Consumer Response to Product-Line Extensions: a Consideration For New Product Planning


J. Taylor Sims (1972) ,"Consumer Response to Product-Line Extensions: a Consideration For New Product Planning", in SV - Proceedings of the Third Annual Conference of the Association for Consumer Research, eds. M. Venkatesan, Chicago, IL : Association for Consumer Research, Pages: 80-95.

Proceedings of the Third Annual Conference of the Association for Consumer Research, 1972      Pages 80-95


J. Taylor Sims, University of South Carolina

[J. Taylor Sims is Assistant Professor of Marketing, University of South Carolina.]

If a household is familiar with a brand, it may be induced by a new flavor introduction to buy an extra unit sooner than it would otherwise. These effects would be part of the short-run impact of a new flavor introduction. Such an interim purchase would, however, be unlikely to affect their long-run purchases of the product category, with reference to the average number of items and volume attributed, in any substantial way. The average number of items within a grocery product class that contribute to the total assortment mix of products within a household should not change, given relatively stable socioeconomic conditions. Thus, it is proposed that the long-run effect of a new flavor introduction on a "repeat" buyer of the brand class to which the flavor belongs is inconsequential with reference to additional total brand volume that could be forthcoming.

If the above hypothesis can be substantiated then the impact of a product-line extension is primarily dependent on the retention of new buyers who currently either fulfill most of their requirements with competing items from other brands or have infiltrated from outside the product market. The ability of the product-line extension to lure and maintain new buyers, i.e., convert them into repeat purchasers, is the primary determinant of success. [Of course, the strategy involving product-line extensions can also be of a defensive nature, e.g., the matching of competitors' products in order to hold one's current market position. The assumptions in this analysis are based on the firm's inherent desire for market growth and its ability to lead from strength.]


The main emphasis of the study is analytical. It seem necessary to point this out because most of the previous studies dealing with brand loyalty have dealt at length on the strengths and weaknesses of methodological techniques instead of the resulting analyses (Brown, 1962; Cunningham, 1961; Fourt & Woodlock, 1966; Frank, 1962; Howard, 1965; Kuehn, 1962; Massy, 1967; and Parfitt & Collins, 1968). In addition, these investigations have examined product loyalty primarily at the brand level with little or no attempt to look at type compositions within brand structure. In contrast, the emphasis of this study is on the analysis of type as well as brand loyaltY.

In order to test the hypothesis of this study, data from the Market Research Corporation of America's National Consumer Panel on the introduction of new types within a particular brand of a specific product class were examined. Other product classes were made available, on a limited basis, for comparative analyses.

The National Consumer Panel consists of 7,500 households stratified to represent the demographic structure of the total United State population. Continuous purchase data are collected by means of a diary which is filled out by the consumer and returned weekly for processing into cumulative computerized data files. These data can be classified in numerous ways (Market Research Corporation of America, 1970). For the purpose of this study data were classified in the following manner:

1. The number of total households purchasing a product class on a calendar quarter to quarter basis;

2. The total volume contributed by purchases of these households on a quarter to quarter basis;

3. The total number of brands and/or types introduced during these periods;

4. The total number of brands and/or types withdrawn during these periods;

5. The average number of items purchased per household by quarterly periods by brand and type;

6. The average volume purchased per household by quarterly periods by brand and type.

The data for the primary analysis in the study were tabulated by U.S. Total consisting of types of the product class shown within brands and also brands shown within types (flavors). Purchases made by over 5,000 households were examined. The first purchase of a new flavor could have been made in any of the four quarterly periods within a year. This quarter of first purchase was the base period in the analysis. Purchase behavior was examined in terms of prior and subsequent purchases to the base period. Data were evaluated on both a cumulative and non-cumulative basis in order to observe new and repeat buyer patterns as well as duplication among purchases of brand and types.

The average number of items purchased and the average volume per household has not changed significantly through time for the product under consideration. There have also been more types than brands introduced with very few withdrawals from the market of either type or brand. This phenomenon is an important factor as will be discussed later.


