their

performance. We structured these questions on the basis of the organizational learning lit-

erature and our experience with earlier exploratory research (Mukherjee & Van Wassenhove

1994, Mukherjee 1992). We conducted a factor analysis on the questions which dealt with

the learning process. Two of the resulting factors are crucial for this paper; they mapped

onto what Kim (1993) calls conceptual learning and operational learning.

The questions that loaded onto conceptual learning measured (i) the use of scientific

models and statistical experiments to assess the cause-and-effect of contingencies, and (ii)

the depth of analysis in the project. In other words, conceptual learning is developing an

understanding of why contingencies occur, i.e. the acquisition of know-why. The questions

that loaded onto operational learning measured the modification of action variables (e.g.

process settings) and the follow-up of experimental results. In other words, operational

learning is developing a skill of how to deal with contingencies, i.e. the acquisition of know-

how.

Our broadest measure of project performance "the ability to change attention rules"

evaluated whether a project resulted in modifications of the set of variables normally mon-

itored by plant personnel. It measured whether the project resulted in e.g. new SOPs or

modifications in the Statistical Process Control. The foundation for this question lies in the

organizational learning literature. March & Olsen (1975) assert that it is possible for the

beliefs or actions of members of the organization not to affect the actions of the organization.

Recall from section 2.1 that role-constrained experiential learning and audience experiential

learning can lead to erroneous myths.

Regression analysis showed that the ability to change attention rules is strongly enhanced

by both conceptual and operational learning. Conceptual learning helped project teams

apply scientific principles to develop falsifiable hypotheses and models of cause-and-effect.

13

Page 15: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Combined with operational learning —observable changes in action variables and follow-up

of experimental evidence— project teams could overturn existing myths. Consequently, the

"judicious mix of conceptual and operational learning" is important for convincing non-

project members to modify attention rules outside the project context.

Given the importance of conceptual and operational learning outside the project context,

we want to investigate whether the conceptual and operational learning in projects had an

impact at the factory level. Using the quality learning curve derived in section 3, we can

explore the cumulative impact of conceptual and operational learning on the factory's waste

performance over time.

5 Data

Our data comes from company records. In January 1984, the Aalter plant introduced a

reporting system that provided standardized production and quality data on a monthly

basis.

• Waste. For all 6 major process steps, the plant reported the percentage of steel wire

scrapped because of irrepairable defects. Let % denote this waste rate for process step

i in month t. The yield for step i is then 1— W. Hence, the yield for the entire factory

is n i( 1 — Wit). Finally, we obtain the factory's waste rate Wt = 1— HA ]. - Wit ) . Figure

1 shows the waste evolution. We correct In Wt for seasonal patterns by substracting

the sample average for that month. Thus we control for effects like increased waste

levels due to start-ups in production months that followed the holidays.

• Production. The factory's production volume x t was readily available in the re-

ports: the tonnage of wire produced in the last step. The proven effective capacity is

constructed straightforwardly as zt = maxr<t xr . See figure 2.

• Projects. For the IKZ projects the project reports provided the project completion

dates. The completion dates for the breakthrough projects were recorded in Mukherjee

(1992). For each project we had factor scores for conceptual and operational learning

14

Page 16: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

at our disposal from our previous study (Mukherjee et al. 1995). 1 Comparing the factor

scores for conceptual and operational learning with the sample averages we classified

each project according to high (H) or low (L) conceptual learning and high or low

operational learning. 2 This allows us to compute C LLt, the cumulative number of

projects with low conceptual learning and low operational learning completed up to

month t. Similarly, we compute C LHt , C H Lt , and C Hilt.

Figure 1 about here

Figure 2 about here

Our sample includes 80 production months: January 1984 to April 1990 (there are 11

production months per year). We now turn to the econometric analysis.

