Promotion Heterogeneity and Consumer Learning: Refining the Deal-Proneness Construct

Caroline M. Henderson, Dartmouth College
ABSTRACT - Researchers have had difficulty clearly defining a segment of deal-prone consumers. One problem with such research is use of an atheoretical and one-dimensional definition of deal-proneness which fails to recognize promotion heterogeneity. Using scanner panel data for two product categories, we find distinct segments that respond to different forms of promotion. On the basis of these results, we develop a theory of promotion learning: that consumers learn to respond in an increasingly complex fashion to their promotion environment. We then develop suggestions for testing this hypothesis.
[ to cite ]:
Caroline M. Henderson (1994) ,"Promotion Heterogeneity and Consumer Learning: Refining the Deal-Proneness Construct", in NA - Advances in Consumer Research Volume 21, eds. Chris T. Allen and Deborah Roedder John, Provo, UT : Association for Consumer Research, Pages: 86-94.

Advances in Consumer Research Volume 21, 1994      Pages 86-94


Caroline M. Henderson, Dartmouth College


Researchers have had difficulty clearly defining a segment of deal-prone consumers. One problem with such research is use of an atheoretical and one-dimensional definition of deal-proneness which fails to recognize promotion heterogeneity. Using scanner panel data for two product categories, we find distinct segments that respond to different forms of promotion. On the basis of these results, we develop a theory of promotion learning: that consumers learn to respond in an increasingly complex fashion to their promotion environment. We then develop suggestions for testing this hypothesis.


One popular research area has centered on consumers who respond to promotionCthe identification of the "deal-prone" consumer (Webster 1965; Carman 1969; Blattberg et al. 1978; among others; see review by Blattberg and Neslin 1990). The deal-proneness studies to date have failed, however, to provide managers with a coherent view of the consumers most sensitive to promotion. One reason for this may be that deal-proneness should be seen as specific to individual types of deals. We propose refining the construct of deal-proneness through recognition of promotion heterogeneity.

Three empirical questions are addressed:

1) Do segments exist which vary in response to certain types of promotion?

2) Are such segments clearly identifiable in the population?

3) Do the segments vary by product category?

We then use this analysis to build a theory of promotion learning. In the final section we present suggestions for testing this theory.


The construct of deal-proneness, first used by Webster (1965), refers to the consumer tendency to respond to sales promotion. The construct is typically measured by behavior; the deal-prone are those who buy promoted brandsCredeem coupons, buy special packs, stock up when price is reduced, or otherwise take advantage of promotions. Difficulties in identifying the deal-prone may be explained by the impossibility of comparing across different studies. Deal-proneness can be specific to individual product categories and to individual types of deals. Yet most identified deal-proneness studies (Blattberg and Neslin 1990) deal with very restricted types of promotions (typically coupons) and a restricted number of product categories. To our knowledge, only Schneider and Currim (1991) measure deal proneness with a heterogeneous set of promotions.

Generalizations across categories can be dangerous; Carman's (1969) measure of deal-proneness produces different results for the two product categories in his study. The Blattberg, Peacock and Sen (1976) segments also show differences among the categories used. Such differences should be expected; product categories differ in motivations to use deals (financial expense, social risk, among others) and in the cost of using promotions (deal availability). If these two factors combine to make deal-proneness category-specific, all single-product studies will be potentially contradictory.

Generalizing across deal types may also be ill-advised. Dodson, Tybout, and Sternthal (1978) provide a compelling analysis of the value of studying promotions as separate types of stimuli. The authors argue that different types of deals should elicit different types of behavior, consistent with the precepts of economic utility theory and self-perception theory. For example, deals requiring substantial consumer effort for a small economic gain should lead to the attribution that a brand is purchased because it is preferred not because of its promotion. Deals requiring less effort relative to gain should be effective in undermining consumer brand loyalty. This theory predicts that price-off promotions, an example of the second type, encourage less loyal behavior than coupons, which involve more consumer effort. While the Dodson study's empirical findings support this theory, the study does not analyze consumer segments. Thus, the findings may actually show that different consumers are susceptible to different kinds of promotion.

To follow this line of reasoning, it may be possible to identify the attributes of promotion to which consumers differentially respond. Such bases should include the fundamental appeal of promotions or the costs of responding to them. The literature has developed theoretical explanations on these two issues.