The basic research design is quasi-experimental and is classified by Campbell and Stanley as having the general characteristics of a time series:

"The essence of the time-series design is the presence of a periodic measurement process on some group or individual and the introduction of an experimental change into this time series of measurements.....It can be diagramed thus following Campbell and Stanley (Campbell & Stanley, 1968):

01 02 03 04 X 05 06 07 08''

With this design an initial examination was made of repeat buyers of the new flavor introduction who were previously loyal to the new flavor's brand. These data were then compared to repeat purchases of new buyers in order to closely determine the source of retention.

The second step in the analysis involved tracing the effects of the above buying patterns on the other types within the brand containing the new flavor as well as on competitive items. A Markov model was used for this portion of the analysis.

Product-Line Extensions and the Product-Life Cycle

Although client non-disclosure pacts limited this writer's investigation, he was able to obtain limited data from selected product classes other than the one used for the primary analysis of this study. Ten such products were made available for the purpose of observing changes in average volumes per household over a three-year period. Five of the product categories were classified as being in the growth phase of their product life cycle. [Data furnished by the Market Research Corporation of America. Product categories and actual counts cannot be disclosed. Definitionally, a product is considered, in this study, to be in the mature phase of its product life cycle if an asymptotic state has been reached in terms of new buyers entering the product class, i.e., increased new buyer penetration for the entire product class is not exceeding the average rate of population growth per year.] Three of the products were purchased as frequently as once in every twenty days; four had purchase cycles of once in every thirty days; the balance had purchase cycles ranging from once in every forty-five to ninety days. Among all mature products, significant increases in purchase volumes were not observed among purchasing households. In addition, long-run market penetration rates and brand shares appeared to be in an asymptotic state among the five mature products. The year to year penetration rates (1967 through 1970), in terms of gains in household buying and buying rates, did not change significantly above population growth which averaged 1-1/2% per year for the three years of observation.

The five mature product classes all revealed total buyer penetration rates in excess of 50% with definite stable-state signs. The mature product class used for the primary analysis in this study has penetrated 80% of total U.S. households, a rate which held steady prior and subsequent to the introduction of new items during the three-year observation period.

The five product classes classified as being in the growth phase of their product life cycle have all been in existence ten years or less. Year to year trends for these products revealed statistically significant growthS in numbers of households buying and in purchase rates, above the average rate of population growth. In addition, these product classes have penetrated a range, thus far, of from only 20% to 50% of total U.S. households. Thus a relative range for continued market penetration for the newer products would appear to exceed that of the mature products.

The above discussion, however, does not change the meaning of the hypothesis in this study. If long-run growth is to be achieved then lasting volume must be obtained from outside the market, from competitors, or from light buyers. The new product brand cannot rely on its hard core heavy buyers to purchase additional volume. What is implied, however, is that with newer product categories, still in the early stages of growth, the hard core buying element has, in all likelihood, not had time to build to its ultimate level leaving room for expansion of lighter buyers into this category. In contrast with a mature product class, growth product classes may expect conversion of light and medium buyers into the heavy category until such time as the product moves into the leveling mature phase of its product life cycle.

The key point in the analysis of product-line extensions may well lie in the nature of the product class from which it comes, i.e., mature versus growth, and not just from the innovative nature of the line extension. It may, in fact, be possible to introduce two or three new flavors within a year in a brand that represents a growth product class with the result that lasting buyers are brought in at no expense to existing flavors, or at a rate that offsets any proliferation between items. On the other hand, a mature product class that has achieved maximum penetration in the market to the extent that additional buyers are not significantly forthcoming, may find extensive proliferation upon existing items as line-extensions are added.

The crux of the whole matter is that the mature product has a more limited means for expansion than does the newer product. The newer growth product is still building penetration in the market as well as gaining volume from competitive brands. The mature product may be limited to only deriving its additional volume from competitors.

Table 1 relates share trends among flavors and brands of a growth phase product. The flavors shown are all within Brand H, and indicate the addition of two flavors to the line, one during period eight, and the other during period eleven. The product has been in existence for ten years.

