Section - 3.2Segmenting in Mature Markets

V. Kasturi Rangan, Rowland T. Moriarty, & Gordon S. Swartz

Segmenting Customers in MatureIndustrial Markets

In mature industrial markets, segmenting customers on size, industry, or product benefits alone is rarelysufficient. Customer behavior in terms of tradeoffs between price and service is an important additionalcriterion. The authors offer a framework for such buying-behavior-oriented microsegmentation of indus-trial customers. They apply the framework to segment the national accounts of a targe industrial companyand show how the results of a segmentation study can be used to redirect the firm's resources andcustomer segments.

THE goal of segmentation is to identify distinctcustomer groups that have homogeneous needs

(Wind 1978). Tailoring the marketing mix for partic-ular segments leads to better planning and more ef-fective use of marketing resources (Kotler 1988; Ma-hajan and Jain 1978). Coles and Culley (1986), forexample, illustrate how DuPont segmented its marketfor Kevlar, an aramid fiber that is lighter yet strongerthan steel. The company focused the unique needs ofcustomers in three different segments.

• Potential fishing boat owners: Kevlar's lightness prom-ised fuel savings, increased speed, and the ability to carrymore fish weight.

• Aircraft designers: Kevlar has a high strength-to-weightratio.

• Industrial plant managers: Kevlar could replace the as-bestos used for packing pumps.

V. Kasturi Rangan is Professor, Harvard Business School. Rowland T.Moriarty is Chairman of Cubex Corporation, Boston, MA. Gordon S,Swartz is a doctoral candidate, Harvard Business School. The authorsthank Russ Flaum of Signode Corporation for his assistance and co-operation in conducting this study, Carol Smith for providing computerprogramming and statistical assistance, and Tom Bonoma, Erich Joach-imsthaler, Tom Kinnear, Ben Shapiro, Fred Webster, and three anony-mous JM reviewers for their comments and suggestions on previousversions of the article.

Though such market segmentation designs basedon product benefits are widely recognized as the stateof the art and superior to traditional segmentationschemes based on industry type or customer size (Car-dozo 1980; Moriarty and Reibstein 1986), sustaininga segnientation strategy based on benefits alone is oftendifficult as the product market matures. Eventually,competitors are able to offer equivalent products andmany buyers may therefore be unwilling to pay a pricepremium. This situation is especially common for in-dustrial raw materials and supplies that are difficult todifferentiate by functions and features alone. Steadilyand deliberately, as the product market turns a com-modity (Kotler 1988; Shapiro ct al. 1987), price andservice become important buying criteria for somecustomers. By further segmentation of each benefitsegment, the heterogeneity in a macrosegment be-comes apparent.

We develop a buy ing-behavior-based frameworksuitable for microsegmenting customers in mature in-dustrial markets. Though the concept of buy ing-be-havior-based segmentation has been recognized for twodecades (Webster and Wind 1972), few applicationsof the approach have been reported in the industrialmarketing literature. We describe how our buying be-havior framework was applied to segment further the

72 / Journal of Marketing, October 1992Journal of MarketingVol. 56 (October 1992), 72-82

national account customers of a large industrial com-pany. In addition, we demonstrate how segmentationanalysis can be used proactively to influence cus-tomers' movements to segments that are mutuallybeneficial to the seller and buyer. In contrast, pre-vious application worit (de Kluyver and Whitlark 1986;Doyle and Saunders 1985; Moriarty and Reibstein 1986)attempted to uncover existing segments as a way toposition products strategically. In our approach the as-sumption is that at a microsegment level, firms caninfluence and shape the buying behaviors of their po-tential customers by tactically altering marketing mixvariables such as price and service.

' The Conceptual FrameworkResearch on industrial market segmentation (see Cheronand Kleinschmidte 1985 and Plank 1985 for a com-prehensive review) offers several bases for segment-ing customers (Frank, Massy, and Wind 1972), in-cluding:

• demographic descriptors such as geography, standard in-dustrial classification code, and account size (HIavacekand Ames 1986).

• pnxiuct end-use or application (Wind and Cardozo 1974),• buying situation (Robinson, Faris, and Wind 1967),• customer benefits (Choffray and Lilien 1978; Haley 1968),• customer buying behavior (Bonoma, Zaltman, and John-

ston 1977; Webster and Wind 1972), and• customer decision-making style (Wilson 1971).

It has been suggested that firms would benefit mostby the successive application of two or more such seg-mentation schemes in a nested fashion (Bonoma andShapiro 1983), similar to the microsegmentation prin-ciple advocated by Webster (1984). Interestingly,however, none of the segmentation schemes cited herecapture the underlying dynamics of a maturing mar-ket.

Product life cycle (PLC) theory contends that pricestend to drop as the product market matures (Curry andRiesz 1988; Day 1981; Simon 1979; Younger 1986).Two underlying forces cause that trend. One is cus-tomer leaming during the PLC. As the product ma-tures, many customers who have become totally fa-miliar with the product's characteristics, functions, andfeatures no longer require the same intensity of prod-uct information that was once provided by its supplier(Day 1986; Schmalensee 1982). As a result, they areunwilling to pay for the cost of such services. Thesecond force is the result of competitive action, whichin a mature market makes equivalent products avail-able to customers at similar or lower prices (Day 1986).

Because of this market dynamic, customers in ma-ture markets can be aligned along the two dimensionsof price and cost-to-serve (see Figure 1; Forbis andMehta 1981; Shapiro et al. 1987). The reason is that

FIGURE 1Potential Buying Behavior Segments

HIGH

PRICE

LOW

PowerAxis

LOW HIGHCost-lo-Serve

customers who demand a tow price will be offered a"no frills" product accompanied by minimal service,and customers who value an augmented product willpay a higher price (Levitt 1980) and receive the fullcomplement of services (Porter 1980). Price differ-entials due to prcxluct quality differences are small be-cause cotnpetitors are able to offer more or less equiv-alent products. Hence, any major price variations aredue to differences among services provided. Cus-tomers who receive the "core" product pay less be-cause it costs less to serve them than to serve thosewho demand and value the full service.

