bohn measuring tech knowledge ocr version

7/29/2019 Bohn Measuring Tech Knowledge OCR Version

1/13

asuring and Managing Technological Knowledgen, Roger En Management Review; Fall 1994; 36, 1; ABI/INFORM Global

61

Measuring and ManagingTechnological Knowledge

Roger E. Bohn

H ow MUCH DOES YOUR ORGANIZATION KNOW? TH E VITAL IMPACT OF ORGANIZA-TIONAL KNOWLEDGE ON PERFORMANCE IS NOW WIDELY RECOGNIZED, BUT THEstudy of how to manage such knowledge is still in its infancy. The author definestechnological knowledge and gives a framework for mapping and evaluating levels ofknowledge. He shows how to apply the framework to measure how much your orga-nization knows and doesn't know about its production processes, to learn whereknowledge resides in your company, and to make better use of what you know. Heshows why automation without adequate knowledge leads to disaster and how tomanage knowledge in a world of continual organizational learning. ~

Roger E. Bohn teaches technologyand operations management atthe University ofCalifornia, SanDiego.

"Knowledge is power. "- Francis BaconA s we move from the industrial age into the infor-mation age, knowledge is becoming an evermore central force behind the competitive suc-cess of firms and even nations. Nonaka has commented,"In an economy where the only certainty is uncertainty,the one sure source of lasting competitive advantage isknowledge.'" Philosophers have analyzed the nature ofknowledge for millennia; in the past half-century, cogni-tive and computer scientists have pursued it with in-creased vigor. But it has turned out that information ismuch easier to store, describe, and manipulate than isknowledge. On e consequence is that, although an orga-nization's knowledge base may be its single most impor-tant asset, its very intangibility makes it difficult to man-age systematically.2

The goal of this paper is to present a framework formeasuring and understanding one particular type ofknowledge: technological knowledge, i.e., knowledgeabout how to produce goods and services. We can use thisframework to more precisely map, evaluate, and comparelevels of knowledge. The level of knowledge that a processhas reached determines how a process should be con-trolled, whether and how it can be automated, the keytasks of the workforce, and other major aspects of itsSLOAN MANAGEMENT REVlEW/FALL 1994

management. Better knowledge of key variables leads tobetter performance without incremental physical invest-ment.

Two examples illustrate the importance of technolog-ical knowledge in the form of detailed process under-standing. Chaparral Steel, a minimill, was able to dou-ble output from its original electric furnace and caster.Semiconductor companies routinely increase yields ontheir chip fabrication lines from below 40 percent toabove 80 percent during a period of several years. Inthese cases, the incremental capital investments are min-imal. Th e improvements are instead due to multiplechanges in the manufacturing process, including differ-ent procedures, adjustments of controls, changes in rawmaterial recipes, etc. Wh y weren't these changes imple-mented at startup? The reason is that the knowledge aboutthe process and how to run it is incomplete and developsgradually through various kinds of learning.

Many authors have noted that there is a difference be-tween data and information. A few have also noted thatthere is a difference between information and knowl-edge. 3 Although no t always clear-cut, the distinctionamong the three in production processes is very impor-tant. Data are what come directly from sensors, reportingon the measured level of some variable. Information is

i "data that have been organized or given structure -BOHN 61


2/13

Figure 1 Diagram _oJ P r o ~ ~ _____ _

Environmentalvariables - ,constantlychanging) Process Y= Ix)Raw materials~ Complex interna I .. Output, Ystructure, onlypartly known

Control StatevariablesvariablesInputs, x

L -_____________________________that is, placed in context - and thus endowed withmeaning."4 Information tells the current or past statusof some part of the production system. Knowledge goesfurther; it allows the making of predictions, causal asso-ciations, or prescriptive decisions about what to do.

For example, consider a stream of measurements ofthe critical dimension of a series of supposedly identicalmanufactured parts - raw data. If the data are plottedon a control chart, they provide information about thestatus of the production process for those parts. Th emeasurements may have a trend, may be beyond theprocess control limit, may be out of the allowed toler-ance, or may even show no discernible pattern. All ofthese are information, but not knowledge. Knowledgeabout the process might include, "When the controlchart looks like that, it usually means machine A needsto be recalibrated" (causal association and prescriptivedecision), or "When the control chart is in control forthe first hour of a new batch, it usually remains that wayfor the rest of the shift" (prediction). This paper is abouttechnological knowledge, not data or information.

To explain why some types of knowledge are morecomplete and useful than others, a colleague and I de-veloped an ordinal scale for describing how much isknown about a process, Originally we studied ramp-upof new production in high-tech industries (VLSI fabri-cation, hard disk drives). Subsequently, we found thatthe same concepts worked well in traditional industriessuch as firearms, pulp and paper, and steel cord.5

In the next section, I give a detailed scheme for measur-ing the extent of technological knowledge and several briefexamples, ranging fro:,:, " e l l : i ~ v : , ! __ '-':'J:-., :0 consulting.The third section l'xa!llines the impli.::ariom of the level of

62 BOHN

knowledge for how to manage production processes.The fourth section looks at learning, i.e., the evolutionof knowledge over time. In the penultimate section, I usea familiar technology, baking, as an extended illustration.In conclusion, I look at some of the implications formanaging technological knowledge itself

A Scale for Measuring Knowledgeabout a ProcessA company's knowledge about its processes may rangefrom total ignorance about how they work to very for-mal and accurate mathematical models.6 For our purpos-es, a process is defined as any repetitive system for produc-ing a product or service, including the people, machines,procedures, and software, in that system, A process has in-puts, outputs, and state variables that characterize whatis happening inside it. The inputs are often further bro-ken down into raw materials, control variables, and en-vironmental variables (see Figure 1). For example, envi-ronmental variables include temperature, humidity, airpressure, dust, seismic vibration, electrical power, etc.

Here I define technological knowledge as understand-ing the effects of he input variables on the output. Mathe-matically, the process output, y, is an unknown functionf of the inputs, x: Y=f(x); x is always a vector (of indeter-minate dimension). Then technological knowledge isknowledge about the arguments and behavior of thefunction f(x).7 The manager's or process engineer's goal isto manipulate the raw materials, controls, and environ-ment to get output that is as good as possible. It is cus-tomary to treat the environmental variables as exogenousand uncontrollable. However, with enough knowledge,the environmental variables can be turned into controlvariables and, therefore, are not exogenous.

I start by looking at well-defined manufacturing pro-cesses such as building a car door or cooking in a fast-food restaurant. Later I will show how knowledge aboutless tangible processes, such as marketing and legal ser-vices, can be described by the same scale. Whatever the

I process, better technological knowledge gives the opera-tors better ability to manage the process effectively.