Periods nine through twelve, as related in Table 1, produced a record volume for the product class. The primary reason for this rate of growth was the continuing acceptance of Brand H's products. The brand represents a strong innovation in its market, and, thus, far, competition has not competed effectively, either in quality or number and types of items available. The squeeze on competitive brands by Brand H, in terms of share, indicates not only gains from competition, but a general monopolization tendency. The brand has been able to bring new buyers into the market and, at the same time, convert competitor's buyers to its market. However, the existing loyal buyers to the brand did not increase their purchasing in total. Transaction size remained a steady 2.08 units throughout this period of growth for these heavy buying households. The brand was able to convert marginal buyers into heavy buyers, at a significant rate, thus expanding the buying base.

During periods one through seven, Brand H marketed eight items with a total brand share ranging from 57% to 59% of the total market. By the end of period twelve this share had increased to 75% of the market. Flavor I was introduced during period eight and immediately gained 18% of Brand H's total share, which increased during the period to 64%. By period nine Flavor I's share had increased to 39% of the total brand's share leveling finally to 30% by the end of period twelve. Following through the data in Table 1, it can be seen that Flavor I contributed substantially to the increased success of Brand H and cannibalized on only one flavor within the brand to any significant degree (Flavor P). The introduction of Flavor J in period eleven should add additional information to the growth pattern of Brand H through time. Additional data is needed, however, for any extensive analysis on this new introduction.

In general, Brand H has increased its share by a significant amount through time by converting non-buyers, competitive buyers, and light buyers to the more loyal hard core of its buyers. This was accomplished primarily through the effectiveness of one specific line extension, Flavor I.



Analysis of Buyer Behavior Among Purchasers of New Flavor of a Mature Product Class

Three analyses were conducted among the new and repeat buyers of the new flavor. The first examined total new and repeat buyers in order to determine the extent of general market acceptance. The second examined market acceptance of the new flavor among buyers of the new flavor brand. The third examined market acceptance of the new flavor among previous buyers of the new flavor brand. Tables 2, 3, and 4 reflect the results of these analyses.

In order to gain a true picture of the market success of the new flavor relative to the buyer groups identified above, it was necessary to convert the purchases to volume. During the first year, the average purchase size among all buyers was 20.5 ounces. These rates did not change significantly in the second year. Thus, this stable rate was applied to both first and second year purchases in order to obtain an estimate of totaL volume for all buyers. Table 2 reflects these estimates. Tables 3 and 4 reflect a breakdown of this total volume by new and previous buyers of the new flavor brand.

In general, the new flavor introduction achieved market acceptance. The data in Table 2 are based on total market penetration rates, among National Consumer Panel households, of 37% with an excellent first repeat rate of 54%. Second and third repeat rates were 33% and 20% respectively.







It must be concluded, however, that the new buyer segment of the new flavor market contributed relatively little in terms of total buyers for the product. The total new buyers shown in Table 3 represent 33% of the total buyer category. Yet, when calculations were made, its total volume represented only 6% of the total new flavor purchase volume. The primary problem was a general lack of repeat volume.

The dominant market for the new flavor was centered in the previous buyers of the new flavor brand. These buyers represented 67% of the total buyers and 94% of the total volume. Analysis of Table 4 suggests an initial stocking up among first purchases of the new flavor with a slight purchase decline among repeaters. Users of new and repeat buying models often seek such indicators in their search for future problem areas leading to market decline (Cunningham, 1961). In the case of this product, it can be expected that other items in the line may have suffered during the stocking up process with the new flavor. This will be discussed in the following section.

The Source and Disposition of New Flavor Buyers

The Markov Process

The Markov process used in this study assumes that the effect on a given state i at condition t is flowing from the probability distribution of all states in the system at only this condition, t - 1, and the transitional probabilities of moving from those states to state i during the conditional interval t - 1 to t. The assumption is that the knowledge of the probability distribution of this conditional period is sufficient to predict the probability distribution of states at the present and future conditional periods at the asymptotic state.

Two type of information are essential to carry forward the analysis. They are (1) an initial probability distribution (brand share), and (2) the transition matrix. The probabilities of the transition matrix are nothing more than the probability that a certain manufacturer will retain its customers. It is computed from a gain-loss summary. Given the transition matrix and brand shares for a particular data period, future brand shares can be predicted, i.e., probabilities, by multiplying the brand shares with the transition matrix raised to higher powers.


