0019-8501/00/$–see front matterPII S0019-8501(98)00056-X

Industrial Marketing Management

29

, 147–155 (2000)© 2000 Elsevier Science Inc. All rights reserved.655 Avenue of the Americas, New York, NY 10010

Intuition Can Helpin Segmenting

Industrial Markets

Paul Millier

After pointing out some differences between theory andpractice in segmenting industrial market, this article reviewsthe literature on the subject. The latter reveals that too mecha-nistic or too quantitative segmentation methods are quite inap-propriate for the industrial environment where the markets areconcentrated and the data sometimes rare. Based on this ob-servation, the article, finally, looks at the problem from an-other angle by mixing formally a little intuition with the analy-sis. The end of the article is illustrated by an application of themethod. © 2000 Elsevier Science Inc. All rights reserved.

A GAP BETWEEN THEORY AND PRACTISE

Among all the management tools, the segmentation ofindustrial markets presents a specificity; there is a stronggap between theory and practice. We find on the onehand literature that presents nice theories. On the otherhand, we find industrial companies who are miles away

from putting in practice these linear and well run-inmethods. It is, indeed, very common to see approximatesegmentation based on the structure of the actual salesdepartment or on the structure of an existing database. Inother words, people make it as easy as possible even ifthey spoil the job. Yet, R. Corey [1] pretends that onlygood segmentation leads to good strategy. Shapiro andBonoma [2] observed that the segmentation—when it isdone—does not play the important role that it should.They claim that most of the time, marketing managersuse segmentation as a means to explain the results morethan as a way to elaborate planning.

In fact, we feel that the subject is neither correctlytreated by literature nor correctly mastered by industrialmanagers. As Saporta [3] said, “When you look at themethods used by industrial companies to segment theirmarket, you feel there is a paradox. On the one hand, themore we study the conditions of work in the companies,the more we are convinced that segmentation should bethe starting point of the marketing strategy. At the sametime we must admit that segmentation is rarely used, and

Address correspondence to Paul Millier, EM LYON, Marches et Innovation,23 Av Guy de Collongue, 69132 Ecully Cedex France.

148

sometimes badly understood by most companies. It isclear that many improvements can be made in this field,that could undoubtedly help companies improve theirperformance.”

Saporta said so for commercialized products. But theobservation of more than a hundred radical innovations re-inforces Saporta’s analysis. The firms do not know how todeal with segmentation when the market does not yet exist.The universe does not exist, and the possible criteria are somany that combining them leads to many more segmentsthan a man can manage. In these conditions, systematicand rigorous methods seem to be a little apart from the in-dustrial reality. We are bound to take into account moreempirical data, more experience and even a little intuitionto deal with the segmentation of industrial markets.

To mix these so different points of view, this article pro-poses first to examine the segmentation literature to showits limits. We shall then present an alternative methodstarting from the actual industrial practices to avoid therisk of rejection. The aim of this method is not too ambi-tious. We do not want industrial managers to become “pre-mium segmenters”. We just want to help them improvetheir empirical practices.

SHORT REVIEW OF THE LITERATURE ABOUT INDUSTRIAL MARKET SEGMENTATION

When we look at papers about segmentation, we findsome common characteristics. The authors start by giv-ing a list of criteria that can possibly explain market seg-

mentation. Then, they give an order to use these criteria.For example, Shapiro and Bonoma [4] with their famous“nested approach” suggested starting with easier criteriaand finishing with the most difficult to get. They classifyby level of difficulty the criteria to use successively.

• Environment criteria (activity, size, geographical loca-tion).

• Exploitation criteria (client technology, type of prod-ucts bought, experience).

• Purchasing criteria (purchasing organization, purchas-ing strategy).

• Circumstance criteria (order urgency, quantity).• Personal features of the decision maker.

Nevertheless, Saporta said again that this approach islimited as underlined by Webster [5]: “The authors proposeno precise rule to decide when we must stop looking forrelevant variables. Some criteria (circumstance) are some-times linked to some others (environment). And moreover,a major problem lies behind the choice of circumstancevariables; can we consider for example that some specificcircumstances (order urgency, for example) are relevantsegmentation criteria?”