I have identified eight stages of technological knowl-edge, ranging from complete ignorance to complete un-derstanding. Each stage describes the knowledge about aparticular input variable Xi'S effect on the process out-put, Y. Why so many stages? We are used to the idea ofa spectrum of knowledge "from art to science," but in-tuition suggests that only three or four stages should besufficient to describe the spectrum, Most analyses ofproduction processes, however, look only at things that

SIn".", MANACE\!ENT REVIEw/FALL 1994

Figure 1 Diagram _oJ P r o ~ ~ _____ _

Environmentalvariables - ,constantlychanging) Process Y= Ix)Raw materials~ Complex interna I - Output, Ystructure, onlypartly known

Control StatevariablesvariablesInputs, x

"--------- ----------- - - - - - - - - - - - - -

that is, placed in context - and thus endowed withmeaning."4 Information tells the current or past statusof some part of the production system. Knowledge goesfurther; it allows the making of predictions, causal asso-ciations, or prescriptive decisions about what to do.

For example, consider a stream of measurements ofthe critical dimension of a series of supposedly identicalmanufactured parts - raw data. If the data are plottedon a control chart, they provide information about thestatus of the production process for those parts. Th emeasurements may have a trend, may be beyond theprocess control limit, may be out of the allowed toler-ance, or may even show no discernible pattern. All ofthese are information, but not knowledge. Knowledgeabout the process might include, "When the controlchart looks like that, it usually means machine A needsto be recalibrated" (causal association and prescriptivedecision), or "When the control chart is in control forthe first hour of a new batch, it usually remains that wayfor the rest of the shift" (prediction). This paper is abouttechnological knowledge, not data or information.

To explain why some types of knowledge are morecomplete and useful than others, a colleague and I de-veloped an ordinal scale for describing how much isknown about a process, Originally we studied ramp-upof new production in high-tech industries (VLSI fabri-cation, hard disk drives). Subsequently, we found thatthe same concepts worked well in traditional industriessuch as firearms, pulp and paper, and steel cord.5

In the next section, I give a detailed scheme for measur-ing the extent of technological knowledge and several briefexamples, ranging fro:,:, " e l l : i ~ v : ' ! __ '-':'J:-" :0 consulting.The third section l'xa!llines the impli.::ariom of the level of

62 BOHN

knowledge for how to manage production processes.The fourth section looks at learning, i.e., the evolutionof knowledge over time. In the penultimate section, I usea familiar technology, baking, as an extended illustration.In conclusion, I look at some of the implications formanaging technological knowledge itself

A Scale for Measuring Knowledgeabout a ProcessA company's knowledge about its processes may rangefrom total ignorance about how they work to very for-mal and accurate mathematical models.6 For our purpos-es, a process is defined as any repetitive system for produc-ing a product or service, including the people, machines,procedures, and software, in that system, A process has in-puts, outputs, and state variables that characterize whatis happening inside it. The inputs are often further bro-ken down into raw materials, control variables, and en-vironmental variables (see Figure 1). For example, envi-ronmental variables include temperature, humidity, airpressure, dust, seismic vibration, electrical power, etc.

Here I define technological knowledge as understand-ing the effects of he input variables on the output. Mathe-matically, the process output, y, is an unknown functionf of the inputs, x: Y=f(x); x is always a vector (of indeter-minate dimension). Then technological knowledge isknowledge about the arguments and behavior of thefunction f(x).7 The manager's or process engineer's goal isto manipulate the raw materials, controls, and environ-ment to get output that is as good as possible. It is cus-tomary to treat the environmental variables as exogenousand uncontrollable. However, with enough knowledge,the environmental variables can be turned into controlvariables and, therefore, are not exogenous.

I start by looking at well-defined manufacturing pro-cesses such as building a car door or cooking in a fast-food restaurant. Later I will show how knowledge aboutless tangible processes, such as marketing and legal ser-vices, can be described by the same scale. Whatever the

I process, better technological knowledge gives the opera-tors better ability to manage the process effectively.

I have identified eight stages of technological knowl-edge, ranging from complete ignorance to complete un-derstanding. Each stage describes the knowledge about aparticular input variable Xi'S effect on the process out-put, Y. Why so many stages? We are used to the idea ofa spectrum of knowledge "from art to science," but in-tuition suggests that only three or four stages should besufficient to describe the spectrum, Most analyses ofproduction processes, however, look only at things that

SIn".", MANACE\!ENT REVIEw/FALL 1994


3/13

Table 1 Stages of Knowledge - - ~ - - - - - - - - - - - ------ - - - -Stage Name Comment Typical Form of Knowledge

1 Complete ignorance Nowhere2 Awareness Pure art Tacit3 Measure Pretechnological Written4 Control of the mean Scientific method feasible Written and embodied in hardware5 Process capability Local recipe Hardware and operating manual6 Process characterization Tadeoffs to red uce costs Empirical equations (numerical)7 Know why Science8 Complete knowledge Nirvana

are already reasonably well understood. Variables in thefirst three stages are usually considered exogenous, inthat it is impossible to control them. Nonetheless, it isimportant to recognize their existence since importantvariables may be at one of those stages, and managementof the process needs to take that into account. The stagesare summarized in Table 1.

In contrast to most approaches for measuring knowl-edge, the nature of the knowledge changes qualitativelywith each stage in this framework. The process of learn-ing from one stage to the next also changes. Each stageis described as follows:

Stage One - Complete ignorance. You do not knowthat a phenomenon exists, or if you are aware of its exis-tence, you have no inkling that it may be relevant toyour process. The history of technology is full of phe-nomena that were initially not recognized, yet had po-tentially major effects on a production process (e.g.,quantum mechanics, germs in the treatment of wounds,contamination in a number of processes). At stage one,there is nothing you can do with the variable, and its ef-fects on the process appear as random disturbances.

Stage Two - Awareness. You know that the phe-nomenon exists and that it might be relevant to yourprocess. There is still no way to use the variable in yourprocess, but you can begin to investigate it in order toget to the next stage. Learning from stage one to stagetwo often occurs by serendipity, by making analogies toseemingly unrelated processes, or by bringing knowl-edge from outside the organization.

Stage Three - Measure. You can measure the vari-ables accurately, perhaps with some effort. This requiresdevelopment and installation of specific instrumenta-tion. Stage three variables cannot be controlled. How-ever, if the variable is important enough, you can alterthe process in response to the variable in order to exploit

SLOAN MANAGEMENT REVIEw/FALL 1994

Scientific formulas and algorithms

or ameliorate its effects. An example of a stage threevariable is weather; many outdoor processes are haltedor done differently during bad weather.

There are two kinds oflearning at stage three. One kindconsists of passive, natural experiments to determine therelationship between this variable and the output. A sec-ond learning process studies ways of controlling the vari-able to reach stage four, control. Knowledge about howto control the variable is, in effect, a subprocess with itsown inputs and output (the level of the input variablefor the main process). For certain variables, knowinghow to measure it (stage three) leads almost automatical-ly to knowing how to control it (stage four). These areprimarily variables where feedback-based control is feasi-ble, such as furnace temperatures.

Stage Four - Control of he mean. You know how tocontrol the variables accurately across a range of levels,although the control is not necessarily precise. That is,you can control the mean level, bu t there is some vari-ance around that level. Stage four provides a quantumleap in process control, since, at a minimum, you cannow stabilize the process with respect to the mean ofthat variable. Variables that were previously viewed asexogenous disturbances to the process can now be treat-ed as control variables. Reaching stage four also makesfurther learning easier, because you can now performcontrolled experiments on the variable to quantifY its im-pact on the process.