The preceding reasoning is consistent with eco-nomic theory. Dortman and Steiner (1954) showed thatwhen higher service implies higher direct costs, a firmcould maximize its profits by charging a price thatequates its marginal revenues to marginal costs. Tellisand Wemerfelt (1987) showed empirically how the priceversus service-quality reiationship would hold in mar-kets characterized by a high level of product infor-mation availability, which is certainly the case in ma-ture markets.

In keeping with this rationale, firms operating inmature environments expect to align their customersalong the equity axis in Figure 1. The lower left quad-rant (C) represents a core, no-frills product withoutmuch service and the upper right quadrant (B) rep-resents an augmented product accompanied by inten-sive value-added services. In both cases, the price-service offering is equitable to the seller and the buyer.The "core product" customer pays a lower price andthe "value added" customer pays a higher price. Thisrationale, however, is based on the seller's expecta-tions of how customers would behave in a maturemarket.

An altemative is given by the power axis in Figure

Segmenting Customers in Mature Industrial Markets / 7 3

1. Customers see only the price dimension of the ma-trix in Figure 1. They do not know the seller's cost-to-serve. Because of market maturity, however, sev-eral sellers usually can offer similar products, and manycustomers therefore may attempt to shop for price.Guaranteed purchase volumes and large order sizesare typically offered as the bait. As a result, cus-tomers may not necessarily align themselves along theequity axis as sellers exjrect, but prefer to operate inquadrants C and D in Figure I, depending on theirknowledge of competitive offerings and their ownmarket power (Scherer 1980). In the health care in-dustry, for example, multihospital buying groups havesuccessfully leveraged their power to seek priceconcessions from their suppliers (In Vivo 1985).

In short, all locations above the equity axis indi-cate that the seller is able to extract more than the fullvalue of the services it renders, because customersperceive the firm's product offering as superior tocompetitive offerings or substitutes. Positions belowthe equity axis indicate that the firm is unable to ex-tract the full value of the services it renders with itsproduct. Along the equity axis itself, the exchange isfair and equitable. We now describe how this frame-work can be used to analyze a company's customersegments in mature industrial markets.

DatabaseThe main purpose of our study was to validate thebuying behavior framework depicted in Figure 1 anddemonstrate its use for managerial action. We there-fore sought a research site where the product-marketenvironment was in the mature phase of its PLC,characterized by price pressures and the availabilityof equivalent competitive products. This criterion ruledout capital goods because suppliers in such marketshave demonstrated their ability to maintain prcxluctdifferentiation, through functions and features, in spiteof market maturity. Our framework is more appro-priate for mature industrial raw materials and sup-plies, for which service rather than product charac-teristics is the basis for competition. We thereforesought and obtained the cooperation of the managersof the packaging division of Signode Corporation forconducting our research. This division produces andmarkets a line of steel strappings used for packaginga diverse range of goods such as brick, steel, cotton,and many manufactured items.

Signode, the market leader since 1948, had lost10 share points because of stiff price competition inthe six yeare before our study. Its managers consid-ered Signode to be the high service supplier of steelstrapping in their market. Signode was the only com-pany to provide parts and service for repair of pack-aging equipment at the user firms. The packaging di-

vision also offered engineering advice on packagingneeds. All other competitors provided only the steelstrapping.

The company segmented its customers by size—small, medium, large, and national accounts—andwithin each of these segments by SIC code. Thoughthe company did not use a buyer-behavior-based seg-mentation scheme, its managers believed that the firm'smarketing policies were structured so that low priceseekers could have a "commodity" product and cus-tomers who sought services could have a "value added"product at a higher price (conforming to the behaviorunderlying the equity axis in Figure 1). The managerssuspected that such buying behavior variations werepresent in each size-based segment.

Because our study required an in-depth analysis ofindividual customer buying behavior, we focused onthe company's 174 national accounts whose individ-ual purchases of Signode's products exceeded $100,(XX)yearly. Collectively, these accounts were nearly 40%of Signode's sales revenues. In general, one wouldexpect to find minimal variation in such a narrow ma-crosegment and if found it would only validate ourconceptual framework. Signode's other segments (large,medium, and small) were much larger, ranging from2000 to 20,000 customers. The i74-company samplewas compact enough to enable us to gather data at theindividual-customer level, but large enough to ensurestatistical conclusion validity (Cook and Campbell1979).

Extensive surveys of multiple members of com-plex decision-making units (DMU) have been foundto be impractical and arduous (Johnson and Flodham-mer 1980; Moriarty and Spekman 1984). The proce-dure is time-consuming, costly, and may influence thevery behavior one is attempting to observe. Further-more, once surveyed, the DMU may not converge inits opinions (Phillips 1981; Silk and Kalwani 1982).We therefore followed the methodology adopted byAnderson, Chu, and Weitz (1987) and Flodhammer(1988), and used the company's five national accountmanagers (NAMs) and 20 dedicated national accountsales representatives as key informants of customerbuying behavior. Because the NAMs and account repsinteract with their customers frequently and becausetheir interactions cover many transactions over an ex-tended time period, their perceptions of buyer behav-ior are likely to be accurate, especially when they aremaking comparisons among customers (Keman andSommers 1966).

To operationalize our framework from Figure I,we constructed 12 variables to capture the potentialbuying behavior variations in Signode's national ac-counts (Platzer 1984; Tutton 1987). Six of them werechosen to reflect the price versus service variationsalong the equity axis and the other six were chosen

74 / Journal of Marketing, October 1992

to reflect the buyer power variations along the poweraxis. Though this list of operational indicators maynot be exhaustive. It does reflect the buying behaviordimensions that Signode's NAMs and sales reps be-lieved to have the most influence on customer pur-chasing behavior. We used two sources for gatheringdata: in-house documents and responses from sales repsand NAMs. Table 1 summarizes the key features ofthe 12 variables measured for 161 of Signode's 174national accounts (93% coverage). The data collectiontask could not be completed in time for the rest of the13 accounts. Though some of the data presented herehave been disguised at the company's request, thesegmentation approach and its implications are ac-curate.