Beyond the very frequent “nested approach” of Shapiroand Bonoma, I shall also mention another dominant modelin industrial marketing, which mixes macro-segmentationand micro-segmentation. Introduced by Frank, Massy,and Wind [6], it was frequently used in literature such asby Wind and Cardozo [7] and by Dorey and Valla [8, 9].This two level method consists in shaping first macro-seg-ments on the basis of descriptive criteria (size, activity,geographical location, etc.). Then, we select one or sev-eral macro-segments according to their advantages. Last,we identify micro-segments inside these macro-segmentsby using the features of the buying centre.

Furthermore, beside this well-known literature there isa great number of models using one or several of the fol-lowing basis of segmentation:

• Geographical segmentation• Demographical segmentation• Psychographic segmentation• Behavioral segmentation• Segmentation by opportunity of purchase• Segmentation by circumstance of use• Segmentation by rate of use• Segmentation by rate of fidelity

As we can see, the list is long and one point can gener-ally be criticized. This point is that the authors never give

PAUL MILLIER is Professor at EM LYON (Lyon Graduate School of Business), where he teaches industrial marketing.He is also Head of Research at EM LYON and visiting professor at Cranfield School of Management, Cranfield University, United Kingdom. He has been conductingresearch on the marketing of technological innovation since 1982 in collaboration with most of the biggest French industrial firms (Saint-Gobain, Thomson, Elf Atochem, Total, and Schneider). He has conducted more than a hundred marketing studies in the mentioned companies and is adviser for a number of start-ups. Paul MILLIER has written three books:

Le marketing des produits high-tech. Outils d’analyse

, Editions d’Organisation, Paris, 1989 (also translated in Spanish.),

Développer les marchés industriels. Principes de segmentation

, Dunod, Paris, 1995,

Stratégie et marketing de l’innovation technologique

, Dunod, Paris, 1997. Also translated into English

Marketing the unknown

, J. Wiley & Sons, April,

1999.

149

any clear pieces of information about how to choose andcombine the criteria.

Among all the existing methods of segmentation, one ofthem seems to be closer to the clients problems. It is called“benefit segmentation” [10]. The principle of this methodconsists in grouping the customers whose needs for ourproduct are the same. The segments of market defined inthis way are very relevant to the marketing strategy.

In summary, in all this literature about segmentation inthe industrial environment, the authors discuss the essen-tial questions to be asked to prepare the segmentation,suggest as many criteria as possible but never properlysay how you do it. Thus, implementing these methods isstill a hard job.

A third type of literature about segmentation concernsmathematical tools. We can find a large sample of thesemathematical methods in Smadja’s book [11]. Thesemethods work properly when the problem is complicatedbut not complex (problems easy to handle, many data toprocess) and when a specialist is at our disposal. On thecontrary, they do not work very well in industrial market-ing where sense must emerge from complexity (the num-ber of answers is low but the dimensions of the problemare numerous and interlinked, the stake is very importantand the points of view opposite) and where data are rare(impossible to make the problem statistic).

In summary, the methods to segment consumer markets(with a very large number of customers) are well run inand are practical for those who know how to use them. Onthe contrary, all these mechanistic methods are limitedwhen the clients are few, when the markets are concen-trated, or when they do not yet exist. As a matter of fact,in this last situation, one single actor can change every-thing on the market due to its role or its weight. One sin-gle event can ruin instantaneously the most serious analy-sis. Furthermore, in industrial environments, we are in thefield of rare, uncertain, changing, and unreliable data,which do not fit with rigorous methods. Consequently, in-

dustrial companies are often in methodological difficultyto segment their markets.

In conclusion to this quick screening of the literature, weshould bear in mind that most methods have their disadvan-tages (too mechanistic or inappropriate for the quality ofdata), and, thus, it is very difficult to find the right one tosegment its market.

INTRODUCE A LITTLE OF INTUITION IN THIS WORLD OF RIGOUR

When we try to understand why industrial marketersseem to be frightened by methods of segmentation, they de-clare that they feel they are loosing contact with reality. Themethod acts more as a screen than as a help to understandthe market. So, how do they do that? As they have notmuch confidence in mathematical tools (too complex), asthey do not read a lot (not enough time), most of them usetheir experience, the “muddling through”, the “trial and er-ror” or, still worse, their intuition. But no manager, no“homo rationnalis” dare say that he/she uses his/her intu-ition for professional purpose because it looks like sorcery.