Stage Five - Process capability (control of the vari-ance). You can control the variables with precision acrossa range of values. When all of the important variablesreach stage five, your process can manufacture productsby following a "cookbook," i.e., a consistent recipe. Theproduct still may not meet quality standards, however,so final inspection will be needed.

Learning from stage four to stage five is a matter of

BOHN 63

Table 1 Stages of Knowledge - - ~ - - - - - - - - - - - ------ - - - -Stage Name Comment Typical Form of Knowledge

1 Complete ignorance Nowhere2 Awareness Pure art Tacit3 Measure Pretechnological Written4 Control of the mean Scientific method feasible Written and embodied in hardware5 Process capability Local recipe Hardware and operating manual6 Process characterization Tadeoffs to red uce costs Empirical equations (numerical)7 Know why Science8 Complete knowledge Nirvana

are already reasonably well understood. Variables in thefirst three stages are usually considered exogenous, inthat it is impossible to control them. Nonetheless, it isimportant to recognize their existence since importantvariables may be at one of those stages, and managementof the process needs to take that into account. The stagesare summarized in Table 1.

In contrast to most approaches for measuring knowl-edge, the nature of the knowledge changes qualitativelywith each stage in this framework. The process of learn-ing from one stage to the next also changes. Each stageis described as follows:

Stage One - Complete ignorance. You do not knowthat a phenomenon exists, or if you are aware of its exis-tence, you have no inkling that it may be relevant toyour process. The history of technology is full of phe-nomena that were initially not recognized, yet had po-tentially major effects on a production process (e.g.,quantum mechanics, germs in the treatment of wounds,contamination in a number of processes). At stage one,there is nothing you can do with the variable, and its ef-fects on the process appear as random disturbances.

Stage Two - Awareness. You know that the phe-nomenon exists and that it might be relevant to yourprocess. There is still no way to use the variable in yourprocess, but you can begin to investigate it in order toget to the next stage. Learning from stage one to stagetwo often occurs by serendipity, by making analogies toseemingly unrelated processes, or by bringing knowl-edge from outside the organization.

Stage Three - Measure. You can measure the vari-ables accurately, perhaps with some effort. This requiresdevelopment and installation of specific instrumenta-tion. Stage three variables cannot be controlled. How-ever, if the variable is important enough, you can alterthe process in response to the variable in order to exploit


Scientific formulas and algorithms

or ameliorate its effects. An example of a stage threevariable is weather; many outdoor processes are haltedor done differently during bad weather.

There are two kinds oflearning at stage three. One kindconsists of passive, natural experiments to determine therelationship between this variable and the output. A sec-ond learning process studies ways of controlling the vari-able to reach stage four, control. Knowledge about howto control the variable is, in effect, a subprocess with itsown inputs and output (the level of the input variablefor the main process). For certain variables, knowinghow to measure it (stage three) leads almost automatical-ly to knowing how to control it (stage four). These areprimarily variables where feedback-based control is feasi-ble, such as furnace temperatures.

Stage Four - Control of he mean. You know how tocontrol the variables accurately across a range of levels,although the control is not necessarily precise. That is,you can control the mean level, bu t there is some vari-ance around that level. Stage four provides a quantumleap in process control, since, at a minimum, you cannow stabilize the process with respect to the mean ofthat variable. Variables that were previously viewed asexogenous disturbances to the process can now be treat-ed as control variables. Reaching stage four also makesfurther learning easier, because you can now performcontrolled experiments on the variable to quantifY its im-pact on the process.

Stage Five - Process capability (control of the vari-ance). You can control the variables with precision acrossa range of values. When all of the important variablesreach stage five, your process can manufacture productsby following a "cookbook," i.e., a consistent recipe. Theproduct still may not meet quality standards, however,so final inspection will be needed.

Learning from stage four to stage five is a matter of

BOHN 63


4/13

learning to control the various disturbances that affectthe input variable. This is a nested subproblem thatpasses through the stages of knowledge on the way togood control of the input variable. That is, producingthe correct level of an input, x, is a process in its ownright and must be learned. Fortunately, accumulatedtechnological knowledge gives cookbook methods forcontrolling many variables. Th e process engineer canlook it up in a catalog or handbook. This means thatyou do not have to "reinvent the wheel" each time; youjust have to learn enough to control the variable usingknown "wheels."Managing in high-techindustries requires bothrapid learning and theability to manufacture with

"immature" (low stage ofknowledge) technologies.

Stage Six - Process characterization (know how). Youknow how the variable affects the result, when smallchanges are made in the variable." Now you can beginto fine-tune the process to reduce costs and to changeproduct characteristics. You can also institute some feed-back control on the output using any stage six variablethat is both easy to change and has a major impact. Thisincreases the quality of the output by reducing its vari-ability. To reach stage six, you run controlled experi-ments with different levels of the variable to determineits effects.

Stage Seven - Know why. You have a scientific modelof the process and how it operates over a broad region,including nonlinear and interaction effects of this vari-able with other variables. At this stage, you can actuallyoptimize the process with respect to the stage seven vari-ables. Feedback and some feed-forward control arebroadly effective. Control can be turned over to micro-processors, which will be able to handle most contin-gencies. You can even use your knowledge to simulatethe process to study settings you have never tried empir-ically, such as ways of making new products using thesame process. Learning from stage six to stage seven in-volves tapping scientific models, running broad experi-ments across multiple variables to estimate the models,and finding interactions among input variables.

Stage Eight - Complete knowledge. You know the com-

64 BOHN

plete functional form and parameter values that determinethe result, Y, as a function of all the inputs. Process andenvironment are so well understood that you can headoff any problems in advance by feed-forward control.Stage eight is never reached in practice because it re-quires knowing all the interactions among variables.However, it can be approached asymptotically by study-ing the process in more and more detail.The stages of knowledge can be applied to diversetasks and industries: High-tech manufacturing requires rapid learning aboutmultiple variables in new products and processes. Wecan frame a definition in terms of the stage of knowl-edge: high-tech processes are those in which many of heimportant variables are at stage four or below. This makesthe process difficult to control and work with, so a lotof effort goes into raising the knowledge level as quicklyas possible. Because of customer and competitive pres-sures, no sooner is knowledge raised for one productthan higher performance products are demanded, whichbrings in new low-stage variables. Thus managing inhigh-tech industries requires both rapid learning andthe ability to manufacture with "immature" (low stageof knowledge) technologies. VLSI semiconductor design and fabrication processesare driven by the ability to reproduce very small featureswith high reliability at high volume. The process is verycomplex, with multiple layers and hundreds of variablespotentially affecting each layer. As feature sizes get small-er with each new generation, new equipment is neededand new variables become important. These new variablesstart at low stages of knowledge. For example, as featuresizes go below one micron, heat dissipation problemsbegin to push designers to engineer chips for three voltsinstead of five volts. This has a number of advantagesbut requires many changes in both chip design and fab-rication. As these changes are made, the variables thatwere at stage six or seven for the old process "regress" tostage five; engineers know how to control them, but don'tknow their effects on the new process. Consumer marketing has made many strides towardhigher stages of knowledge in the past thirty years. Manyof the breakthroughs have been based on developing ef-fective ways to measure variables (stage three). For ex-ample, bar-code scanners at supermarket checkouts haveprovided masses of disaggregated data about who is buy-ing what, whether they use coupons, etc. Some stores arenow using customer ID cards to match this data with in-formation about individual households, their demo-graphics, what TV commercials they received, and otherenvironmental variables to allow development of stage