Buying Behavior Variables From In-HouseDocuments

Given the mature stage of the market and the stan-dardization of product features across suppliers, priceversus service tradeoffs were common (Ross 1984;Shapiro et al. 1987). However, because our focus waswithin rather than across a market segment, we mea-sured price and cost-to-serve in relation to Signode'sother national accounts. Called "relative price" and"relative service," these two measures correspond tothe price and cost-to-serve dimensions, respectively.Hence, the first four variables we measured were rel-ative price, relative service, account size, and marketshare.

1. Relative price is a measure of the higher or lower pricethat an account paid in relation to Signode's other na-tional account,s. Almost all national accounts receiveddiscounts from standard carload prices. Using ac-counting data on all transactions completed in the mostrecent 12 months, we computed a volume-weightedaverage discount for each national account. Averagediscounts ranged from 0% to as much as 11.3%.

2. Relative service is the higher or lower level of servicethat an account received in relation to other nationalaccounts. Signode's managers identified three impor-tant components of service; (1) field sales calls, (2)unbilled parts, tools, and repair work, and (3) appli-cation and engineering services. To aggregate thesethree components, we first computed quantitativemeasures for each component using their natural units.Thus, field sales calls were measured as calls per year,unbilled work was estimated in dollars, and applica-tions engineering was measured in hours. These in-dividual components were converted linearly to a 1-to-10 scale and the three rescaled measures then wereaveraged to yield a composite score. To verify thejustification for our summation (Churchill 1983, p.255; John and Weitz 1988), we ran confirmatory fac-tor analysis models. The three-factor oblique modelfit best with the data (x' = 8.09. p = .620, adjustedgoodness of fit = .969, root mean residual = .013).The second-order model that assumed three specificfirst-order factors and one general second-order factoralso fit the data well (x" = 8.59, p = .378, adjustedgoodness of fit = .972, root mean residual = .011).In either case, therefore, summation of the serviceconstructs is a justifiable measurement strategy, Allother altemative models provided a poorer fit.

3. Account size. Various aspects of account size havebeen identified as influencing buying behavior. It has

TABLE 1National Account Database

Buying Behavior VariablesIndicators

Of Source Units Mean

In-House Documents1. Relative price

2. Relative service

3. Account size (annual purchases)]4. Market share J

Salesforce Judgmental Data(Sales Elasticity)

5. For decrease rn price "̂6. For increase in price 1 - ,7. For decrease in service I ,.8. For increase in service )

9. Product importance"]10. Switching potential [11. Market knowledge [12. DMP complexity J

Price andservicetradeoffs

Buyer power

Price andservicetradeoffs

Buyer power

Sales records/NAMjudgment

Sales records/NAMjudgment

Sales recordsSales records

1

NAM/sales repjudgment




Sales rep judgmentSales rep judgmentSales rep judgmentSales rep judgment

Discount %

1-10 scale

Dollars%

' 1 . .

1 1

%

%

1-5 scale1-5 scale1-5 scale1-5 scale

1

4.6

556,00063

7 *2ZA'M.19

2,94.14i3

Segmenting Customers in Mature Industrial Markets / 75

been argued, for example, that the larger the pro-curement scope (Cardozo 1968) and order quantity(Assaei and Ellis 1976; Bonoma and Shapiro 1983),the higher the level of buyer-seller interdependency(Corey 1978). Using notions of "power" (Emerson1962; French and Raven 1959), one could argue thata large volume buyer would be able to negotiate rel-atively lower prices and higher services than its smallercounterparts. By the same loken, one can argue thatif the buyer is overly dependent on one source for itslarge volume requirements, the seller could exertcountervailing power (Gaski 1984). Regardless of thedirection of influence, account size obviously affectsbuying behavior. We measured account size simplyas the total purchase volume of all Signode productsin the most recent 12 months.

4. Market share. A different measure of dependency isthe proportion of business a single supplier has in thebuyer's total purchases of a product category. This isa measure of vendor reputation (Levitt 1965; Sheth1973) and the associated purchasing strategy—howmany suppliers would be sought and how the bidswould be allocated among the bidders (Bonoma andShapiro 1983; Cardozo and Cagley 1971; Corey 1978).We used Signode's share of dollar sales volume foreach account. As part oftheir routine sales reporting,Signode's salespeople estimated the total dollar vol-ume of each account's steel-strapping purchases.Knowing Signode's actual sales to the account, wecomputed Signode's market share—an indicator of thebuyer's preference for, as well as reliance on, Sig-node's products.

Buying Behavior Variables From NationalAccount Reps and NAMsIn line with research by Anderson, Chu, and Weitz(1987), Corey (1978). Duncan (1966). HIavacek andAmes (1986), Moriarty and Reibstein (1986), Parket(1972), and Webster and Wind (1972), we consideredcustomer sensitivity to price and service changes tobe two important aspects of buying behavior. Thus,using the decision calculus methodology suggested byLittle (1970), we estimated customer demand elastic-ities with respect to price and service. Sales reps wereasked the following question for each of their ac-counts: "If you were able to drop (increase) prices by7%, what is your best estimate of the percentage ofincrease (decrease) in sales volume that would re-sult?" TTie 7% represented a level of price discountingthat the managers had selectively used in the past toretain certain large volume accounts in the face ofcompetitive activity. Service elasticity was measuredin the same way, except that the unit of change wasbroken down by its individual components for bettercomprehension. To add to our list of variables, thedecision calculus methodology revealed four customerdemand elasticities:

5. For decrease in price.6. For increase in price.7. For decrease in service.8. For increase in service.

To overcome the drawbacks of the decision cal-culus methodology (Chakravarti, Mitchell, atid Stae-lin 1979), the data collection method included an "an-choring" process to improve the standardization of theinformants' responses (Anderson 1974). Each salesrep first tentatively evaluated a typical account andreceived feedback from his or her national accountmanager (NAM). Next, the sales vice president andthe other four NAMs provided benchmarks from theirown experiences with other national accounts. Usingthese anchors as a guide, each salesperson made a fi-nal assessment for the typical account. With this eval-uation as the comparison point, the sales rep com-pleted the questionnaires for the other accounts. At nopoint did we ask sales reps or NAMs for estimatesoutside their normal operating ranges; we thus avoideda common criticism of such Judgmental data collec-tion methods.