6 Quality Learning Curve Estimates

We first estimate the traditional learning curve with the various experience variables pro-

posed in the literature, time (t), cumulative volume (q), and proven effective capacity (z):

In Wt = al + bi ln t (7)

In Wt = a2 -I- b2 In qt (8)

In Wt = a3 + b3 In zt (9)

l A factor analysis reduces a set of variables (responses to Likert scale or numerical questions) to a smaller

set of new variables called factors. This set accounts for the larger part of the variation in the original set. In

standard factor analysis (principal components followed by varimax rotation) the newly constructed factors

are uncorrelated with one another. Typically, each original variable is highly correlated with one factor,

and relatively uncorrelated with the other factors. For each observation factor scores can be constructed. A

factor score for an observation on a particular factor is a weighted average of the standardized responses to

the original questions. The weights are determined by the correlations between the original questions and

the specific factor.

2Aalter produces three products. The waste and production data concern Aalter's most important prod-

uct, tire cord. Five IKZ projects did not deal with tire cord. Hence, we only retained the 57 projects that

did.

15

Page 17: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Next, we estimate the quality learning curve derived in section 3:

In Wt = a + Po + OiCILLt+ thCLHt+ #3CHLt+ thCHHOzt. (10)

The Aalter plant introduced Statistical Process Control for practically all product and pro-

cess variables in 1986. As figure 1 shows, the variance in the waste rate decreased after mid

1986. We therefore test whether we have to control for heteroskedasticity, by splitting up the

sample into two subperiods: one before mid 1986, and one after mid 1986. We estimate (10)

for each subperiod, and test whether the error variance in the first subperiod is larger than

in the second (see e.g. Judge et al. 1988, p.363). This is in fact the case for equation (10).

Hence, we employ weighted least squares with weights determined by the two estimated

error variances. Table 1 shows the results. The traditional learning curve models explain

less variation than the quality learning curve (10). Moreover, the corresponding Durbin

Watson statistics show high degrees of autocorrelation indicating ill-specified models. How-

ever, there is no clear evidence of autocorrelation for the quality learning curve (the Durbin

Watson statistic of 1.56 is inside the inconclusive region (1.36, 1.62) at the 1 % significance

level, and close to the upper bound).

In section 2, we discussed several shortcomings of the traditional learning curve. We

derived the quality learning curve to overcome these shortcomings. Therefore, it makes

sense that the theory based quality learning curve yields a better fit than the traditional

learning curve. We will focus our discussion on the quality learning curve estimates. In

particular, we discuss the impact of the five types of learning that constitute the learning

rate At = f3o+ 131CLLt + /j2 CLHt+133CHLt + 134CHHt•

Table 1 about here

• The sign of A) is negative and significant, indicating clear evidence of autonomous

learning. This finding endorses the importance of learning by new experiences. As

Mishina (1992) asserted, the scale-up of production introduces new problems for factory

personnel. Learning occurs by addressing these problems. Aalter's production data

16

Page 18: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

supports his thesis. Indeed, 50 % of the scale-up during the 1980s occurred at the

end of 1987 and the beginning of 1988 (figure 2). Figure 1 shows that this scale-up

coincided with a major reduction in waste. The scale-up of production changes the

production environment (higher machine speeds or new machines). Factory personnel

have to learn how to produce good output in this changed production environment.

The results for the four types of induced learning confirm our earlier cross-sectional results.

• The insignificance of /el indicates that projects with little conceptual and little op-

erational learning did not affect the learning rate. In these projects the team hardly

reflects on the causes of contingencies, implements only minor changes, and obtains lit-

tle follow-up. We label these projects as "firefighting." The lack of root cause analysis

and the lack of operational results lead other people to ignore these efforts.

• Projects with little conceptual learning but substantial operational learning generate

know-how. Recall from section 4.3 that operational learning basically means modifying

action variables and obtaining follow-up of the experimental results, i.e. these projects

generate successful solutions within the project context. However, according to the

insignificance of they did not affect the learning rate. Due to the lack of concep-

tual learning, the engendered know-how is not well understood. It is more art than

science. We therefore label these projects "artisan skills." They often take the form

of ad hoc experiments which Bohn (1987, p. 15) defines as "deliberate changes made

to the process ... without a careful control group or experimental design." Instead of

the deliberate changes, many confounding factors may have caused the results. This

holds in particular in dynamic production environments with incomplete technological

knowledge like Bekaert's. Consequently, people form their own subjective interpre-

tations of purely know-how based experimental results. March & Olsen (1975) call

this breakdown between environmental response and individual beliefs learning under

ambiguity. It leaves other people's strongly held beliefs, or myths, unchallenged. This

may explain why artisan skills did not receive factory personnel's attention outside the

17

Page 19: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

project context, and hence did not affect the learning rate.