SLOAN MANAGEMENT REVIEW/FALL 1994

learning to control the various disturbances that affectthe input variable. This is a nested subproblem thatpasses through the stages of knowledge on the way togood control of the input variable. That is, producingthe correct level of an input, x, is a process in its ownright and must be learned. Fortunately, accumulatedtechnological knowledge gives cookbook methods forcontrolling many variables. Th e process engineer canlook it up in a catalog or handbook. This means thatyou do not have to "reinvent the wheel" each time; youjust have to learn enough to control the variable usingknown "wheels."Managing in high-techindustries requires bothrapid learning and theability to manufacture with

"immature" (low stage ofknowledge) technologies.

Stage Six - Process characterization (know how). Youknow how the variable affects the result, when smallchanges are made in the variable." Now you can beginto fine-tune the process to reduce costs and to changeproduct characteristics. You can also institute some feed-back control on the output using any stage six variablethat is both easy to change and has a major impact. Thisincreases the quality of the output by reducing its vari-ability. To reach stage six, you run controlled experi-ments with different levels of the variable to determineits effects.

Stage Seven - Know why. You have a scientific modelof the process and how it operates over a broad region,including nonlinear and interaction effects of this vari-able with other variables. At this stage, you can actuallyoptimize the process with respect to the stage seven vari-ables. Feedback and some feed-forward control arebroadly effective. Control can be turned over to micro-processors, which will be able to handle most contin-gencies. You can even use your knowledge to simulatethe process to study settings you have never tried empir-ically, such as ways of making new products using thesame process. Learning from stage six to stage seven in-volves tapping scientific models, running broad experi-ments across multiple variables to estimate the models,and finding interactions among input variables.

Stage Eight - Complete knowledge. You know the com-

64 BOHN

plete functional form and parameter values that determinethe result, Y, as a function of all the inputs. Process andenvironment are so well understood that you can headoff any problems in advance by feed-forward control.Stage eight is never reached in practice because it re-quires knowing all the interactions among variables.However, it can be approached asymptotically by study-ing the process in more and more detail.The stages of knowledge can be applied to diversetasks and industries: High-tech manufacturing requires rapid learning aboutmultiple variables in new products and processes. Wecan frame a definition in terms of the stage of knowl-edge: high-tech processes are those in which many of heimportant variables are at stage four or below. This makesthe process difficult to control and work with, so a lotof effort goes into raising the knowledge level as quicklyas possible. Because of customer and competitive pres-sures, no sooner is knowledge raised for one productthan higher performance products are demanded, whichbrings in new low-stage variables. Thus managing inhigh-tech industries requires both rapid learning andthe ability to manufacture with "immature" (low stageof knowledge) technologies. VLSI semiconductor design and fabrication processesare driven by the ability to reproduce very small featureswith high reliability at high volume. The process is verycomplex, with multiple layers and hundreds of variablespotentially affecting each layer. As feature sizes get small-er with each new generation, new equipment is neededand new variables become important. These new variablesstart at low stages of knowledge. For example, as featuresizes go below one micron, heat dissipation problemsbegin to push designers to engineer chips for three voltsinstead of five volts. This has a number of advantagesbut requires many changes in both chip design and fab-rication. As these changes are made, the variables thatwere at stage six or seven for the old process "regress" tostage five; engineers know how to control them, but don'tknow their effects on the new process. Consumer marketing has made many strides towardhigher stages of knowledge in the past thirty years. Manyof the breakthroughs have been based on developing ef-fective ways to measure variables (stage three). For ex-ample, bar-code scanners at supermarket checkouts haveprovided masses of disaggregated data about who is buy-ing what, whether they use coupons, etc. Some stores arenow using customer ID cards to match this data with in-formation about individual households, their demo-graphics, what TV commercials they received, and otherenvironmental variables to allow development of stage



5/13

six and seven models of the marketing mix'seffects on consumer behavior.9 Figure 2 AKnowledge Tree Professional services such as legal servicesrun the range of knowledge stages. For exam-ple, preparing a will has reached stage six oreven seven for many people, so that it can bedone by a $30 software program. At the otherextreme, high-profile criminal trials used tobe at stage three or below. Recently, a numberoflaw firms have attempted to move jury se-lection to stage six, using methods such as cus-tomized polling of population groups fromwhich a particular jury will be drawn. Other

Process Output Y

aspects of trial strategy, presumably, remain atstage three or four; they can be measured butnot controlled well. For example, an impor-tant type of"input" to litigation is judicial rul-ings on motions. Lawyers can use the judge'sruling to measure whether the judge agrees

Key- Each branch represents avariableaffecting the process.

with them on a motion, but they have onlylimited control over that decision (stage four).Pretrial aspects of litigation, on the other

(invisible)

Stage 7Stage 6Stage 5Stage 4Stage 3Stage 2Stage 1 The thickness of each branch representsthe importance of that variable.

hand, are generally better understood. In strategic consulting, the Boston Consulting Group'sfour-quadrant matrix (cash cows, dogs, stars, and ques-tion marks) was an attempt to reduce acquisition anddivestiture decisions to two quantitative variables -market share and growth rate. 1O It is possible to writeequations that describe the effects of market share andgrowth rate on business unit profit, so these two vari-ables are at stage six. But there are many other impor-tant variables that also influence the outcome and thatare at much lower stages of knowledge. Many consult-ing firms claim knowledge about these other variables,but they perform strategic analysis using a heavy mix ofexpertise, implying an awareness that some of theirknowledge is at a low stage.Dynamic Evolution of Knowledge andPerformanceImportant variables are those that, in fact, have great eco-nomic implications for the process. Ideally, a companywould like to have a high stage of knowledge about allthe important variables and a low stage about all the vari-ables that have negligible effects. But, instead, the organi-zation is likely to know very little about some importantvariables, especially for immature processes. Conversely, itmay have stage six knowledge about unimportant vari-ables, such as the color of paint on the machine and thetype of clothing workers wear. Of course, in certain pro-


cesses, these variables may be important, but there maybe little way to know this until you learn enough to bringthem to a high stage. For example, paint inside a machinemay affect process chemistry, paint outside a machinemay affect worker morale, and worker clothing can affectcontamination-sensitive processes.

One way to visualize overall technological knowledgeis as a tree (see Figure 2). The trunk of the tree is Y, theprocess output that we want to control. The branchesfrom the trunk are variables that directly affect Y, (Xl> X2'.. . ). Branching off from each of these are subvariables(Xl.l> xu' .. . ) that collectively determine Xl> and so on,to any level of detail. The shading of each branch repre-sents the organization's stage of knowledge, with white(invisible) representing stage one, while black is stageseven. The thickness of the branch represents its impor-tance. Every knowledge tree trails off into a haze ofdimly seen bu t potentially important variables andeventually becomes invisible, because there are alwayssome variables at a still finer level of detail whose exis-tence is unrecognized.