In addition to these demand elasticities, four morebuying behavior indicators—product importance,switching potential, market knowledge, and decision-making process complexity—were estimated by thenational account sales reps through the data collectionprocess described above.

9. Product importance. The degree of risk (Cardozo 1980;Moriarty and Galper 1978; Sheth 1973) as well ascompatibility and complexity (Roger* 1983) of theproduct or its application have been identified as keydeterminants of organization buying behavior. Degreeof risk here pertains to the "line-stopping" potentialof the product. If the product's quality or its deliveryis unreliable. It could significantly influence the op-erations of the user firm. Compatibility and complex-ity refer to the significance of the product's fit withthe buyer's operations. Depending on the applicationand extent of usage, the importance of steel strappingvaried over Signode's national accounts. Customersthat perceived tbe product line to be critical werethought to devote more energy and consideration tothe buying process.

10. Switching potential. Though Robinson, Faris, and Wind(1967) suggested that altemative suppliers are notusually considered in straightforward rebuy situations(such as Signode steel strapping). Anderson, Chu, andWeitz (1987) provide evidence that incumbent sup-pliers may have to "prove themselves" again to "tie-win" the bid. We believe the latter observation ap-plies particularly for mature commodities becausesuppliers and their products generally are consideredundifferentiated by customers (Kotler 1988). Switch-ing potential (Gensch 1984) is also related to the no-tion of supplier reputation and previous performancehistory (Parket 1972; Sheth 1973). Over the years,several customers had built a trusting relationship withSignode because of the prtxluct or the service or both.These customers were expected to deviate less fromnormal purcha.sing pattems than other customers, whomight be more likely to switch at lower levels of dis-satisfaction.

i 1. Market knowledge. Regardless of the stage of marketmaturity, customers vary substantially in their knowi-


edge of competitive products and prices (Nagle 1987);they search for infonnation (Sheth 1973) in varyingdegrees. Bonoma and Shapiro (1983) pointed out thatsome customers actively .seek altemative informationas a matter of policy. Naturally, customers with a de-tailed knowledge of altemative suppliers' steel-strap-ping offerings were expected to use somewhat moreaggressive negotiation strategies (Corey 1978).

12. Decision-making process (DMP). The complexity ofthe buying decision-making process is a reflection notonly of product and vendor characteristics but also ofthe buying organization's priorities and purchasingstrategies (Johnston and Bonoma 1981). Thus, deci-sion-making style and structure (Duncan 1966; Mor-iarty and Bateson 1982; Moriarty and Galper 1978)are indicative of customer buying power. At Signode.national account reps suggested that customers withconsiderable leverage usually required .several salespresentations and often contract-by-contract negotia-tion before an agreement could be reached.

Analysis and ResultsTo identify buying behavior microsegments, we per-formed a hierarchical cluster analysis based on the 12variables of Table 1. Though there is no consensus onthe best clustering algorithm, we selected Ward'smethod of minimum variance to maximize homoge-neity within clusters and heterogeneity between clus-ters.' To determine the number of clusters, we ex-amined three statistics.' The cubic clustering criterionand the pseudo F-statistic both showed local peaks atthe four-cluster solution, whereas the pseudo t'̂ -sta-tistic revealed a large drop at the fourth cluster. Wealso used discriminant analysis to check the clustergroupings (Churchill 1983, p. 654; Green 1978). Thefour-group discriminant analysis indicated the pres-ence of two significant discriminant functions (eigen-values 12.0144 and .4616, respectively), which re-covered 92.8% of the classification accurately.

Mean values of the buying behavior variables arereported for each microsegment in Table 2 and thestandardized discriminant function coefficients are re-ported in Table 3.

Buying Behavior MicrosegmentsSegment I: programmed buyers. Customers in thismicrosegment were small and viewed the product asa routine purchase item. They had the lowest averagesales of any group and were not particularly price or

(1980). in examining 15 clustering algorithm.s, recom-mended Ward's meihod as the best available. In ihis method the dis-lance between two clusiers is the ANOVA sum of squares betweenthe two clu.sters, added over all the variables. At each iteration, thewithin-cluster sum oC squares is minimized over all partitions.

Mn comparisons of 30 methods for estimating the number of clusters(Cooper and Milligan 1984; Milligan and Cooper 1985), the cubic-clustering criterion (Sarle 1983), ihe pseudo F-statistic (Calinski andHarabasz 1974). and Ihe pseudo t'-statistic (Duda and Hart 1973) havebeen successful in idenlilying the appropriate number of underlyingclusters.

service sensitive. The product was not very importantor central to their operations. In comparison with thosein the other three microsegments, these customers hadthe lowest market share of Signode products.

We subsequently learned that many of the.se ac-counts used rules-of-thumb to allocate their pur-chases. They split orders among two or three vendorsin fixed proportions. Signode, because of its market-leader reputation, received a major share of these pur-chases—on average, about 54%. Perhaps because oftheir routinized procedures, these accounts investedlittle effort in the buying process, either in negotiatingpurchases or in investigating alternative sources. Inretum, Signode charged them the full list price andprovided below-average service. Because customersin this segment tended to allocate market share sys-tematically rather than evaluate the price-volumetradeoffs, we characterized the purchasing behavior ofthis microsegment as "programmed buying."

Segment 2: relationship buyers. Customers in thismicrosegment were also relatively small. The productitself was moderately important in their operations and,unlike the programmed buyers of segment 1, they weremore knowledgeable about competitive offerings.Though customers in this microsegment paid lowerprices and received more service than programmedbuyers, they also gave Signode a higher market share(67.8%).