A basic economic theory of deal-proneness is that consumers are motivated by a promotion's savings but must trade off such savings against a promotion's costsCparticularly the cost of time spent taking advantage of the promotion (Blattberg, Buesing, Peacock and Sen 1978; Narasimhan 1984). While all consumers are intrinsically motivated by possible savings, they are heterogeneous with respect to time costs and thus vary in their deal-proneness. (Narasimhan (1984) shows additionally that those consumers having the lowest costs are also the most price elastic.) Blattberg, Eppen and Lieberman (1981) stress the importance of inventory holding costs which will also vary across the population and lead to varying degrees of deal-proneness. Previous researchers have used this approach to generate hypotheses that identify the deal-prone by relating demographic variables to the consumer's promotional costs. For example, Narasimhan (1984) hypothesizes that heavy usage, unemployment, higher education, and absence of children reduces the cost of using coupons. Thus, promotions can be seen to motivate by providing a basic economic utility but differences in the costs of using promotions may account for heterogeneity of response.

We propose to refine this view further by building on the Dodson, Tybout and Sternthal (1978) observation that all deals are not equivalent. In this approach each deal may be described by its attributesBBthe cost/benefits pattern uniquely pertaining to the type. As Figure 1 illustrates, promotional types may vary considerably in such attributes. For example, coupons may provide large economic incentives but incur more "time preference" costs due to the extensive pre-use activities required. Many deal pack offers force stockpiling by providing extra product; price-off offers may require purchase of multiple units to attain full savings, etc. Figure 1 is not intended to show a fully-formed theory of promotional differences, simply to suggest that cost/benefit patterns are sufficiently varied to require separate investigations for each promotional type. To illustrate, a consumer with high time costs who is sensitive to price may bypass coupons yet still buy price-off products. Thus, consumer segments which respond only to certain types of promotion may be found. Such segments should also be category-specific. If such segments are found, deal-proneness should be considered multidimensional across deal types.




Two separate scanner panel data bases were used. Data set 1 covered 28 weeks of consumer purchases in a "northeast scanner market"Ca metropolitan area of about 250,000 population. This market included three major grocery chains, with a total of fifteen stores. These stores kept a complete scanner record of all purchases made by the 2463 members of the panelBBcoverage amounting to 85 percent of the all-commodity-volume of the area. Data included purchase records and panel demographics. Available promotional variables included price changes, deal packs, coupons and two types of local advertising; retailer advertising was for weekly price specials and manufacturer advertising carried coupons.

Two product categories were studied to mitigate some of the weaknesses of single-product studies. We analyze purchases of bathroom tissue ("paper") and caffeinated instant coffee ("coffee"). These two product categories provide an interesting contrast. Private label and generic brands are purchased less frequently in coffee than paper. Paper is more likely to be promoted to the trade, and coffee promoted directly to the consumer. The average interpurchase time for paper is 19 days, for coffee it is 43 days.

Data set 2 covered 24 months of consumer purchases from a separate scanner market and included purchase records only for the coffee category. The data were analyzed for instant coffee and provided a replication of the analysis done with data set 1. The categories are equivalent and the time periods roughly comparable, although there are geographic differences in the consumer panels. There are also a few differences in the variables available for study. Data set 2 includes price changes and deal packs in common with data set 2. Coupons, however, are analyzed as store coupon and manufacturer coupons. Retailer advertising, called features, in data set 2 is also comparable but, in addition, a display variable was available.

Proneness Measures

A common measure for deal-prone behavior is percentage of purchases made on deal (Montgomery 1971, Wierenga 1974) and adjusted for the relative prevalence of deals (Webster 1965, Carman 1969). Percentage of purchase measures were the primary measures selected for this study. To develop these measures, the data files were combined and summarized to profile the consumers purchasing each of the product categories (1292 coffee purchasers making 4910 purchases and 2198 paper purchasers making 23727 purchases in a static sample for the 28-week period in data set 1). The percentage of total category purchases representing deals was computed for four types of promotional conditions: Deal Packs (special packages including extra product, reusable containers, or premarking with a price discount.), Retail Ad/Feature (retail advertising in local newspapers), Manufacturer Ad (manufacturer advertising, including coupons, in local newspapers), Coupon ( purchase was made using a coupon from any source), Display (Data set 2 onlyCspecial display within the store).

As percentage of purchase measures, these variables assume that purchase samplings for each consumer represent an overall behavior pattern. Data set 1 is limited by a six-month time frame, since, on average, there are only four purchases of coffee per buyer. We have assumed that use of a coupon on five out of ten purchases is equivalent to use of a coupon on one out of two purchases. Thus, we separate deal-proneness from frequency of category usage. While this small sample size may tend to overstate degrees of brand loyalty among infrequent purchasers, a period of six months has been used previously to study loyalty (Johnson 1984). This time frame should introduce no systematic bias in promotional usage. Data set 2 covers a two-year period and is not subject to this potential weakness.