• Projects with substantial conceptual learning and little operational learning develop

a possible conceptual understanding but fail to produce the corresponding know-how.

Hence, we label these efforts as "unproven theories." Projects of this type had a

significantly disruptive impact on the learning rate witnessed by the positive sign of

143. Unproven theories can lead individuals to change their beliefs even if there is no

empirical foundation to do so. March & Olsen call this process superstitious learning:

there is no link between organizational action and environmental response, yet people

update their individual beliefs. New erroneous myths result, on which people can take

future detrimental actions.

• The negative and significant sign of /̂34 implies that projects with both substantial

conceptual and operational learning enhanced the learning rate. In these projects,

the team uses scientific models and statistical experiments to develop a theory that

explains the occurrence of contingencies. Based on this theory the team implements

changes and obtains empirical evidence. Therefore, we label these projects "empirically

proven theories." Conceptual learning guides the team in determining the key variables

to modify. Instead of changing variables by trial-and-error, the team applies scientific

principles to develop falsifiable hypotheses and build models of cause-and-effect. Con-

sequently, the team is less likely to make erroneous links of cause-and-effect. Moreover,

supporters of erroneous myths will have difficulty disputing the science based evidence.

The power of conceptual learning to overturn myths combined with the changes in

action variables and the follow-up of experimental evidence forces even the most skep-

tical of factory personnel to recognize an improved method for production. (See also

Mukherjee, Lapre & Van Wassenhove 1995.)

To appreciate the magnitude of the estimates in table 1 one might ask, how much does

one additional project of type i change the learning rate (30? For each type of project weA A

calculated ,3i / i30 x 100% to measure this effect. Table 2 shows that one additional empirically

18

Page 20: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

proven theory would improve the learning rate with 3.5%, whereas one additional unproven

theory would worsen the learning rate with 3.6%.

Table 2 about here

The mix of conceptual and operational learning is important in enhancing waste reduc-

tion. The question arises why some projects with high conceptual learning failed to generate

the corresponding operational learning, whereas others succeeded. In-depth case studies of

two key projects illustrate the difference. The following descriptions are entirely based on

Mukherjee (1992) and Mukherjee & Jaikumar (1993). Both projects dealt with "drawabil-

ity," the number of dies that wear out during the production of one metric ton of wire. Wear

out of dies is an important process variable, for the shape of the die determines, among

others, the quality of the wire and the amount of wire fractures (Bekaert's major source of

process interruptions).

One key project sought to introduce SPC to control the soap circuits. Bekaert person-

nel widely believed that soap variables like 'fatty acid' concentration, temperature, levels of

several chemicals, etc., affected drawability. The key problem was that the chemical compo-

sition of the soap changed over time. This process of 'ageing' was ill-understood. Aalter used

flow-charts to codify its beliefs and corrective actions that specified how much fresh soap

to add/old soap to remove under different conditions. R&D had specified several corrective

actions. The project team modified the flow-charts several times based on new information

from R&D and their own experiences. The team strongly suspected that wire-related vari-

ables affected drawability, but they chose not to analyze these variables because it would be

difficult to get the necessary data from an upstream process step. At the end of the project,

the team believed they could achieve any reasonable process capability as long as they did

not have to worry about production! However, they believed that the theory and practice

of soap steering were a long way apart. One team member had begun to doubt the utility

of applying SPC in an area where the plant lacked fundamental knowledge.