Customers in this microsegment had a propensityto switch, but they were less prone to switching thantheir counterparts in the third and fourth segments. Inaddition, in comparison with their more aggressivecounterparts in the third and fourth segments, thesebuyers did not push Signode for price and serviceconcessions and they paid higher prices for relativelyless service. This difference in "value received" prob-ably explains their extreme sensitivity to price in-creases. On average, a 7% price increase in this mi-crosegment would decrease purchase volumes by asmuch as 28%. These customers seemed to prefer Sig-node's partnership to a mere price exchange. We la-beled the behavior of this segment as "relationshipbuying."

Segment 3: transaction buyers. Customers in thismicrosegment were, on average, twice as large as therelationship buyers. They received price discounts av-eraging about 10% and an above-average service level;they had the highest sensitivity to decreases in ser-vice. The product itself was very important to theiroperations. Customers in this group were very knowl-edgeable about competitive offerings and. though val-uing Signode's service offerings, they would not hes-itate to switch suppliers. Because these customersactively considered the price versus service tradeoffs.

Segmenting Customers in Mature industrial Markets / 77

BehavioralSurrogates

1. Relative price (%)2. Relative service [%)3. Account size (sales)4. Market share (%)5. Percentage increase in sales for price

drop6. Percentage decrease in sales for price

raise7. Percentage decrease in sales for ser-

vice drop8. Percentage increase in sates for ser-

vice raise9. Product importance {%)

10. Switching potential (%)n . Market knowledge (%)12. DMP complexity (%)

TABLE:IGroup Means'

Segment 1(54 accounts)

0.03.6

$122,00054.2

5.6

15.5

5.1

1.2

2.53.84.03.2


-7.94.9

$472,00067.8

8.9

27.9

9.2

3.0

3.04.44.53.6


-10.1&6

$1,100,00071.98.7

24.5

12.5

5.2

3.54.54.63.3


-11.37.1

$2,100,00068.311.8

22.7

12.3

7,3

3.54.64.73.4

'Of the 161 complete data records, the clustering algorithm omitted nine cases as outliers. The numbers in rows 5 and 6 shouldbe read as the percentage increase or decrease in sales for a 7% price change. The numbers in rows 8 and 9 should be read asthe percentage increase or decrease for a unit of service change.

TABLE 3Standard Discrinninant Function Coefficients

Behavioral Variables Function 1 Function 2

1. Relative price2. Relative service3. Account size (sales)4. Market share5. Percentage increase in sales for price drop6. Percentage decrease In sales for price raise7. Percentage decrease in sales for service drop8. Percentage increase in sales for service raise9. Product importance

10. Switching potential11. Market knowledge12. DMP complexity

-.1044.0417

1.0103.0247.1207

-.0280-.0940

.0443-.1164-.2000-.0378

.2012

.8019-.0875

.2270-.5487-.0909

.3779

.0614-.2598

.0677

.0357

.2054-.4069

but often favored price over service, we labeled them"transaction buyers."

Segment 4: bargain hunters. Customers in thismicrosegment were large volume customers that re-ceived the largest price discounts (averaging 11.3%)as well as the highest level of service. They were sen-sitive to any changes in price or service; the productwas very important to their operations. They were mostknowledgeable about alternative suppliers and mostlikely to switch suppliers at the slightest dissatisfac-tion. Customers in this segment were the ultimate bar-gain hunters.

DiscussionThe purpose of our study was to identify buying be-havior variations in macrosegments such as nationalaccounts. We argued that such an analysis would be

useful in redirecting a company's price versus serviceofferings in mature industrial markets. Tables 2 and3 confirm the presence of four such buying behaviorsegments. Even though sales (or account size) pro-vided the single most important discrimination acrossthe four groups (see Tables 2 and 3), without the buy-ing behavior information provided by the rest of thevariables, it would have been impossible for Sig-node's managers to devise strategies to realign cus-tomers along the equity axis of Figure 1—an impor-tant objective of our study.

Figure 2 shows the alignment of the behavioralmicrosegments with respect to the relative price andrelative service variables. As mentioned previously,Signode's managers expected to find their accountsaligned along the equity axis. In particular, the com-pany believed that customers paid for the services theyreceived. Moreover, because Signode had positioned


FIGURE 2Segment Profile

2.5HEIATIVE SERVICE

itself as a full-line, high-value supplier to its nationalaccounts, the managers expected to see a large con-centration of national accounts in the upper rightquadrant (augmented product buyers). The analysis ofbuying behavior microsegments, however, high-lighted the fact that the company's national accountswere aligned in exactly the opposite formation. Onlyone segment of customers, the relationship buyers, waspositioned on the diagonal; the other segments wereon a croSvS-diagonal axis.

Because Signode's products and performancecharacteristics were well understood, the company'smanagers found it plausible that, in this mature mar-ket, many large national accounts did not want to payfor costly support service that they no longer needed.What the managers found surprising was the fact thata large number of accounts wanted a high level ofservice as well as low price (segment 4 buyers). Theprofile of these customers from Table 2 shows theywere clearly the larger customers that, perhaps takingadvantage of the price and service rivalry among thevarious vendors in this competitive marketplace, hadmanaged to extract steep concessions from the com-pany.

Signode's managers also expressed concern aboutthe high price-low service customers (segment I buy-ers), because they were open to competitive inroads.On further examination, the managers were reassuredthat the higher price was correlated with smaller ordersizes (and lower Signode market shares). If and whenthese accounts could be persuaded to increase theirorder sizes, they would be eligible for lower prices.The biggest challenge to the company, after the iden-tification and alignment of the four segments, was howto reorient price and service offerings to improve prof-its without losing sales volume.