But why are intuition and reason so rarely used to-gether? It could be the fault of intuition, which is preju-diced by managers who do not dare say that they use it.And it is still worse in literature. No scholar would runthe risk of ruining his reputation by saying that intuitioncan be a marketing tool. If we want to look serious inmanagement, we need hard and logical methods, we needcriteria, marks, charts and so on (a.s.o.)

Equations, charts, and matrices will alas never makemanagement decisions perfectly reliable, certain, andcompletely rational. In the best case we shall be able torationalize them. But there will always remain this slightuncertainty due to the fact that we can never be sure weknow all the conditions of the experience. We can neverbe sure to have collected all pieces of information neces-sary to make the decision.

Well run-in methods exist for segmenting consumer goods markets. It is far from being

the same with industrial markets.

150

As doubt is creeping in, why shouldn’t we admit com-pletely that intuition can play an important role in pro-fessional decisions. Let us go a little further and let ussee if we could not use intuition in segmenting industrialmarkets.

As a matter of fact, if we start saying that industrialmanagers use their intuition to get a projection in whichthey recognize themselves and recognize their market,why shouldn’t we go further and help them in “profes-sionalising” their intuition to do still better.

For those of you who still have doubts about intuitionefficiency, we shall take a common example of using in-tuition. Ninety-five percent of all drivers driving their caron a small country road and seeing an old mustard yellowAustin Allegro, driven by an old Scottish farmer, smokingan unlit fag, do the same. They overtake, they slow down,they go away from him. They do so because their experi-ence tells them that this chap is dangerous and it is betterto be far away from him if they want to avoid a catastro-phe. You need neither statistical series nor a representa-tive sample. Just two or three experiences are sufficient torecognize instinctively the situation. Intuition is, in thiscase an excellent way to make a quick synthesis of a greatnumber of data collected through experiences. We get avision of the world through our experience as say the con-structivists.

There is just a small step to cross to assume that any-body is able to use his/her intuition for professional aims.The only condition is that this intuition is based on strongfield experience. In this sense, intuition is much more de-fined as an empirical tool than the one used by a fortuneteller who feels the hidden things.

Nevertheless, a problem still remains with intuition; itis difficult to convince other people just on the basis ofour intuition. Even if we are sure of our intuition, it is dif-ficult to convince our boss, our banker, or our share-holder. They may have their own intuition, and, as theyare boss, banker, or shareholder, their intuition will bebetter than ours. We must then jump another step and ra-tionalize, justify, rebuild the intuitive reasoning until it

resists any criticism from our boss, banker, or share-holder.

AN EXAMPLE TO SHOW HOW TO USE INTUITION AND RATIONALIZATIONIN SEGMENTATION

We shall examine in this example how to segment thefuture market of a technological innovation. In otherwords, the market does not exist yet, and we typically can-not use historical or statistical data to segment the marketfor this data does not exist.

We shall first make an assumption based on the obser-vation of many cases of industrial purchases; industrialclients never buy a product if they do not need to solve atechnical problem. Taking this into account, we shall firstsegment the market from a technical point of view. Weshall call “application” a technical problem that the cli-ents want to solve.

But experience shows that it is not sufficient that aproblem exists to say that there is a market. We mustanalyse and understand the reasons why a customer is in-terested in this specific innovation. For this reason, weshall also segment the market from a behavioral point ofview. Once we have these two segmentations (technicaland behavioral), we cross them to get the segmentationmatrix that represents the virtual future market of innova-tion (see Figure 1)

In this paper, we shall only explain how to get thetechnical segmentation (i.e., the applications), but the in-tuition/rationalization principle is also available for thebehavioral segmentation. To illustrate, we shall take theexample of a new aluminum alloy that a company wantsto sell in the automotive industry.

The segmentation process that we shall now presentruns four steps:

• Uses check-off• Intuitive grouping of individual uses• First rationalization by two logical tests

Intuition is a dramatically efficient way to

begin the segmentation process.

151

• Second rationalization by identification and combina-tion of criteria

List of Uses

To make the technical segmentation, we first need tocollect on the market all the ideas of use suggested by theprospects. For example, automotive companies could sug-gest using this new aluminium alloy for making the fol-lowing products: piston, intervalve bridge, crankshaft,connecting rod, piston axle, rocker, temperature gauge,gear box forks, brake caliper, bumper bracket.