A fifth measure of deal-prone behaviorCPrice ChangeCwas computed as the difference, measured in dollars, between the price paid on the purchase and the average shelf price during the week preceding the purchase. (A value of -.05 indicated that the brand had been discounted five cents.) As this value captures both promotional and nonpromotional price changes, it represents the consumer's response to lowered price. This operationalization avoids imputing promotional price changes from sales data (see, for example, Guadagni and Little 1983). Price changes for paper were computed from average shelf prices across all color combinations of a particular brand-size. Coffee price changes were identified for each brand-size.

The promotion variables are not highly correlated across the population. One exception is a .55 correlation in the coffee category between price change and retail ads. The decision to analyze each product category separately was substantiated by cross-category correlation. For example, the highest correlation was .26 for percentage of coupon purchases between the two categories in data set 1.


Two random samples of 320 each (the maximum allowable because of computer constraints) were drawn from each category's consumers. To ensure that these samples did not differ in any way, mean values were compared across all variables in both samples from each category. The variables were standardized and used to create a Euclidean distance dissimilarity matrix, which then was clustered by two algorithms: nearest neighbor and minimum squared error (Schlaifer 1981). The dendrograms were analyzed to find the most suitable number of clusters using both the "elbow" and "intuitive" (Calantone and Sawyer 1978) rules.

Analysis procedure follows the three-step process needed to assess the strength of a cluster solution (Punj and Stewart 1983). This process tests that the cluster solution is 1) different from what could be obtained by a random assignment to groups (F-test across cluster variables); 2) reliable (cross-classification between samples); and 3) related to differences between individuals (F and chi-square tests across non-clustering variables).

In the second step, the output was analyzed for split-half stability by classifying each individual with a discriminant function developed from each sample's cluster analysis (as recommended by Calantone and Sawyer 1978). The ability of each sample's function to correctly classify its own and the other sample was tested by the percentage of correct classification across samples according to the proportional chance criterion (Morrison 1969).

The third test of the cluster resultsCgeneralizabilityCinvolves the relationship between the clusters and demographics and purchase behavior variables. Consumers in each category were first classified into a cluster with the discriminant function generated from one of the samples of 320. 1276 out of 1292 coffee purchasers and 2196 out of 2198 paper purchasers were successfully classified. Next, all available demographics and purchase variables were analyzed across the groups using F-tests for continuous variables and chi-square tests for nominal or ordinal data.

Data set 2 was used to replicate the analysis in data set 1. The same series of steps were performed on the instant coffee category.


Data Set 1

Five clusters were developed for each sample in each product category, using the minimum squared error cluster analysis. The nearest neighbor algorithm did not yield interpretable resultsCa common problem when cluster boundaries are not completely clear (Schlaifer 1981). As no strong elbow could be found to select an optimal number of clusters, the final number of clusters was selected judgmentally to provide usable results with an adequate number of consumers per cluster and a clear pattern within the cluster. As Table 1 shows for the entire data base of consumers in each category, the clusters have significantly different mean values on each promotion variable. The paper clusters can be described as:

Cluster 1: Deal Packs (13 percent of the sample)

This group has almost three-quarters of its purchase concentrated in deal pack merchandise. The group is very low on all other variables.

Cluster 2: Retail Ads (5 percent)

This small group appears to have a very high percentage of purchases (almost one-third) that coincide with the retail featured brand.

Cluster 3: Coupons (9 percent)

Over half of this group's purchases are made with a coupon.

Cluster 4: Manufacturer Ads/Coupons (1 percent)

This tiny group appears most responsive to manufacturer ads as well as coupons.

Cluster 5: No Response (72 percent)

The majority of the consumers in the paper category do not appear to be deal-prone.

The coffee cluster results are somewhat different:

Cluster 1: All Promotion (4 percent)

This segment is responsive to four of the five types of promotionmeasured. One-third of customer purchases are made with a coupon, over three-quarters are the retail featured brand, and, on average, the group saves $0.24 on the regular price of the brand. Seventeen percent of purchases also are made during the week of a manufacturer ad for the brand. Deal packs, however, are not an important part of purchasing.