At the time of the study, Signode's managers were

under severe pressure from its national accounts to re-duce prices. The buying behavior microsegmentationand the concurrent analysis of the judgmentally gen-erated sales elasticities, however, suggested that priceand service changes would not be equally effective inall four segments. When sales variations were esti-mated for each microsegment, Signode found thatbreaking even on a 7% price change would requiretotal sales volume to change by about 20%. Similarestimates for changes in service showed that breakingeven on a one "unit" service change would requiretotal sales volume to change by 8%. Though thesenumbers should be viewed with caution because thedata were gathered from salespeople in a judgmentalexercise, the estimates generally suggest the follow-ing outcomes.

• Decreasing price is unprofitable for Signode hecause theestimated increase in sales is far below ibe required 20%for every microsegment in Table 2 (see row 5 in Table2).

• Increasing price is profitable in the programmed buyermicrosegment—sales decrease only 15.5% in compari-son with a breakeven of 20%. The sales decrease ex-ceeds the breakeven number for all other microsegments(see row 6 in Table 2).

• Dectieasing service is profitable in the programmed buyermicrosegment because the estimated sales drop is 5.1%.The sales decrease exceeds the breakeven of 8% in allother segments (see row 7 in Table 2).

• Increasing service is barely profitable in the bargain huntermicrosegment—-sales increase 7.3% in comparison wilha breakeven of 8%. The sales increase is far below thisnumber for all other micro.scgments (.see row 8 in Table2).

It is interesting to note that programmed buyerswere willing to pay a relatively higher price withoutdemanding additional service. In a sense, these ac-counts were willing to pay a premium to maintain theirruie-of-thumb purchase allocations and have the flex-ibility of buying a high proportion (46%) of non-Sig-node products. In comparison with other national ac-counts, programmed buyers were not as sensitive toprice or service changes. Hence, to increase marketshare in these accounts, Signode's sale.speople wouldneed to influence their customers' underlying deci-sion-making processes. Signode managers thereforedirected the sales reps handling these accounts to fo-cus their efforts on changing the buying decision-making strategies that limited Signode's share.

The bargain hunters posed a more immediateproblem. These accounts were critical to the companybecause of their very large size—11 accounts con-tributed nearly 25% of national account revenues. Theseaccounts, however, also demanded the lowest pricesand the highest levels of service. Worse still, bargainhunters had the highest propensity to switch to com-petitive suppliers. Hence, managing the bargain hunt-


ers would require considerable tact and skill to keepthem from switching while countering their discountrequests. To protect company margins, Signode man-agers decided that price cutting should be used onlyas a defense against price cuts of competitors. Instead,Signode offered additional service in hop>e that it wouldimprove sales volume beyond the estimated 8%breakeven point. In addition, the NAMs were directedto take an active role in handling these II accounts.We caution industrial marketers who want to realigntheir accounts similarly to consult the provisions ofthe Robinson-Patman Act, which allows price dis-crimination only on the basis of specific factors suchas cost differences or competitive activity (Bowersoxet al. 1980).

The relationship buyers and transaction buyers ofthe middle microsegments had somewhat similar pro-files, except that the former were less likely to pressSignode for price and service concessions. Though therelationship- and transaction-oriented customers didnot pose an immediate problem, both microsegmentswere sensitive to price and service tradeoffs. BecauseSignode managers were concerned about the potentialmigration of these accounts to the bargain hunter mi-crosegment, a separate service management group wascreated to explore ways of adding service value forthis group of customers.

ConclusionsWe believe the practice of industrial market segmen-tation has lagged behind the theoretical developmentsin the field. Though the concept of buy ing-behavior-based segmentation was advanced two decades ago,virtually no application of the concept has been re-ported to date. Many important and valid reasons canbe found for this applications gap but, as our researchdemonstrates, considerable value can be gained by at-tempting to move toward buying-behavior-based seg-mentation. A knowledge of segment behavior helped

the Signode Company to redirect marketing resourcesfor profit gain.

After the analyses at Signode, we believe that evena simple framework, such as the two-dimensional plotof price versus cost-to-serve in Figure I, is capableof unearthing a rich subsegment of behaviors in in-dustrial accounts. As can be seen from the figure, thediagonal equates the price to cost-to-serve for the seller.We hypothesize that the seller's profit would be roughlyequal for accounts located on this axis—when cus-tomers want services (augmented product), they arewilling to pay higher prices. The cross diagonal, how-ever, represents an axis of product differentiation.Clearly, customers that demand and get high levels ofservices for low prices must have altematives, just ascustomers that pay high prices must find the productattractive even though they do not receive the full bat-tery of services. Obviously, the seller's profits are likelyto be higher in the upper left quadrant and lower inthe lower right quadrant than on the diagonal axis.The segment descriptor variables and dimensions arelikely to vary across applications, but nevertheless afew variables could provide rich diagnostics for man-agement actions.

In addition, our research demonstrates a practicaland implementable method for constructing buyer-be-havior-based segments from readily available datasources. The key is to identify variables that ade-quately capture the variance in buying behavior andthat address a specific management problem. By se-lecting the segment descriptor variables to addressmanagers' concems with price and service—two vari-ables under Signode's control—we designed the buy-ing behavior microsegmentation to provide usefulguidance for Signode's account management policies.

We acknowledge that one application does notnecessarily prove the rule, but it at least provides abenchmark for future studies. Our purpose here ismerely to show that industrial market segmentationtheories can be usefully applied to management prob-lems.

REFERENCESAnderson, Erin, Wujin Chu. and Barton Weitz (1987), "In-

dustrial Purchasing; An Empirical Exploration of the Buy-class Framework." Journal of Markeiing., 51, (July), 71—86.

Anderson, N. H. (1974), "Infonnation Integration Theory: ABrief Survey," in Contemporary Developments in Mathe-matical Psychology. Vol. 2, D. H. Krantz et al., eds. SanFrancisco: W. H. Freeman and Company.

Assael, Henry and R. Ellis (1976), "A Research Design toPredict Telephone Usage Among Bell System BusinessCustomers," European Research, I, 38-42.

Bonoma, Thomas V. and Benson P. Shapiro (1983), Seg-menting the Industrial Market. Lexington, MA: LexingtonBooks.