Tables 5 through 7 illustrate flavor and brand loyalties of the new flavor buyers both prior and subsequent to their quarter of first purchase of the new flavor. The data in these tables were compared with predictions from the Markov model. The X goodness of fit test was used to evaluate how well the model predicted actual brand and flavor shares prior and subsequent to the quarter of first purchase of the model to actual brand share data

Table 5 reflects shares and predictions for five flavors and subsequent to the quarter of first purchase of the new flavor. The new flavor is an extension of Flavor A. The first impression upon reviewing the total flavor share information prior to the quarter of first purchase is one of stability of the shares period to period for the four quarters. It appears that the buyers may be in an asymptotic state.

To be in an asymptotic state should not be considered unusual given the maturity of the product class involved. The product has been in existence for decades and can definitely be classed in the maturity phase of the product life cycle.

The goodness of fit of the model reflects this stability. The transition matrix was fixeded at t - 1 the fourth buying period prior to the quarter of first purchase. The predictions of the model were quite close to the actual observed shares in all periods. Using equivalent volumes for shares, at x2 statistic was computed for all prior expected versus observed observations. The resulting 7.68 statistic did not exceed the .95 significance level at 19 degrees of freedom. The measures showed no significant differences.

The flavor share for the quarter of first purchase was not included in the predictions as it was considered an independent variable. The purpose in the analysis was to observe differences between expected and actual observations prior and subsequent to the insertion of the atypical variable. If the quarter of first purchase probabilities had been included in the predictions, the stability of the transition matrix would have been disturbed for subsequent periods. By using the quarter of first purchase as an independent variable it was possible to continue observing effects on the stable state. The greatest volatility on the disposition of buyers side of Table 5 appeared in those quarters adjacent or near adjacent to the quarter of first purchase. In terms of total periods, however, it can be observed that flavor shares fell near to ones observed in prior periods. The x2 text for goodness of fit in the subsequent periods reflected no significant differences between predicted and actual observations. Thus, in terms of the complete set of data the new flavor introduction caused little long-run volatility between flavors.

Table 6 reflects shares and predictions for seven brands prior and subsequent to the quarter of first purchase of the new flavor. Again, a relatively stable state can be observed between quarter to quarter purchase shares in the prior period. x2 for the total prior period was computed at 14.65 which was below the .95 significance level with 27 degrees of freedom.

Brand A contains the new flavor. It can be observed that from the fourth period prior to the fourth period subsequent to the quarter of first purchase, Brand A increased its market share by only one point. In terms of share, Brand A was unable to hold long run purchase volume equivalent to the volume generated in the quarter of first purchase. An examination of competitive brands will reveal a converse condition to Brand A in most instances.

In general, all competitive brands to Brand A, with the exception of Brand F, suffered share losses during the quarter of first purchase of the new flavor. All recovered, however, to levels approximating brand shares generated in the prior periods. The volatility that was generated by the new flavor introduction did cause a significant difference between the model's predictions and observed values during the subsequent period. x2 was computed at 75.32 indicating significance at the .999 level. The values that contributed primarily to the high statistic occurred in the first subsequent period to the period of first purchase. Differences in expected and observed shares between Brands B, D, and F, contributed heavily to these differences. The implication, again, is that short-run and not long-run changes were prevalent.







Table 7 reflects shares and predictions for seven brands, prior and subsequent to the quarter of first purchase of the new flavor, for total Flavor A purchases only. This view affords a direct comparison of the effects of Brand A's flavor extension relative to similar flavors within brands.

Although more variance between expected and observed shares can be observed in this table than in the previous two, a relatively good fit still existed in the prior observations. x2 was computed at 25.80, less than the .95 level of significance with 27 degrees of freedom. Yet, this statistic was the highest recorded for the prior observations.

A striking point in Table 7 is the fact that Brand A's share of Flavor A fell to its original prior brand share during the fourth quarter subsequent to the first purchase of the new flavor. It should also be recalled that Brand A gained an overall one share point (Table 6) by the end of the fourth subsequent period. This indicates that in terms of Flavors A's family of products the overall long-run effect for Brand A reflected no change. Yet the activity generated by the new flavor buyers.