Intuitive Grouping of Uses into Applications

Intuitively, we group together, all uses (i.e., piston,con-rod, etc.) showing the same kind of technical problemto solve, and we call this group an “application”. In thepresent case, we can for example build three groups ofuses, that is, three applications (see Figure 2).

First Rationalization with Two Logical Tests

Once the intuitive grouping has been made, we use twological tests to rationalize the job. These two tests are:

• The Functions/Applications matrix,• The Competing Technologies/Applications matrix.

How can these two matrices, be considered as logicaltests? In functional analysis, we consider that the functionis the thing bought by the customer to solve his problem.But, if two technical problems are different they must ad-mit two different solutions. In other words, we cannot findmore than one combination of functions for each applica-tion (defined as technical problem solving). Furthermore,all the uses grouped inside an application must require thesame functions to ensure that the application is a homoge-neous group of uses.

If that was not the case, we would have to group to-gether the applications requiring the same combination of

Consolidate your intuition with some simple

logical tools.

FIGURE 1. Segmentation matrix.

FIGURE 2. Grouping uses into applications.

152

functions and split heterogeneous application. We can il-lustrate this second case with the aluminium alloy exam-ple. The function/application matrix built around theoriginal intuitive grouping of uses is shown in Figure 3.

Now, a more precise analysis of the third column re-vealed that there is no need to reduce friction in inter-valve bridge and temperature gauge. Thus, the matrix hadto be cut as shown in Figure 4.

Then we have four tougher applications because they arewell justified by the list of functions required in each case.

On the contrary, we can group two applications thathave been abusively split. Then, we put together applica-tions in which the same functions are needed. As a matterof fact, as the sum of functions required is no more thanthe “list of requirements” of the product; it is, in fact, thesolution to the client’s problem. Now, if two problems(i.e., two applications) admit the same solution, we cansuppose that they are so similar that we can group them.

The rationalization of the segmentation can now stillbe reinforced by the competing technologies/applications

Construct the segmentation tree.

FIGURE 3. Function/application matrix used as a first logical test.

FIGURE 4. Improved technical segmentation.

153

matrix. This matrix plays exactly the same role as thefunctions/applications matrix for the technologies incompetition with ours are analytical solutions (i.e., thecontents of the offer) to the client’s problem.

Searching and Combining the Criteria of the Technical Segmentation

Once the applications have been consolidated by theprevious tools, the next step consists in describing themwith as many details as possible.

For example, in the case of a nondestructive control de-vice, we shall describe the materials to be analyzed, thefeatures of the device, the competing technologies, the re-quired functions, the required performance, the type ofanalysis to be done.

Then, we try to point out the similarities and the differ-ences between all these characteristics. These similaritiesand differences become no more than the modalities of thesegmentation criteria. For example, some applications canbe defined by the fact that composite materials are analyzedwhile metal is analyzed in the others. “Composite materi-als” and “metal” may be considered as the two modalitiesof the same criterion called “material to be analyzed.”

If we do the same systematically to all characteristics, wecan progressively draw up a list of segmentation criteria.

Finally, different combinations of these modalities arescreened, and only the combination that leads to the con-solidated applications remains. In the case of the nonde-structive control device, we could for example keep thesegmentation tree shown in Figure 5.

Once this job is accomplished, a similar operation isdone on the clients’ behaviour, and both segmentationsare crossed to produce the final matrix of segmentation.The complete segmentation of the non destructive controldevice could look as shown in Figure 6.

At the end, the segments (the nonempty boxes of thematrix) just have to be named by selecting the commondescriptive features of the clients inside the segment. Tofinish the job, the segments are quantified and the sum ofall quantities gives approximately the market size.

CONCLUSION

In conclusion, a first draft of segmentation can be veryquickly got by intuition. This draft is far from being per-fect, but it is not a problem as long as we know that it is

FIGURE 5. Building the segmentation tree.

Well-built segmentation is very convincing.