, Gerald Zaitman, and Wesley J. Johnston (1977)."Industrial Buying Behavior," Report No. 70-177 (De-cember). Cambridge, MA: Marketing Science Institute.

Bowersox, Donald J.. M. Bixby Cooper, Douglas M. Lam-bert, and Donald A. Taylor (1980), Management in Mar-keting Channels. New York: McGraw-Hill Book Com-pany. 129-47.

Calinski, T. and J. Harabasz (1974), "A Dendrite Method for


Cluster Analysis," Commumcations in Statistics., 3, 1-27.Cardozo, Richard N. (1968), "Segmenting Ihe Industrial Mar-

ket," in Marketing and the New Science of Planning, R.L. King. ed. Chicago: American Marketing Association,433-40.

(1980), "Siluational Segmentation of IndustrialMarkets," European Journal of Marketing, 14 (5/6). 264-76.

and James W. Cagley (1971), "An ExperimentalStudy of Industrial Buyer Behavior." Journal of MarkeiingResearch. « (Augu.st). 329-34.

Chakravarti, Dipankar, Andrew Mitchell, and Richard Siaelin(1979). "Judgment Based Marketing Decision Models: AnExperimental Investigation of the Decision Calculus Ap-proach," Munagertient Science, 25 (3), 251-63.

Chcron. Emmanuel J. and EIko J. Klein.schmidte (1985), "AReview of Industrial Market Segmentation Research and aProposal for an Integrated Segmentation Framework," In-ternational Journal of Research in Marketing, 2 (2). 101-15.

Choffray, Jean-Marie and Gary L. Lilien (1978). "A New Ap-proach to Industrial Market Segmentation." Sloan Man-agement Review., 3 (Spring). 17-29.

Churchill. Gilbert A., Jr. (1983), Marketing Research: Meth-odological Foundations, 3rd ed. Hinsdale, IL: The DrydenPress.

Coles, Gary J. and James D. Culley (1986), "Not All Pros-pects Are Created Equal." Business Marketing, 71 (May).52-8.

Cook. Thomas D. and Donald T. Campbeil (1979). Quasi-Experimentation: Design and Analysis Issues for Field Set-lings. Chicago: Rand McNally, 39-49.

Cooper. ManhaC. and Glen W. Milligan (1984), "The Effectof Error on Determining ihe Number of Clusters." Collegeof Administrative Science Working Paper Series, 84-2, OhioState University.

Corey, E. Raymond (1978). Procurement Management: Strat-egy, Organization, and Decision Making. Boston: CBIPublishing Company.

Curry, David J. and PeterC. Riesz (1988), "Prices and Price/Quality Relationships: A Longitudinal Analysis," Journalof Marketing, 52 (January), 36-51.

Day, George S. (1981), "The Product Life Cycle: Analysisand Applications Issues," Journal of Marketing, 45 (Fall),60-7.

(1986), Analysis for Strategic Market Decisions.St. Paul. MN: West Publishing Company.

de Kluyver, Comelus A. and David B. Whitlark (1986), "Ben-efit Segmentation for Industrial Products." Industrial Mar-keting Management. 15 (November), 273-86.

Dorfman, Robert A. and Peter O. Steiner (1954), "OptimalAdvertising and Optimal Quality," American Economic Re-view. 44 (December). 826-36.

Doyle. Peter and John Saunders (!985), "Market Segmenta-tion and Positioning in Specialized Industrial Markets,"Journal of Marketing. 49 (Spring), 24-32.

Duda, R. D. and P. E. Hart (1973), Pattern Classification andScene Anatysis. New York: John Wiley & Sons, Ine.

Duncan. Delbert J. (1966). "Purchasing Agents: Seekers ofStatus^Personal and Professional." Journal of Purchas-ing, 2 (August). 17-26.

Emerson, Richard M. (1962), "Power-Dependence Rela-tions," American Sociological Review, 27 (February), 3 1 -41.

Flodhammer, Ake (1988), "A Sales Force Approach to In-dustrial Marketing Segmentation," Quarterly Review ofMarketing, 13 (Winter), 5-9.

Forbis. John L. and Nitin T. Mchta (1981). "Value-BasedStrategies for Industrial Products." Business Horizons, 24(3), 32-42.

Frank. Ronald E.. William F. Massy, and Yoram Wind (1972),Market Segmentation. Englewood Cliffs, NJ: Prentice-Hall,Inc.

French. John R. P. and Bertram Raven (1959), "The Bases ofSocial Power." in Studies in Social Power, D. Carlwright,ed. Ann Arbor: University of Michigan Press.

Gaski. John F. (1984). "The Theory of Power and Conflict inChannels of Distribution." Journal of Marketing, 48 (Sum-mer), 9-29.

Gensch. Dennis H. (1984), "Targeting the Switchable Indus-trial Customer." Marketing Science, 3 (Winter). 4I~54.

Green. Paul E. (1978), Analyzing Multivariate Data. Hin.s-dale, IL: The Dryden Press. 423-37.

Haley. Russell I. (1968), "Benefit Segmentation: A Decision-Oriented Tool," Journal of Marketing, 32 (Summer), 30 -5.

Hlavacek, James D. and B. Charles Ames (!986). "Seg-menting Industrial and High>Tech Markets," Journal ofBusiness Strategy, 17 (Fall). 39-50.

//) Vivo: The Business and Medicine Report (1985), "The De-volution of Prime Vendors" Contraets" (January-February),20-7.

John, George and Barton A. Weitz (1988). "Forward Integra-tion Into Distribution: An Empirical Test of Transaction CostAna\ys\fi.'' Journal of Law. Economics, and Organization,4 (2), 337-54.

Johnson. Hal G. and Ake Flodhammer (1980), "Some Factorsin Industrial Market Segmentation." Indu.strial MarketingManagement, 9 (July), 201-5.

Johnston, Wesley J. and Thomas V. Bonoma (1981), "TheBuying Center: Structure and Interaction Pattems," Journalof Marketing, 45 (Summer), 143-56.