Brand A did, however, maintain an above-average share of Flavor A for three periods subsequent to the first quarter of purchase. During these periods most competitive brands retained a below average market share. The X value for subsequent periods was the highest observed among the subsequent purchase categories. A 90.86 statistic was computed indicating significance at the .999 level for 27 degrees of freedom.


The hypothesis of this study suggested that long-run purchase volume of households loyal to a particular brand will not be affected by purchases of a new type introduced within that brand. It was discovered that the bulk of volume for the flavor did indeed come from previous loyal buyers of Brand A. This finding, coupled with the fact that Brand A was unable to hold, in the long-run, purchase volume generated by the new flavor introduction, reflected the general behavior pattern suggested in the hypothesis. The indication is that the loyal buyers did not increase their total consumption of the product in the long-run market growth for Brand A was not achieved.

A buyer may experience genuine discontent with a product implying the need for an extended search process for new products. Or, the need for temporary or intervening purchases may arise to relieve the boredom of routine related to content purchases of the same products. After one or two intervening purchases this type of buyer will normally return to previous purchase patterns. The inference from the data in this study is that, in general, purchasers of the new flavor behaved in this latter manner. An examination of the inferences resulting from the data should lend additional insight into why this conclusion has been reached.

A general stability of period to period brand shares among households was evident from the source and disposition analysis. This apparent asymptotic state was not considered unusual given the maturity of the product class and its buyers. The basis assumption for the application of the Markov model, i.e., that learning had occurred was, thus, fulfilled.

In the early periods of a product class, the gains and losses from brand to brand can be of high magnitude. But as buyers travel up the slope of the learning curve to its asymptote (equilibrium), gains and losses between brands become smaller and smaller until just before the asymptote they are infinitesimally small. This was the condition noted in the source and disposition analysis.

The asymptotic state of equilibrium may also reflect a leveling out of buyer penetration. In other words, a mature product class in a state of equilibrium relative to the matrix of transition probabilities between brand shares may also be in a state of equilibrium concerning the number of new buyers entering the market. With the same buyers purchasing the product from period to period, it should not be surprising that a rather static long-run market situation can develop given a relatively low level of product innovation.

Different decision strategies vary with respect to the position of the consumer on the product learning curve. The pure strategy of invariably choosing the previously purchased item may be expected to yield less disturbance in the consumer's overall routine. With respect to variability of behavior, mixed strategy of selecting different brands and/or flavors on different occasions may actually reinforce loyalty for the more frequently purchased item by reducing overall boredom for the product class in general.

The new flavor buyers were willing to give up income in the short-run by purchasing the new flavor in addition to other items within the same line. This behavior was not retained by Brand A in the long-run, however.

The buyers of the new flavor item originated primarily from the overall group of loyal buyers of the new flavor brand. The inference is that these buyers had reached an upper limit of product that they are willing to buy. Thus, in terms of building market share and volume, Brand A was limited to deriving long-run volume from outside the current market and from competitors who might also contain light buyers of Brand A's products. Since an equilibrium ceiling had been reached in the overall product category with reference to buyer penetration (penetration currently exceeds 80% of total U.S. households), the primary means of expansion for Brand A, via the new flavor introduction, was limited to competitive sources and occasional buyers. The findings point to a definite failure on the part of Brand A to exploit its one source of long-run volume. For example:

1. The new flavor introduction was not responsible for long-run volatility between the various flavors.

2. Brand A was unable to hold long-run purchase volume anywhere near the volume generated in the quarters of first purchase of the new flavor.

3. Conversely, competitive brands suffered only short-run losses in volume and share because of the introduction of the new flavor, and enjoyed a return to previous higher levels by the end of the fourth subsequent quarter after the periods of first purchase of the new flavor.

4. Brand A's share of Flavor A fell to its original prior level by the end of the fourth subsequent purchase period, although Brand A maintained an above-average share for this flavor for three of the subsequent quarters.