As the tree illustrates, a single process has many vari-ables that are inevitably at different stages of knowledge.As more is learned about part of the process, old vari-ables are brought to higher stages, but new variables alsoemerge from the mists of ignorance. The process as awhole can do no better than the knowledge about itsmost important drivers. If even a few key variables are at

BOHN 65


6/13

Figure 3 Learning Curve and Knowledge-Based views of Organizationallmproveme_n_t ~ ~ __ ~ ~ ~ ~ ~ _A. The Learning Curve

Production

ImprovedCost andQuality

i i i B. Inside the Learning CurveProduction ~ICP LI L e _ a _ r n _ i n _ g - - - - - - ,Deliberate /

BetterOrganizationalKnowledge

ChangedBehavior

ImprovedCost andQuality

Activities

low stages of knowledge, the process can be consideredat a low stage of knowledge overall.Relationship to Theories ofOrganizational LearningExperience in conducting a task generally leads to im-provement, a concept formalized in the literature onIeven a few key variables are atlow stages of knowledge theprocess can be considered at a

low stage of knowledge overall.- ~ ~ - - - - - - - - -

learning curves. ll Most learning curve models skip the in-termediate stages of causality and statistically link cumu-lative production directly to costs (see Figure 3, part A).But it is clear that how the production and learning pro-cesses are managed has a big impact on whether and howfast learning occurS. 12 Indeed, the large amount of litera-ture on quality improvement concerns systematic learn-ing methods to achieve more improvement in a shorterperiod of time. Thus learning can be a directed activity,not just a by-product of normal production. Part B in

66 BOHN

Figure 3 shows a more complete model of technologicallearning, with explicit recognition of knowledge. 13

It is no coincidence that the knowledge tree of Figure2 resembles causal trees like those used in quality im-provement efforts.14 These trees, also called fishbone orIshikawa diagrams, are often used as a way of listing po-tential causes of problems. A process engineer may havefifty variables (or corresponding problems) at stages twothrough four that are potentially important. Variousmethods can be used to guess which ones will turn outto be the most important. 15 The stages of knowledgeprovide a way of mapping current knowledge and esti-mating how hard it will be to go further on particularvariables. That is, they provide a detailed scorecard forprocess improvement efforts.How to Manage at Each Stage ofKnowledgeThe knowledge stage of different process variables is im-portant because it determines how to manage both theknowledge and the production process. The higher thestage of knowledge, the closer the process is to "science,"and the more formally it can be managed. Conversely,low-stage processes, such as creative endeavors, do not do

SWAN MANAGEMENT REVIEw/FALL 1994


7/13

well under formal management methods, and should betreated more as "art."

One of the most basic system-design decisions is thedegree of procedure. There are different ways of per-forming a given task, requiring different kinds of peo-ple, training, and tools. At one extreme is pure proce-Te higher the stage ofknowledge, the closer theprocess is to "science,"and the more formally it can

be managed.dure, i.e., a completely specified set of rules about whatto do under every possible set of circumstances. At theother extreme is something we can call pure expertise orpure art - a style of action in which every situation isdealt with as if it were new and unique. This requiresexperienced and skilled people who use their own judg-ment at each moment. These people have tacit knowl-edge, meaning that although they can carry on a task,they are not able to explain it.

Managers can attempt to operate a

edge, it is inefficient to use lots of expertise to carry itout. An expertise-based process may still work (althoughpeople lose attentiveness in purely repetitive situations),but you will pay extra for experts who are not reallyneeded. This is the area below the diagonal in Figure 4.

Why do companies find themselves off the diagonalof Figure 4? A common reason during the early 1980swas hubris: overoptimism about the firm's knowledge ofproduction processes and its associated ability to build,debug, and operate new factories. This led to numerousattempts to solve manufacturing competitiveness prob-lems by automation, as exemplified by the slogan, "au-tomate, emigrate, or evaporate." When automation wasundertaken without a solid base of process knowledge,the results were counterproductive: "The automation of alarge, complex, poorly understood, conventional manu-facturing process leads to a large, complex, poorly under-stood, unreliable, expensive, and automated manufactur-ing proceSS."I? Perhaps one of the most conspicuous andexpensive examples of this syndrome was General Motors,which, in the early 1980s, invested approximately $40billion to build a number of automated auto assemblyplants, many of which never worked properly.

At the other, perhaps less common extreme are compa-nies that use expensive labor to perform repetitive tasks,

process anywhere along the spectrumfrom pure expertise to pure procedure.The microprocessor has made it possi-ble to execute very complex procedures

Figure 4 Ideal Operating Method and the Stage of Knowledge

at very low cost.16 But this does no tmean that procedural approaches arealways best. There is a natural relation-ship between degree of procedure andstage of knowledge (see Figure 4). Forexample, in order to automate a pro-cess, all key variables should be under-stood at least to stage six, and preferablyto stage seven. If they are not, unanti-cipated problems will crop up frequent-ly, and the system will not be able todeal with them effectively. Those por-tions of processes that are at low stagesof knowledge should be done using ahigh degree of expertise and little au-tomation. Locations above the diagonalin Figure 4 correspond to inexpensivebut ineffective processes, which do notproduce consistently good output.

Conversely, if a process or portion ofa process is at a high stage of knowl-

SLOAN MANAGEMENT REVlEW/FALL 1994

100% , - - - - - - - - - - - - - - - - - - - - - - ,

'0uI!!enuCI

Zone of Ineffectiveness - I... : .'.. If.' . "'"11.;' -- -..; 111,-. . . : . . "_.. . ..- -.1.. ~ ~ J I .. ~ . : t.-.. :'.. ~ ' . - ~ ." J i t ~ ... to.... ... U .. rI. " . _ C ~." - __ -.. IJI : ."* ~ . ~ . ." ' . ,. i :". " , ~ 1 . ; - ."'." ,' . ,. ":.\."1-."" .- ... ,. '. 1 \ ~ " ! ' " : ; ~ " . . . . .. , , ~ ..... + $ _" I C " ~ t , ........... : .. ~ ~ ....... ...-... ..' . ".... ~ ... .. f ' . ' 4, Zone of Inefficiency

.... 0% '.1 2 3 4 5 6 7 8Stage of Knowledge

Source: R.E. Bohn and R. Jaikumar, "The Development of Intelligent Systems for Industrial Use: An EmpiricalInvestigation," in Research on Technological Innovation, Management and Policy, ed. R. Rosenbloom (London andGreenwich, Connecticut: JAI Press, 19861, pp. 213-262.