The model line at Aalter (MLA) attempted to improve drawability by an order-of-

19

Page 21: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

magnitude. The MLA routinely collected data on hundreds of process parameters and

environmental variables across the entire length of the production line. The MLA man-

ager pulled together fragmentary knowledge on unrelated experiments from several R&D

units. He combined these insights into a formal chemical model. He also included factors

based on prior production experience and regressions on natural data. The MLA team tested

the model with controlled and natural experiments. The data showed that none of the soap

parameters had any statistically discernable effect on drawability, but several wire-related

variables determined by upstream process settings did. Based on this knowledge, the MLA

achieved and sustained sharply improved drawability performance.

Both projects used insights from R&D, and pursued a deep understanding of drawability.

So, both projects engendered high levels of conceptual learning. Yet the nature of the learning

process was in stark contrast. Due to several factors the first project remained an unproven

theory, whereas the second project became an empirically proven theory.

First, in the soap steering project tacit knowledge concerning the role of wire-related

variables was ignored, because the team lacked the means to collect relevant data. Note that

the MLA showed the importance of this tacit knowledge. Its data collection system allowed

the MLA to consider any variable it deemed relevant.

Second, the soap steering project incorporated scientific knowledge from R&D, but, con-

trary to the MLA, it did not combine insights from R&D to build a scientific model. Not

only did the MLA construct a scientific model based on R&D insights, it also incorporated

prior production experience into the model. So, the MLA recognized that knowledge from

R&D laboratories often needs to be adapted to full scale manufacturing.

Third, the MLA used natural and controlled experiments as opposed to ad hoc experi-

ments to validate scientific models. Doing so, it reduced the probability that confounding

factors instead of deliberate changes caused the results. We believe that the distinctions

between these two projects are typical for unproven vs. empirically proven theories.

Figure 3 shows the evolution of the two types of projects that significantly affected

20

Page 22: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

the learning rate, unproven theories and empirically proven theories. Figure 4 depicts the

estimated dynamic learning rate = -00 - OiCLLt - 132 C LHt — ,83CHLt — 134CHHt.

Figure 3 about here

Figure 4 about here

We distinguish three phases in Aalter's TQC evolution.

• Until 1987, Aalter paved the road for TQC. It introduced most of the commonly

accepted quality tools like a structured approach to problem solving, SOPs, SPC, a

functional TQC organization, and invested heavily in training its shop floor personnel

in the use of these tools.

• In the years 1987 and 1988, Aalter achieved a major improvement in waste which

seems to have been driven by two forces. First, there was a significant scale-up of

production (see figure 2) triggering autonomous learning. Second, the investments in

TQC started to pay off: plant management required first line managers to conduct

several IKZ projects. An increasing number of IKZ projects were empirically proven

theories (figure 3) enhancing the learning rate (figure 4). However, the learning rate

dropped back to its old level due to a number of unproven theories.

• In 1990, the investment in the model line —a learning laboratory in the factory— explic-

itly set up in 1988 to acquire fundamental knowledge on the production process started

to pay off. The model line produced the knowledge that accelerates the learning rate:

theory driven controlled and natural experiments which marry conceptual learning and

operational learning (figure 4).

This remarkable evolution of TQC at Bekaert has some implications for scholars and man-

agers. We will discuss these in the next section.

7 Implications

Figure 5 summarizes the logic of the quality learning curve. It distinguishes between au-

tonomous and induced learning (top and bottom half in figure 5 resp.). The scale-up of

21

Page 23: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

production changes the production environment. In this new production environment ac-

tual waste performance differs from the target of zero defects, i.e. there is a performance

gap. The performance gap in turn triggers learning by new experiences, which leads to waste

reduction. The effectiveness of learning by new experiences is determined by the learning

rate, which is dynamic in nature. Knowledge acquired through quality improvement projects

(deliberate learning activities) modify the learning rate.

Figure 5 about here

Our quintessential contribution to the learning curve literature is the introduction of

organizational learning variables into the learning curve. Instead of merely linking production

and cost improvement, we incorporate the learning process that acquires knowledge into the

learning curve. Moreover, we provide empirical evidence for Dutton & Thomas's (1984)

seminal piece: the learning rate is definitely not constant, in fact, it can be modelled as

a dependent variable, and autonomous and induced learning are important explanatory

variables for the learning rate.