5. During the periods of first purchase of the new flavor, Brand A's other flavor items suffered share losses indicating proliferation from the new flavor item. Gradually, however, this share has been increasing in subsequent periods and is nearing prior levels.

6. A general pattern of substitution between various Brand A flavors was noted as subsequent volume totals returned to prior levels.

These findings reflect definite signs of buying households engaging in the "psychology of complication." In the words of Howard and Sheth:

A surprising phenomenon, we believe, occurs in many instances of frequently purchased products, such as grocery and personal care items. The buyer, after attaining routinization of his decision process, may find himself in too simple a situation. He is likely to feel monotony or boredom associated with such repetitive decision making. It is also very likely that he is satiated with even the most preferred brand. In both cases, he may consider all existing alternatives including the preferred brand to be unacceptable. He, therefore, feels a need to complicate his buying situation by considering new brands, and this process can be called the psychology of complication. The new situation causes him to search for identity with a new brand, and so he begins again to simplify in the manner described earlier. Thus with a frequently purchased item, buying is a continuing process with its up and downs in terms of information seeking, analogous to the familiar cyclical fluctuations in economic activity (Howard 6 Sheth, 1969).

In general, the findings indicate a group of buyers unwilling to purchase additional product in the long-run, but willing to vary their behavior in the short-run, perhaps because of the reasons noted by Howard and Sheth. Identification of where the buyer is located relative to the "simplification-complication" process in terms of information seeking is clearly very important to the marketing manager. If it can be determined, for example, that a group of buyers are at a level of routinization where satiation or monotony with regular brand buying is apparent, then a new brand or flavor could possible be introduced that would provide the needed source of change that will ultimately build a greater share of market for the innovator. What is difficult to determine, however, is the extent of the discontent with existing routinization. If, as in the case of the product under review in this study, it is a short-run phenomenon that does not have the strength to hold switching buyers from other brands then long-run effects will more than likely be negligible in terms of increased overall sales of product. A true innovation, on the other hand, could satisfy the need for new product searching, and thus, become a vehicle toward increased market growth.


Brown, G. Brand Loyalty--Fact or Fiction. Advertising Age. 1962, 23, 53-55.

Campbell, D. T. 4 Stanley, J. C. Experimental Designs for Research. Chicago: Rand, McNally, 1968.

Cunningham, R. Customer Loyalty to Store and Brand. Harvard Business Review, 1961, 39, 127-137.

Fourt, L. A. 4 Woodlock, J. W. Early Prediction of Market Success For a New Grocery Product. Journal of Marketing, 1960, 25, 31-38.

Frank, R. E. Brand Choice as a Probability Process. Journal of Business, 1962, 35, 35-56.

Howard, J. W. 4 Sheth, J. N. The Theory of Buyer Behavior. New York: John Wiley & Sons, 1969, 27-28.

Howard, R. A. Dynamic Inference. Operations Research, 1965, 13, 35-42.

Kuehn, A. Consumer Brand Choice as a Learning Process. Journal of Advertising Research, 1962, 2, 10-17.

Massy, W. F. A Stochastic Evolutionary Adoption Model for Evaluating New Products. Working Paper No. 95, Graduate School of Business, Stanford University, 1967.

Market Research Corporation of America. The National Consumer Panel. Unpublished Monograph, 1970.

Parfitt, J. H. 4 Collins, B. J. [. The User of Consumer Panels for Brand-Share -Predictions. Journal of Marketing, 1968, 5, 131-145.



J. Taylor Sims, University of South Carolina


SV - Proceedings of the Third Annual Conference of the Association for Consumer Research | 1972

Share Proceeding

Featured papers

See More


Saving for Experiences Versus Material Goods

Grant E. Donnelly, Harvard Business School, USA
Masha Ksendzova, Boston University, USA
Michael Norton, Harvard Business School, USA

Read More


K1. The Impact of Moral Violation and Advertising Appeals on Brand Attitude

Chunya Xie, Renmin University of China
En-Chung Chang, Renmin University of China
Beixi Wen, Renmin University of China

Read More


The Mystique of Masculine and Feminine Choices: How Aversive Feelings Underlie Preferences

Niusha Jones, University of North Texas
Blair Kidwell, University of North Texas

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.