BOHN 67


8/13

Table 2 Some Effects of Knowledge StagesKnowledge at Stage . .. 2 3 4 5 6 7 8Nature of production Expertise based ... Procedure based

Role of workers Everything Problem solving Learning and improvingLocation of knowledge Workers' heads Written and oral In databases or in

softwareNature of learning Artistic Natural experiments Controlled experiments.simulationsNature of problem T ial and error Scientific method Table look-upsolvingMethod of training Apprenticeship. ... Classroomnew workers coachingNatu ra I ype of Organic Mechanistic Learning orientedorganizationSuitability for None ... HighautomationEase of transfer to Low ... Highanother siteFeasible product variety High Low High*Qual ity control approach Sorting Statistica I Feed forwardprocess control

leasee ding to inefficiency. Examples include information-based services such as manual letter sotting (e.g., the U.S.Postal Service) and routine telephone services (directoryassistance). Although human judgment is very useful inthese processes for handling exceptions, the bulk of thework is routine, well understood, and uses mainly the pat-tern recognition abilities of the human brain. Industrieshave taken several approaches to dealing with the resultinginefficiency, including high proceduralizing of workers,which risks dehumanizing the work and suppressing theirexpertise (e.g., United Parcel Service industrial engineer-ing and automated monitoring of telephone operators),and finding ways of getting data into machine readableform so that human operators do not have to keypunch it(optical character recognition and bar coding).

The degree of procedure is not the only managerialdecision affected by the stage of knowledge. Methods oforganizing, methods of problem solving, learning, andtraining, and many other aspects of the process shouldalso be adjusted (see Table 2).

Yet, as shown in Figure 2, most processes have im-portant variables at widely differing stages of knowl-edge. The ideal management style for the process as awhole is an uncomfortable hybrid. The traditional ap-

68 BOHN

proach tifferentfunctional departments, which are then managed ac-cording to their own needs. A common example of thisin traditional manufacturing companies is R&D (lowstages) versus manufacturing functions (high stages, orso it was believed). This Taylorist approach has brokendown in modern manufacturing, especially for tech-nologies that are evolving rapidly, because the less ma-ture portions of the process are inevitably at low stagesof knowledge.18

There are at least two other approaches to this para-dox. One is to use microprocessors (or other automa-tion) to execute procedures, but with human oversight toselect the appropriate program and to recognize unpro-grammed contingencies and take control. Examples in-clude accounting, continuous manufacturing processessuch as paper mills, and commercial aviation. A final ap-proach is to use low-skilled workers to execute the betterunderstood tasks, with experts monitoring and directingthem. The low-skilled workers may be apprentices to theexperts or on a separate career track. For example, lawoffices use both junior associates (apprentices) and para-legals.

All three approaches have weaknesses. For example, it



9/13

is difficult for pilots to monitor autopilots reliably during seem to make any difference. Control of the ingredients islong flights without taking an active role themselves, yet straightforward, using a standard measuting cup; that is,respond quickly and appropriately in emergencies. 19 If for the raw materials, stage three leads immediately tolower skilled workers perform the better understood and stage four.therefore more procedural tasks, this can lead to excessive Stage Five - Process capability and a recipe. You prac-division oflabor, poor coordination, and lost opporcuni- , tice measuring ingredients until you can do it with 95ties for learning. In addition, cultural conflict is a com- I percent repeatability. You write down a set of instruc-mon result when an organization is split into sections tions (recipe) that seems to produce "adequate" cookies.operating at different stages of knowledge. Thus there is Your cookies now have a reasonably consistent taste, butno ideal solution to the problem of working at multiple texture and appearance are still variable and some cook-stages of knowledge, or if there is one, we don't yet know ies are burned.it. Nonetheless, this situation is increasingly common. Stage Six - Process characterization. You run a series ofA Simple Example ofKnowledgeProgression over TimeKnowledge increases through learning. Much learning issimply increasing the precision and accuracy of parame-ter estimates within a single stage, but sometimes learn-ing shifts the knowledge to the next stage. To illustrate,using familiar technology, suppose you are baking cook-ies for the first time. You hope to make chocolate chipcookies, but have only a vague idea of a good recipe (rawmaterials) and procedure (control variables). You have astandard oven, which you were told to set at 350 de-grees. 20

The first step is to define your output measure, Y. Itconsists of a combination of taste, texture (hard or soft),and appearance.

Stage One - Complete Ignorance. You don't even knowwhat influences cookie characteristics, so when the resultschange, you consider it "random."Stage Two - Awareness. You rack your memory, ob-serve others in the kitchen, and begin to build a list ofpossibly relevant input variables, including the list of in-gredients, baking time, outdoor weather (rainy, cloudy,clear), time of day, amount and brand name of each in-gredient, and a vaguely defined "mixing procedure."

Stage Three - Learning to measure key variables. Youuse your watch to measure cooking time, measuringcups to measure raw materials, an outdoor thermometerand hygrometer for the weather, and a clock for thetime of day. You have no detailed metric for mixing pro- Icedure, so you throw everything into one bowl andcount strokes of the mixing spoon.

Stage Four - Control of he mean. You get a count-down timer and develop a procedure to take the cookiesout of the oven after a set amount of time. You can con-trol outdoor weather only crudely, by baking on dayswhen the weather is of a particular type. You decide notto bother controlling for time of day since it does not


experiments on many variables, including baking time,baking temperature, mixing time, and the exact amountsof flour, sugar, and liquid ingredients. You discover theeffects of a 10 percent change in each of these variableson the cookie characteristics. If a friend asks for a betterbaked cookie, you can now achieve it by varying eitherMUch learning is simplyincreasing the precisionand accuracy ofparameter estimates within a single

stage, but sometimes learning shirtsthe knowledge to the next stage.

- - - --- --------- -------the time or the temperature. You discover that somevariables, including weather and time of day, have nodetectable effect on the output.

Stage Seven - Know why, including interactionsamong input variables. You go to the local university li-brary and take out textbooks on baking, which givemathematical formulas for outcome variables such assweetness and surface texture. You calibrate those mod-els using data from your own baking process. You cannow produce a "near perfect" chocolate chip cookie. Ifsomeone asks for a healthier cookie (less sugar), you canproduce it, and you know how much to adjust the bak-ing temperature. Similarly, if you are in a hurry, youknow how to increase the temperature and decrease thebaking time without burning the cookies.

Repeat for secondary variables. Although you nowhave stage five control (a recipe) for about ten variablesand a stage seven understanding (know why) of five ofthem, there will always be a host of secondary variablesin your knowledge tree that have smaller effects. Andthere is no guarantee that you will learn about the most

BOHN 69


10/13

important variables first. For example, you may not re-alize that cookie size is important (stage two) until youare well into stage five for other variables. You can sub-ject these additional variables to the same progressionthrough the stages of knowledge. Variables include thebrand and characteristics of raw materials (butter versusmargarine versus inexpensive margarine, types of flour),the importance of sifting dry ingredients together beforemixing, type of baking tray (aluminum versus glass ver-sus iron), and use of a scale instead of measuring cupsfor more accurate measurement of raw materials. Forcasual baking, you would never bother to learn aboutsome of these variables, bu t if you wanted to reducecosts or improve consistency, you would have to delvemuch deeper into these secondary variables.