Cluster 2: Deal Packs/Coupons (15 percent)

This group is similar to the paper deal pack segment with the addition that over one-third of their purchases are made with a coupon.

Cluster 3: Manufacturer Ads/Coupons (9 percent)

This group also approximates a paper segment (cluster 4), as it has a very high percentage of purchasing coinciding with manufacturer advertising.

Cluster 4: Coupons (32 percent)

The coupon group is the largest of the coffee clusters. These consumers use a coupon on almost two-thirds of their coffee purchases but do not appear particularly responsive to other forms of promotion.

Cluster 5: No Response (40 percent)

Although smaller in size, this group is equivalent to paper cluster 5 in showing little response.

Split-half clustering yielded similar groups for the two samples. Sizes of the clusters were quite comparable and means were not significantly different between samples. Clusters emerged in the same order for coffee and with one reversal (clusters 2 and 3) in paper. This cluster structure is internally reliable since two samples for each category could be used for fairly accurate cross-classification for members of the other sample on the clustering variables. (On average, 72 percent of paper consumers and 81 percent of coffee consumers were accurately cross-classified, compared with accurate reclassification on the same sample of 91 percent for each and with the proportional chance criterion of 55 percent for paper and 29 percent for coffee.) Such results are impressive given the sensitivity of cluster analysis to minor variations in data sets (for example, Funkhouser 1983). (Detailed study results are available from the author upon request.)





Data Set 2 - Replication

Results for the instant coffee category are shown in Table 1a and can be described as:

Cluster 1: No Response (46 percent)

This cluster is comparable in size and purchasing pattern to cluster 5 in data set 1.

Cluster 2: Coupon/Feature/Display (4 percent)

This cluster is somewhat unique compared with data set 1. A variety of promotions, with the exception of price changes, are used.

Cluster 3: Manufacturers Coupon (39 percent)

This cluster is comparable to cluster 4 in data set 1.

Cluster 4: All Promotion (10 percent)

Although the size of this segment is larger than cluster 1 in data set 1, the cluster is similar in using of a variety of promotions.

Cluster 5: Deal Purchases/Coupons (2 percent)

A much smaller cluster than cluster 2 in data set 1, this cluster has a similar tendency to buy the brand in special packaging.

This analysis reveals a number of similarities with data set 1 (Table 2).


The clusters were found to have statistically significant differences on brand loyalty, demographics, and other purchase variables, as shown in Tables 3 and 4. These tables show that consumers who are sensitive to nearly every type of promotion (coffee cluster 1) are also very loyal. They achieve savings through a combined strategy of coupons, price-offs, and large package sizes; they are older and nonworking, presumably able to incur the extra stock-up and holding costs associated with the combination of brand loyalty with deal proneness. This finding validates the importance of distinguishing loyalty, promotion usage, and brand choice in segmentation studies, as initially conceived in Blattberg and Sen (1974). These results disagree with the position of Hackleman and Duker (1980, p. 172) that deal-proneness "by definition" involves low brand loyalty; these results suggest that certain deal-prone consumers are high in brand loyalty.


Nonresponders to promotion (paper cluster 5 and coffee cluster 5 in data set 1 and cluster 1 in data set 2 ) are likely to buy in small quantities. These consumers are susceptible to private labels. Responding to sales promotion and buying private labels or generics appear to be antithetical consumer strategies, in distinction to the Wierenga results (1974). Consumers do not appear to switch between major brands on deal and private labels, which do not offer deals but are frequently priced lower. Instead, those with the highest percentages of private label purchases are only infrequent users of major brand promotions.

To provide a multivariate classification of the clusters, five-group discriminant analysis was performed for each product. For both categories the discriminating power of the independent variables was statistically significant using the proportional chance criterion. The paper functions reclassified 82% of the data base compared with a proportional chance classification of 55%. For coffee, the functions correctly classified 43% of a holdout sample, a significant improvement beyond a proportional chance of 29%. The functions indicate that promotion segments can be successfully identified with demographics and purchase behavior variables. Of the two variable types, knowledge of purchase habitsBBparticularly brand loyaltyBBmay be the more helpful in identifying groups of deal-prone consumers. Since purchase habits can be category specific, it may not be realistic to generalize segments across product categories.

To explore this issue of generalizability, clusters were compared across the two categories. The cross-classification in Table 5 was developed for the two categories by making certain assumptions about similarities in cluster description. Cluster 5BBnonresponsive consumersBBwere roughly equivalent. Coffee cluster 2 was equated with paper clusters 1 and 2 as referring to a "non-coupon" response. Coffee cluster 1 appears to have no counterpart in paper. This most highly deal-sensitive segment is only a small percentage of the market and is only found for the one product category.