This paper sheds new light on Adler & Clark's (1991) findings. They found that induced

learning can disrupt as well as facilitate the learning process. Our analysis confirms this.

However, we can root our explanation in the dimensions of the learning process we identified

in earlier work. Induced learning that provides both the know-why and the know-how

enhances learning, whereas induced learning that only yields know-why disrupts the learning

process.

Our message to managers is a clear one: (at least part of) the learning process for quality

improvement is manageable. Our findings show that it is in fact possible to accelerate the

quality learning curve. However, the key may well be different from current norms for quality

improvement. Often factory personnel identify problems using SPC, assess magnitudes with

Pareto analyses, determine possible causes using Ishikawa diagrams to structure strongly held

beliefs of cause-and-effect (i.e. myths), tentatively implement one or more (uncontrolled)

changes, observe the effects using scatter diagrams and SPC charts, and either label the

22

Page 24: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

problem as solved or cycle back to try another change. Such efforts are generally lauded

quintessential examples of TQC or Kaizen implemented by empowered workers.

This approach relies on operational learning. We do not dispute that ad hoc experiments

can yield useful solutions for local problems. However, in a dynamic production environment

characterized by detail and dynamic complexity and ambiguity such locally acquired know-

how does not affect other people's strongly held beliefs, or myths. It takes conceptual learning

to challenge myths. Our findings suggest that accelerating learning for quality improvement

requires both theory and practice. Empirically proven theories are capable of overturning

received wisdom.

The disruptive effect of unproven theories underscribes our warning in earlier work

(Mukherjee, Lapre & Van Wassenhove 1995). If an organization has so far mainly relied

on operational learning, it may be difficult to incorporate conceptual learning. Henderson

& Clark (1990) suggest that the appropriate organizational structure might be absent. Our

findings seem to be consistent with their suggestion. The model line -a learning laboratory

in the factory- is an organizational structure that fosters the mix of conceptual and oper-

ational learning. The non-MLA key projects were unproven theories, whereas all MLA key

projects were empirically proven theories.

8 Suggestions for Future Research

We propose some pieces that could follow our work in addition to the issues we identified in

Mukherjee, Lapre & Van Wassenhove (1995) concerning the reproducibility of our work and

the development of more rigorous definitions of the concepts of operational and conceptual

learning.

Having confirmed our project level findings at the factory level, we re-emphasize the

importance of researching the organizational systems for consistently producing and sup-

porting conceptual and operational learning. Both the projects conducted on Aalter's model

line for learning and earlier research at Bekaert provide a clue where to start. Further

23

Page 25: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

study of sophisticated experimentation in the factory could possibly build on Bohn's work

on experimentation (1987, 1995).

This paper focuses on the autonomous/induced learning dimension from Dutton &

Thomas's (1984) seminal work. Further research should consider their full framework by

including the endogenous/exogenous dimension in the learning curve. This requires the

study of multiple production units. Such research could help develop an understanding of

how knowledge gets transferred across sites. Which types of knowledge are easier to transfer?

How should knowledge acquisition be managed in a network of plants? Argote, Epple and

others have studied the transfer of autonomous learning (e.g. Argote et al. 1990), yet the

transfer of induced learning has never been addressed.

The current learning curve study focuses on a quality measure. Another important area

for future research would be the development and estimation of learning curve models for

productivity/cost measures with a theoretical footing.

We believe that answers to these questions will enhance-firms' efforts to manage and

measure their learning curve processes.

Acknowledgements

Roger Bohn provided useful comments. We thank the management and employees of N.V.

Bekaert, S.A. for their unstinting cooperation. This research was supported by INSEAD

R&D budget 2272, and (for a part of the field research) by the Division of Research, Har-

vard Business School. The work of Michael Lapre is in part supported by the Sasakawa

Foundation.