Stage Eight - Complete knowledge. Since there is aninfinitude of potential secondary variables, you can neverhave complete knowledge of the cookie-making pro-cess. 21 But for practical purposes, you can say that youhave reached stage eight when you have a model thatwill predict output (cookie) characteristics to an accuracyof one-tenth of the tolerance band, for changes in inputsacross a 2: 1 range, and including all interactions.

Amateurs may stop when they have stage five knowl-edge about the primary variables that affect taste. Theycan then bake decent cookies and throwaway batchesruined by low knowledge about secondary variables. Butprofessional bakeries must track down additional sec-ondary variables, especially those that influence costs.Here is a description of the situation at one famous bak-mg company:

Since early this decade, Nabisco has been worried aboutits bakery technology, which, according to a 1981 study,had allen far behind that of ven some tiny rivals. . . . Thebiscuit company, to this day, uses a lot ofequipment madedecades ago at Nabisco's former Evanston, Illinois, machineshop.

And to this day, baking at Nabisco remaim somethingofan art. Oreos have uneven swaths ofcream filling. Theexact number ofRitz crackers in a box is anybody's guess.Some 5 percent to 7percent ofNabisco's cookies and crackers emerge from its ovens broken.

Similarly, the company still has poor inspection methodsfor the tons ofcommodities it purchases, such as flour andcocoa, according to a former executive of he baking unit.The bakers must repeatedly test-bake batches of ookies andcrackers to adjust ovens and other gear to slight variatiomin commodity composition. [In our terms, they had stagefour knowledge of raw materials and were attempting tocompensate for it by using stage six knowledge about

70 BOHN

how to adjust the ovens.] Such trial-and-error methodsmake quality control among other things, difficult.

So, sixteen months ago, . . . the company planned tospend some $1.6 billion on complete retrofitting offour existing bakeries and close five other plants.

The plan called for a microchip revolution in Nabisco'sbakeries. At least one-third of he project's cost was to be forthe purchase ofcomputerized weighing, mixing, packaging,and process-control equipment, says a senior Nabisco manufacturing engineer who recently resigned

Such high-tech gear would eventually halve the company's 8 percent "give-away" rate - the overweight amountin an average package ofNabisco biscuits - and sharplyreduce its 5percent to 7percent breakage.22

Nabisco's automation will be most effective with stageseven knowledge (know why) about all of the key vari-ables. It is possible that Nabisco's equipment vendors sellmachines that already embody that knowledge, but it islikely that some of it (including the specific variablesuniquely affecting Nabisco's cookies) would have to bedeveloped as part of the automation program.

Applying the Stages ofKnowledgeNow that we have a framework for measuring and un-derstanding technological knowledge, we can look atsome principles for managing knowledge to improveproduction processes.Understand How Much You Know and Don't KnowIn order to understand how much you already knowabout a process, you need to ask a number of questions: What are the important variables for the process? At what stages are these variables? Which variables inthe process would give the most leverage if you couldget them to a higher stage?

, How can you manage the process well at these stagesof knowledge? What limits and opportunities does theprocess impose? Are your management methods consis-tent with knowledge levels (Figure 4 and Table 2)? Howshould you handle the inevitable variables that youknow less about yet are still important? How can you learn to reach higher stages of knowledge?You also need to beware of what you think you knowabout a process that you really don't. One of the mostpainful forms of ignorance is false knowledge. If yourcompany believes that it has stage six or higher knowl-edge about a variable, bu t in fact that knowledge isbased on past experience and is incorrect for the presentprocess, you will operate the process in an inferior way.



11/13

A common version of this is the belief that "variable xdoes not matter." It may not have mattered ten yearsago became of a small contribution to process variance.But what was considered small ten years ago may bequite important today. A newer competitor, unbur-dened with this false knowledge, can control or changethe level ofx to get superior quality or lower cost.

Th e countermeasure for this problem is to realizethat as your company's process changes, its effectiveknowledge regresses to earlier stages. In particular, stagesix knowledge, which is generally derived by empiricalobservation, often regresses to stage five for a new pro-cess. You still know how to measure and control thevariable, but you no longer know its true impact.Understand and Manage the Locations ofKnowledgeKnowing where knowledge resides for the process you aremanaging is important for effectively managing and usingthat knowledge. It has implications for accessibility, trans-mission to new locations, and ability to extend theknowledge, among other things. Technological knowl-edge may be located in people's heads, word of mouth, orother informal mechanisms; in formal procedure sheetsfor operators, handbooks, other written documentation;or embodied in machinery, firmware, and software. Howwell is it documented? How easy is it to change? Howmuch do users know about how to use its features?

As I have discussed, the feasible and desirable loca-tions of knowledge depend on its stage. There are alsobroader issues surrounding more general forms of orga-nizational memory. 23Be Wary of Deskilling the Workforce and FreezingProcessesTh e Taylorist model of manufacturing, as it is com-monly applied, moves technological knowledge aboutthe process away from line workers and puts it in theheads of staff engineers. These engineers will be lessavailable when problems come up, or they may leave thecompany. If workers do not understand the process,they cannot handle unanticipated situations, nor canthey do much to improve the process, even if they aremotivated. Therefore, one of the revolutionary effects ofthe total quality management movement has been to re-turn knowledge to the workers and make them capableof doing process improvement in small groups, withoutrelying on the traditional staff experts.

Even if you fully understand a process today, theworld will change in a few years. Some of your currentknowledge will be obsolete, and it will be important to


reevaluate it. Once a firm assumes, for whatever reason,that it has nothing more to learn about a productionprocess, it tends to "lock in" the present productionmethods by specifYing rigid procedures that can deskillthe workforce and cut back on product and process engi-neering. A firm may use time and motion studies to findthe "one best way" to produce and lose interest in rootcause analysisY While this may work well in the shortrun, five years from now the company may find com-petitors making superior products at two-thirds its cost.

For example, Jaikumar compared the developmentand use of flexible manufacturing systems (FMS) in theUnited States and in Japan. 25 He found that the U.S.systems had been developed with overly ambitious goalsfor flexibility, up-time, labor use, etc. These goals werenot achieved by the initial designs; the knowledge basewas not adequate to make them possible. Yet the pro-jects were often declared complete, and workers withmuch lower skills were brought in to run the FMS. Theresult was that the users were afraid to experiment andIworkers do not understand aprocess, they cannot handleunantiCipated situations, nor

can they do much to improve theprocess, even if they are motivated.

learn about the systems, and the systems were in factused in a very inflexible way. In contrast, in the success-ful Japanese systems, the original developers stayed withthe system for the first year or more of operation, andcontinued to improve it during that time. Th e result, was systems that were very flexible and robust enoughto run unattended.Learn Carefully and SystematicallyAs we have seen, different stages of knowledge requirevery different methods of learning. For example, Chewand others recommend sequential use of four differentmethods of learning about problems that occur duringthe installation of new technology: Vicarious learning - learning from other organiza-tions with similar situations. Simulation - building a model of your process andexperimenting with the model. Prototyping - taking a subset of your process andusing it for testing and refining.

BOHN 71


12/13

On-line learning - experimenting systematically onthe full process.26

Many organizations become proficient at only one ora few methods of learning, which makes it difficult forthem to deal with variables that are at different stages ofknowledge. 27 For example, many plants avoid the use ofpilot lines and simulators to pretest process changes.