Keman. J. B. and M. S. Sommers (1966), "The BehavioralMatrix: A Closer Look at Industrial Buyers." Business Ho-rizons, 9 (Summer). 59-72.

Kotler. Philip (1988). Marketing Management: Analysis,Planning. Implementation, and Control, 6lh ed. Engle-wood Cliffs. NJ: Prentice-Hall. Inc.. 448.

Levitt. Theodore (1965), Industrial Purchasing Behavior.Boston: Division of Research. Graduate School of BusinessAdministration. Harvard University.

(1980), "Marketing Success Through Differentia-tion—of Anything," Harvard Business Review. 58 (Janu-ary-February), 83-91.

Little. John D. C. (1970), "Models and Managers: The Con-cept of a Deeision Calculus." Management Science, 16(April). B466-B485.

Mahajan. Vijay and Arun K. Jain (1978), "An Approach toNormative Segmentation," Journal of Marketing Research,i5 (August). 338-45.

Milligan, Glen W. (1980). "An Examination of the Effect ofSix Types of Error Perturbation on Fifteen Clustering Al-gorithms," Psychometrika, 45, 325-342.

and Martha C. Cooper (1985). "An Examinationof the Procedures for Determining ihe Number of Clustersin a Data Set." Psychometrika, 50 (2). 159-79.

Moriarty, Rowland T. (1983). Industrial Buying Behavior.Lexington. MA: D.C. Heath & Company.

and John E. G. Bateson (1982). "Exploring Com-plex Decision Making Units: A New Approach," Jourrmlof Marketing Research, 19 (May), 182-91.

and Mort Galper (1978). "Organizational BuyingBehavior: A State-of-the-Art Review and Conceptualiza-


tion." Report No. 78-101. Cambridge, MA: MarketingScience Institute.

and David J. Reibstein (1986). "Benefit Segmen-tation in Industrial Markets," Journal of Business Re-search, 14 (May). 463-86.

and Robert E. Spekman (1984), "An Empirical In-vestigation of Information Sources Used During the Indus-trial Buying Process," Journal of Marketing Research, 21(November), 137—47.

Nagie, Thomas T. (1987). "Segmented Pricing: Tactics forSeparating Markets." chapter 7 in The Strategy and Tacticsof Pricing. Englewood Cliffs, NJ: Prentice-Hall, Inc.

Parket, I. Robert (1972), "The Effects of Product Perceptionon Industrial Buyer Behavior," Industrial Marketing Man-agement, I (April). 339-45.

Phillips, Lynn W. (1981). "Assessing Measurement Error inKey Informant Reports: A Methodological Note on Orga-nizational Analysis in Marketing," Journal of MarketingResearch, 18 (November), 395-415.

Plank, Richard E. (1985), "A Critical Review of IndustrialMarket Segmentation." Industrial Marketing Management,14 (May). 79-91.

Platzer. Linda C. (1984), "Managing National Accounts," TheConference Board Report No. 850. New York: The Con-ference Board.

Porter, M. (1980), Competitive Strategy. New York: The FreePress. 34-44.

Robinson. Patrick J., Charles W. Faris, and Yoram Wind(1967). Industrial Buying and Creative Marketing. Boston:Aliyn and Bacon.

Rogers, Everett M. (1983). Diffusion of Innovations, 3rd ed.New York: The Free Press.

Ross. Elliot B. (1984). "Making Money With Proactive Pric-ing," Har\-ard Business Review, 62 (November-Decem-ber), 145-55.

Sarle, W. S. (1983). "The Cubic Clustering Criterion," SASTechnical Report A-108. Cary. NC: SAS Institute, Inc.

Scherer, Frederick M. (1980), Industrial Market Structure and

Economic Performance., 2nd ed. Chicago: Rand McNally,297-313.

Schmalensee, Richard (1982). "Product Differentiation Ad-vantages of Pioneering Brands." American Economic Re-view, 72 (June). 349-65.

Shapiro, Benson P. and Rowland T. Moriarty (1980), "Na-tional Account Management," Repori No. 80-104. Cam-bridge, MA: Marketing Science Institute.

, V. Kasturi Rangan, Rowland T. Moriarty, and El-liot P. Ross (1987). "Manage Customers for Profits (NotJust Sales)," Harvard Business Review, 65 (September-Oc-tober), 101-8.

Sheth, J. (1973), "A Model of Industrial Buyer Behavior,"Journal of Marketing, 37 (October), 50-6.

Silk. Alvin J. and Manohar U. Kalwani (1982). "MeasuringInfluence in Organizational Purchase Decisions." Journalof Marketing Research, 19 (May). 165-81.

Simon. Hermann (1979), "Dynamics of Price Elasticity andBrand Life Cycles: An Empirical Study." Journal of Mar-keting Research, 16 (November), 439-52.

Tellis, Gerard J. and Birger Wemerfelt (1987), "CompetitivePrice and Quality Under Asymmetric Information," Mar-keting Science, 6 (3), 240-53.

Tutton. Merrill (1987). "Segmenting a National Account,"Business Horizons 30 (January-February), 61-8 .

Webster. Frederick E. (1984). Industrial Marketing Strategy,2d cd. New York: John Wiley & Sons, Inc.

— — and Yoram Wind (1972), Organizational BuyingBehavior. Englewood Cliffs. NJ: Prentice-Hall, Inc.

Wilson. David T. (1971), "Industrial Buyers* Decision-Mak-ing Styles," Journal of Marketing Research, 8 (Novem-ber), 433-6.

Wind, Yoram (1978), "Issues and Advances in SegmentationResearch," Journal of Marketing Research, 15 (August),153-65.

and Richard N. Cardozo (1974), "Industrial Mar-keting Segmentation," Industrial Marketing Management,3 (March), 153-65.

Younger. Michael (1986). "Achieving Product Growth in aMature Industry," Chief Executive (March). 19-20.

Reprlnl N«. JM5641IM


Section - 3.2Segmenting in Mature Markets

Documents

Transcript of Section - 3.2Segmenting in Mature Markets