References

Adler, P.S. and K.B. Clark "Behind the Learning Curve: A Sketch of the Learning Process,"

Management Science, 37 (1991), 267-281

Argote, L., S.L. Beckman and D. Epple, "The Persistence and Transfer of Learning in

24

Page 26: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Industrial Settings," Management Science, 36 (1990), 140-154

Bohn, R.E., Learning by Experimentation in Manufacturing, Harvard Business School

Working Paper 88-001, 1987

Bohn, R.E., "Measuring and Managing Technological Knowledge," Sloan Management Re-

view, (Fall 1994), 61-73

Bohn, R.E., "Noise and Learning in Semiconductor Manufacturing," Management Science,

41 (1995), 31-42

Deming, W.E., Quality, Productivity, and Competitive Position, MIT Center for Advanced

Engineering Study, Cambridge, 1982

Duncan, R. and A. Weiss, "Organizational Learning: Implications for Organizational De-

sign," Research in Organizational Behavior, 1 (1979), 75-123

Dutton, J.M. and A. Thomas, "Treating Progress Functions as a Managerial Opportunity,"

Academy of Management Review, 9 (1984), 235-247

Epple, D., L. Argote and K. Murphy, "An Empirical Investigation of the Micro Struc-

ture of Knowledge Acquisition and Transfer Through Learning by Doing," Operations

Research, 44 (1996), 77-86

Garvin, D.A., "Building a Learning Organization," Harvard Business Review, (July-August

1993), 78-91

Hayes, R.H. K.B. Clark, "Why Some Factories Are More Productive Than Others,"

Harvard Business Review, (September-October 1986) , 66-73

Hax, A.C. and N.S. Majluf, "Competitive Cost Dynamics: The Experience Curve," Inter-

faces, 12 (1982), 50-61

Hedberg, B.L.T., "How Organizations Learn and Unlearn," in Handbook of Organizational

Design, P.C. Nystrom and W.H. Starbuck (Eds.), Oxford University Press, 1981

Henderson, R.M. and K.B. Clark, "Architectural Innovation: The Reconfiguration of Ex-

isting Product Technologies and the Failure of Established Firms," Administrative

Science Quarterly, 35 (1990), 9-30

25

Page 27: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Huber, G.P., "Organizational Learning: The Contributing Processes and the Literatures,"

Organization Science, 2 (1991), 88-115

Imai, M., Kaizen, the Key to Japan's Competitive Success, Random House, New York, 1986

Ishikawa, K. and D.J. Lu, What Is Total Quality Control? Prentice-Hall, Englewood Cliffs,

1985

Ittner, C.D., "The Economics and Measurement of Quality Costs: An Empirical Investiga-

tion," HBS doctoral dissertation, 1992

Jaikumar, R. and R.E. Bohn, "A Dynamic Approach to Operations Management: An

Alternative to Static Optimization," International Journal of Production Economics,

27 (1992), 265-282

Judge, G.G., et al. "Introduction to the Theory and Practice of Econometrics," 2nd Edition,

Wiley, New York, 1988

Juran, J.M., and F.M. Gryna, Quality Planning and Analysis, 3rd Edition, McGraw-Hill,

New York, 1993

Kim, D.H., "The Link between Inidividual and Organizational Learning," Sloan Manage-

ment Review, (Fall 1993), 37-50

Leonard-Barton, D., "The Factory as a Learning Laboratory," Sloan Management Review,

(Fall 1992), 23-38

Levy, F.K., "Adaptation in the Production Process," Management Science, 11 (1965), B-

136-B-154

Lieberman, M.B., "The Learning Curve and Pricing in the Chemical Processing Industries,"

RAND Journal of Economics, (Summer 1984), 213-228

March, J.G. and J.P. Olsen, "The Uncertainty of the Past: Organizational Learning under

Ambiguity," European Journal of Political Research, 3 (1975), 147-171

Mishina, K., "Learning by New Experiences," HBS Working Paper 92-084, 1992

Mukherjee, A.S., The Effective Management of Organizational Learning and Process Con-

trol, Unpublished Doctoral Dissertation, Harvard Business School, Boston, 1992

26

Page 28: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Mukherjee, A.S. and R. Jaikumar, Paradigms of Process Control, INSEAD Working Paper