Using these assumptions we classified 37 percent of consumers as belonging to an equivalent deal-prone cluster in each category. The 37 percent is indistinguishable from a random assignment of 36 percent. Thus, knowledge of a consumer's deal-proneness in one category provides no information for classification in another category. This is support for the importance of category-specific research.

Although a no-response segment (cluster 5 in both) is found in both categories, its size varies; almost three-quarters of the paper consumers, but fewer than one-half of the coffee consumers, are non-responsive. This discrepancy might be explained by the relative importance of promotion in these two categories. The coffee industry has led in consumer promotion, particularly in the form of coupons, while the paper industry tends to use trade deals (Quelch 1982). While trade deals can affect consumers when the product is advertised, displayed, or reduced in price, market areas studied in this research show such pass-throughs as being primarily replaced by "everyday low pricing." The net effect is more availability of promotional opportunities for coffee purchasers than for paper buyers. The relative importance of these categories on the shopping list may also be a factor. Coffee, although less frequently bought, costs more than the paper product. Price may motivate consumers to take advantage of deals. Such considerations may shift the costs and benefits of specific types of promotion depending on category.


The research has shown that promotion heterogeneity should be recognized in the deal-proneness construct. Distinct and identifiable segments can be developed which isolate consumers who respond only to particular forms of sales promotion. These results can be replicated in a second data set. Extending this work, researchers might analyze wider varieties of markets and product categories.

Our research, although largely descriptive, also sees the need for more theory building in the area of deal-proneness. Our empirical findings can provide a basis for generating one such theory. Two noteworthy findings are: the most deal-prone consumers in each product category are older; and, in the coffee category, both ends of the response spectrum are quite brand loyal. In tying these threads together, we speculate on the evolution of deal-proneness segments.

We hypothesize that consumers will changeBBbecoming more or less willing to trade their time for economic gains, becoming more price-sensitive, or learning better use of promotion information. Such changes may occur through several mechanisms: the consumer's changing economic environment or circumstances, or adult socialization processesBBas they learn new shopping styles that revise their expectations. Therefore, consumer response to individual promotion types may vary over time. The question isBBhow can change patterns be predicted? Figure 2 illustrates this process.

Economic incentives and inventory costs may change in a relatively predictable fashion over time. All things being equal, we might expect economic incentives of promotions to become less important to consumers as their own economic status becomes more secure with time. Similarly, inventory holding costs may become less salient as home sizes increase, etc. Thus, the economic attractiveness of promotions may diminish eventually, but may be easier to take advantage ofBBin an inventory sense. Since these mechanisms may balance out, they provide no clear average changes over time.









The attributes loosely grouped under "time preferences," however, are a set of consumer skills at which consumers can learn to become more proficient. We therefore predict that consumers will learn such skills and begin to respond to complex forms of promotion. Non-responsive consumers might select brands without considering available coupons, ads, price cuts or other promotions. The learning process might begin with consumer behavior requiring a slight adaptationBBmaking the regular purchase under a price-off or feature condition. Consumers might then move slowly to take advantage of promotions requiring more effort as they become increasingly adept. Consumers are "rewarded" for their successful usage of promotion and thus an operant conditioning mechanism (Rothschild and Gaidis 1981) may serve to escalate a consumer's involvement with promotion. At this point, promotion-sensitivity may increase to the level at which offers for brands other than the regular brand are attractive. Thus, brand loyalty is undermined. At the end of this continuum is the stereotypical "price shopper"BBone who always uses coupons and rebates and switches easily. This view is supported by Demsetz (1962), who shows a relationship between consumer experience with a product category and price paid. Consumers who have been in the market longer buy less expensive brands. At this extreme point, however, we might further hypothesize that brand loyalty could again strengthen. Presumably a consumer with the skills to identify and take advantage of multiple forms of promotion might successfully use these skills to take advantage of deals for preferred brands without needing to switch.

This study cannot provide a real test of the hypothesis since we cannot track the different stages in an ordered progression over time. We strongly suggest that future research be directed toward an over-time research design, such as utilized by Calantone and Sawyer (1978), that tracks such changes as they occur. If confirmed, the promotion learning hypothesis may provide a unifying explanation for deal-proneness studies and measures previously not comparable. It is most important for marketers to gain insight into primary demands for sales promotion, plus knowledge of the points at which they can influence this demand.