ConclusionLord Kelvin, in the 1890s, commented on the value ofknowledge:

When you can measure what you are speaking about,and express it in numbers, you know something about it;but when you cannot measure it, when you cannot expressit in numbers, your knowledge is ofa meager and unsatis-factory kind it may be the beginning of knowledge, butyou have scarcely, in your thoughts, advanced to the stage o fsctence.

In terms of my framework, Kelvin was advocatingthe value of stage three k n o ~ l e d g e (measure) over stagetwo knowledge (awareness). As I have shown, being ableto measure is only the beginning; the stages of knowl-edge beyond stage three (control, capability, characteri-zation, and know why) give additional power and eco-nomic value to a company's processes. The stages-of-knowledge framework provides powerful leverage to ef-forts to improve processes and conveys informationabout how to manage. A company can make explicitdecisions about which portions of the knowledge tree topursue most vigorously.For example, a high-volume, forty-year-old, continu-ous process was controlled using incremental extensionsof the original sensors. These operated on a time scalefrom seconds to hours. A consultant recognized that thecompany did not have knowledge of the variables at timescales below a second. Once it learned how to measureevents in the millisecond range, a large new subtree ofvariables became visible. By learning about these vari-ables and their implications for the process, the processengineers were able to reduce quality problems by a fac-tor of three within a few months. Development and ex-ploitation of the new variables continues today.

ReferencesMy thanks toJim Cook, Therese Flaherty, and two reviewers for especially helpfol comments on earlier drafts of his article; to Steve Furbush andLiz Bohn for research assistance; and to hundreds ofmanagers andHaroard Business School students for allowing me to test these ideas onthem. Remaining errors of mission and commission are my responsibility.

72 BOHN

1. I. Nonaka, "The Knowledge-Creating Company," Haroard BusinessReview, November-December 1991, pp. 96-104.2. Peter Drucker has commented, "In fact, knowledge is the onlymeaningful resource today. The traditional 'factors of production'have not disappeared, but they have become secondary." See:P.F. Drucker, Post-Capitalist Society (New York: Harper Business,1993), p. 42.3. Harlan Cleveland distinguishes dara, information, knowledge, andwisdom. However, he then intermixes the four concepts. See:H. Cleveland, "The Knowledge Dynamic," The Knowledge Executive(New York: Human Valley Books, 1985).4. R. Glazer, "Marketing in an Information-Intensive Environment:Strategic Implications of Knowledge as an Asset,"Journal ofMarketing55 (1991): 1-19.5. R. Jaikumar, "From Filing and Fitting to Flexible Manufacturing:A Study in the Evolution of Process Control" (Boston: HarvardBusiness School, working paper, 1988); an dA.S. Mukherjee, "The Effective Management of OrganizationalLearning and Process Control" (Boston: Harvard Business School,doctoral dissenation, 1992).6. R.E. Bohn and R. Jaikumar, "The Structure of TechnologicalKnowledge in Manufacturing" (Boston: Harvard Business School,working paper 93-035, 1992); andR.E. Bohn and R. Jaikumar, "The Development ofIntelligent Systemsfor Industrial Use: An Empirical Investigation," in Research on TechnologicalInnovation, Management and Policy, ed. R.S. Rosenbloom(London and Greenwich, Connecticut: JAI Press, 1986), pp. 213-262.7. This formalism is pursued in Bohn and Jaikumar (1992).8. af/mq in a local region.9. Glazer (1991); andN.R. Kleinfield, "Targeting the Grocery Shopper," New York Times,26 May 1991.10. J.A. Seeger, "Reversing the Images of BCG's Growth/Share Matrix,"Strategic Management Journal 5 (1984): 93-97.11. J. Dutton and A. Thomas, "Treating Progress Functions as aManagerial Opportuniry ," Academy ofManagement Review 9 (1984):235-247.12. P.S. Adler and KB. Clark, "Behind the Learning Curve: A Sketchof the Learning Process," Management Science 37 (1991): 267-281.13. R. Jaikurnar and R.E. Bohn, "A Dynamic Approach to OperationsManagement: An Alternative to Static Optimization," InternationalJournal ofProduction Economics 27 (1992): 265-282.14. J.M Juran and F.M. Gryna, eds., Juran's QJuzlity Control Hand-book (New York: McGraw-Hill, 1988), Chapter 22.15. These methods include Pareto chans, use of analogies to similarbut better understood processes, screening experiments, and othermethods discussed in the qualiry controlliterarure. Notice that screen-ing experiments are possible only if the variable is already at stage fouror higher.16. G.V. Shirley and R. Jaikumar, "Turing Machines and GutenbergTechnologies: The Post-Industrial Marriage,"ASME ManufacturingReview 1 (1988): 36-43.17. J. Flanagan, "GM Saga a Lesson for America," Los Angeles Times,27 October 1992, p. AI.18. Bohn and Jaikumar (1992).19. KE. Weick, "Organizational Culture as a Source of High Relia-bility," California Management Review, Winter 1987, pp. 112-127.20. Experienced bakers will realize that the following account is highlysimplified. A case simulation of some of the following issues is provid-ed in:



13/13

R.E. Bohn, "Kristen's Cookie Company (B)" (Boston: Harvard Busi-ness School, Case 9-686-015, 1986).21. For example, eggs, flour, and chocolate are relatively complex agri-cultural products, of imperfect consistency over time.22. P. Waldman, "Change ofpace: New RJR ChiefFaces a DauntingChallenge at Debt-Heavy Firm," WaH StreetJ()UmaJ, 14 March 1989.23. J.P. Walsh and G.R. Ungson, "Organizational Memory,"ActtdemyofManagement Review 16 (1991): 57-91.24. Bohn and Jaikumar (1992).25. R. Jaikumar, "Postindustrial Manufacturing,"Harvard BusinessReview, November-December 1986, pp. 69-76.26. W.B. Chew, D. Leonard-Barton, and R.E. Bohn, "BeatingMurphy's Law,"Sloan Management Review, Spring 1991, pp. 5-16.27. Learning is obviously of central importance in knowledge-basedcompetition, but detailed analysis is beyond the scope of this paper. Avery interesting s tudy of how machine developers become aware ofnew variables (stage two) through field use is provided by:


E. von Hippel and M. Tyre, "How Learning by Doing Is Done:Problem Identification in Novel Process Equipment," Research Policy,forthcoming.For a description of how one company manages learning as an integralpart of the manufacturing process, see:D. Leonard-Barton, "The Factory as a Learning Laboratory," SloanManagement Review, Fall 1992, pp. 23-38.For a discussion of the characteristics of organizations that learn suc-cessfully, see:D.A. Garvin, "Building a Learning Organization," Harvard BusinessReview, July-August 1993, pp. 78-9l.For a general typology of methods of technological learning, see:R.E. Bohn, "Learning by Experimentation in Manufacturing" (Boston:Harvard Business School, working paper 88-001, 1987).

Reprint 3615

BOHN 73

bohn measuring tech knowledge ocr version

Documents