93/76/TM, 1993

Mukherjee, A.S. and L.N. Van Wassenhove, The Impact of Knowledge on Quality, INSEAD

Working Paper 94/26/TM, 1994

Mukherjee, A.S., M.A. Lapre and L.N. Van Wassenhove, Bekaert: Beyond the Quality

Prizes, INSEAD-CEDEP case 01/94-4253, Fontainebleau, France, 1994

Mukherjee, A.S., M.A. Lapre and L.N. Van Wassenhove, Knowledge Driven Quality Im-

provement, INSEAD Working Paper 95/48/TM, 1995

Muth, J.F., "Search Theory and the Manufacturing Progress Function," Management Sci-

ence, 32 (1986), 948-962

Senge, P.M., "The Leader's New Work: Building Learning Organizations," Sloan Manage-

ment Review, (Fall 1990), 7-23

Von Hippel, E. and M.J. Tyre, "How Learning by Doing is Done: Problem Identification

in Novel Process Equipment," Research Policy, 24 (1995), 1-12

Wadsworth, H.M., K.S. Stephens and A.B. Godfrey, Modern Methods for Quality Control

and Improvement, Wiley, New York, 1986

Yelle, L.E., "The Learning Curve: Historical Review and Comprehensive Survey," Decision

Sciences, 10 (1979), 302-328

27

Page 29: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

W

84 85 86 87 88 89 90 91

Figure 1: Waste evolution. To protect Bekaert's proprietary data, we do not report the scale on

the vertical axis, nor do we report the estimate of the intercept.

x

- Z

84 85 86 87 88 89 90 91

Figure 2: Production and scale-up history

28

Page 30: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Unproven theories– — – Empirically proven

theories8 –

6 –

4--

2 –

16 –

14 –

12 –

10 –r

r 1

0 ■ ' t 1 1

84 85 86 87 88 89 90 91

Figure 3: Evolution of unproven theories and empirically proven theories

0.0003 –

0.00025

0.0002 T

0.00015 –

0.0001 –

0.00005 –

084 85 86 87 88 89 90 91

Figure 4: Estimated dynamic learning rate

29

Page 31: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Scale-up Performancegap

Learning bynew experiences

Wastereduction

facilitates disrupts

Deliberatelearningactivities

Empirically proven theories

Unproven theories

Artisan skills

Firefighting

Figure 5: The quality learning curve: linking learning activities to waste reduction.

30

Page 32: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

Traditional Learning Curve Estimates

blb2 b3 Adjusted R2 DW

In t In q In z

(1) —0.22*

(-11.0)

(2) —0.21*

(-11.6)

(3) —1.34*

(-18.0)

03 04

z CLL x z CLH x z CHL x z CHH x z

(4) —229* —0.56 —1.79 8.29* —7.97*

(-3.59) (-0.14) (-0.47) (3.97) (-3.81)

0.60 0.591

0.63 0.631

0.80 1.141

Adjusted R2 DW

0.84 1.56u

Quality Learning Curve Estimates

Dependent variable In W. Sample size 80. Estimates in row (4) x 10-6.

T-statistics in parentheses. * signifies significant at 0.1% in a 2-tail test.

DW statistic does not exceed the lower bound at the 1% significance level

DW statistic exceeds the lower bound but not the upper bound at the 1% significance

level

Table 1: Learning Curve Estimates

31

Page 33: BEHIND THE LEARNING CURVE: LINKING LEARNING ACTIVITIES … · 2008-06-03 · • Investments have resulted in shifts to steeper learning curves (Hax & Majluf 1982). Mishina's (1992)

unproven theories empirically proven

high theories

+3.6% * -3.5%*

firefighting artisan skills

low

—0.2% —0.8%

low high

conceptual

learning

operational learning

Table 2: Impact of four types of induced learning: percentage change in the learning rate by

adding one project (a negative sign means faster waste reduction). * signifies significant at 0.1%

in a 2-tail test.

32