Bearden, W.O., D.R. Lichtenstein and J.E. Teel (1984), "Comparison Price, Coupon, and Brand Effects on Consumer Reactions to Retail Advertisements," Journal of Retailing, Vol 60, #2, Summer, pp. 11-34.

Beem, E.R. and H.J Shaffer (1981), Triggers to Customer ActionCSome Elements in a Theory of Promotional Inducement, Marketing Science Institute.

Blattberg, R.C. and S.K. Sen (1974), "Market Segmentation Using Models of Multidimensional Purchasing Behavior," Journal of Marketing, Vol. 38, #4, October, pp. 17-28.

Blattberg, R.C., P. Peacock and S.K. Sen (1976), "Purchasing strategies Across Product Categories," Journal of Consumer Research, Vol. 3, #3, December, pp. 143-154.

Blattberg, R.C., T. Buesing, P. Peacock, and S.K. Sen (1978), "Identifying the Deal-Prone Segment," Journal of Marketing Research, Vol. 15, #3, August, pp. 369-377.

Blattberg, R.C., and S.A. Neslin (1990), Sales Promotion: Concepts, Methods and Strategies, Englewood Cliffs, New Jersey: Prentice-Hall.

Blattberg, R.C., G.D. Eppen and J. Lieberman (1981), "A Theoretical and Empirical Evolution of Price Deals for Consumer Nondurables," Journal of Marketing, Vol. 45 #1, Winter, pp. 116-129.

Calantone, R.J. and A.G. Sawyer (1978), "The Stability of Benefit Segments," Journal of Marketing Research, Vol. XV, August, pp. 395-404.

Carman, J.M. (1969), "Some Insights into Reasonable Grocery Shopping Strategies," Journal of Marketing, October, pp. 69-72.

Dodson, J.A., A.M. Tybout, and B. Sternthal (1978), "Impact of Deals and Deal Retraction on Brand Switching," Journal of Marketing Research, Vol. 15, #1, February, pp. 72-81.

Densetz, H. (1962),"The Effect of Consumer Exposure on Brand Loyalty and the Structure of Market Demand," Econometrica, January.

Funkhouser, G.B. (1983), "A Note on the Reliability of Certain Clustering Algorithms," Journal of Marketing Research, Vol. XX, February, pp. 99-102.

Guadagni, P.M. and J.D.C. Little (1983), "A Logit Model of Brand Choice Calibrated on Scanner Data," Marketing Science, Vol. 2, #3, Summer, pp. 203-238.

Hackleman, E.C. and J.M. Duker, (1980), "Segmenting the Deal Prone Consumer Using Consumption Volume," in R. Bagozzi, et al, eds., Marketing in the 80's, American Marketing Association, pp. 172-175.

Johnson, T. (1984), "The Myth of Declining Brand Loyalty," Journal of Advertising Research, Vol. 24, #1, February/March, pp. 9-17.

Montgomery, D.B. (1971), "Consumer Characteristics Associated with Dealing: An Empirical Example," Journal of Marketing Research, Vol. 8, #1, February, pp. 118-120.

Morrison, D.F. (1969), "On the Interpretation of Discriminant Analysis," Journal of Marketing Research, Vol. VI, May, pp. 156-163.

Narasimhan, C. (1984), "A Price Discrimination Theory of Coupons," Marketing Science, Vol. 3, #2, Spring, pp. 128-146.

Punj, G. and D.W. Stewart (1983), "Cluster Analysis in Marketing Research: Review and Suggestion for Application," Journal of Marketing Research, Vol. XX, May, pp. 134-148.

Quelch, J.A. (1983), Trade Promotion by Grocery Products Manufacturers: A Managerial Perspective, Marketing Science Institute.

Rothschild, M.L. and W.C. Gaidis (1981), "Behavioral Learning Theory: Its Relevance to Marketing and Promotions," Journal of Marketing, Vol. 45, #2, Spring, pp. 70-78.

Schlaifer, R. (1981), Users' Guide to the AQD Collection, Harvard Business School.

Schneider, L.G. and I.S. Currim (1991), "Consumer Purchase Behaviors Associated with Active and Passive Deal-Proneness," International Journal of Research in Marketing, Vol. 8, #3, pp. 205-222.

Webster, F.E. (1965), "The Deal-Prone Consumer," Journal of Marketing Research, Vol. 2, #2, May, pp. 186-189.

Wierenga, B. (1974), An Investigation of Brand Choice Processes, Rotterdam University.