154

just a starting point. The aim of the operation is just to getsome first classification poles to begin the rationalizationprocess. In the case of the above segmentation, the logicaltest matrixes and the research of the segmentation criteriaplayed this role of rationalization [12]. If we look care-fully at what we have done, we have made no more thanputting in practice some principles used in mathematicalmethods such as the “heap method” or the “dynamicclouds method”. In these two methods, some groupingpoles are first chosen (randomly or not), and some mathe-matical tests are done to maximize intraclass homogeneityand interclass heterogeneity. The first unfinished classifi-cation is progressively improved until we reach an ap-proached result that can always be endlessly modified.The segmentation is adjusted to the market step by step, itis compared to the market, and it is validated by this com-parison at the same time as it validates the market. Inthese conditions, the segmentation can be considered asan “ideal type” in the meaning of Max Weber as quotedby Bourdieu, Chamboredom and Passeron [13]. He saysthat the ideal type is a guide to build hypothesis, a coher-ent fiction to which the situation or the action is com-

pared. It is a construction that allows a comparison to thereality. The reality can be measured by the ideal type be-cause the ideal type is measured against the reality as ad-justments of both the ideal type, and the reality lead to anever greater similarity between the two.

But, even partly invalid the segmentation is useful be-cause, at least, it exists. Now, it is always much easier tocriticize and improve something than to build for the firsttime. In fact the reason of being of the segmentation is tobuild a first and original representation of the market. So,the segmentation becomes a heuristic tool, a tool to dis-cover the reality. It rationalizes the market and remindsus that “the reality is not rational but only rationalizable”.

The practice of this way of segmenting the marketproves that rationalization gives a strong power of convic-tion to the segmentation because it relies on a construc-tion, more logical more justified and more communicablethan the only intuition. But this benefit is small comparedto what the exercise brings to the one who does the joband appropriates progressively the representation that heproduces. The process of segmentation forces us to write,to project on the paper the ideas turning endlessly in our

FIGURE 6. Final segmentation matrix.

155

mind. Now, as long as the ideas turn in mind it is impossi-ble to know if they are valid, complete, and justified. Assoon as we draw the ideas on the paper we get a jigsawpuzzle effect. We just begin to see the image when westart putting the parts together. Furthermore, the only wayto know if we have all the parts, or if some are missing, isto build the puzzle. In this latter case, we have to searchfor the missing pieces of information, but the job is veryfocused on what is missing, hence very efficient.

REFERENCES

1. Corey, R.:

Industrial Marketing Strategy: Cases and Concepts

. Prentice-Hall, Englewood Cliffs, NJ, 1975.

2. Shapiro, B., and Bonoma, T.: La Segmentation des Marchés Industriels.

Harvard l’Expansion

(Automne),

37–45 (1984).

3. Saporta, B.: La segmentation en marketing industriel.

Recherche et appli-cations en marketing

4,

39–52 (1987).

4. Shapiro, B., and Bonoma, T.:

Segmenting the Industrial Market

, Lexing-ton Books, Lexington, MA, 1987.

5. Webster, F.:

Industrial Marketing Strategy

. J. Wiley & Sons, London,1974.

6. Frank, R., Massy, W., and Wind, Y.:

Market Segmentation

. Prentice Hall,Englewood Cliffs, NJ, 1972.

7. Wind, Y., and Cardozo, R.: Industrial Market Segmentation.

IndustrialMarketing Management

7,

153–166 (1974).

8. Dorey, F., and Valla, J. P.: Pour un système de segmentation adapté àl’entreprise industrielle,

Acte du 11e Séminaire International de Recher-che en Marketing

, IAE, Aix-en-Provence, 13–14 Janvier 1984.

9. Dorey, F., and Valla, J. P.: La Segmentation en Marketing Industriel, ClubMarketing Industriel RA, IRE, Mars, Lyon, France, 1986.

10. Haley, R. I.: Benefit Segmentation: A Decision Oriented Research Tool.

Journal of Marketing

32 (July),

30–35 (1968).

11. Smadja, Alain: Segmenter Ses Marchés. Applications Pratique des Tech-niques de Segmentation dans le Marketing. Presses polytechniqueromandes, Paris, 1988.

12. Millier, Paul:

Développer les Marchés Industriels. Principes de Segmenta-tion

, DUNOD, Paris, 1995.

13. Bourdieu, P., Chamboredon, J. C., and Passeron, J. C.:

Le métier de socio-logue

, Ed Mouton, Paris, 1983.