Advertising Exposure, Loyalty, and Brand Purchase: A Two-Stage Model of Choice
Jan 1, 1987
Influence of repetitive advertising on purchase behavior.
Type of Report
Combines experimental research on attitudinal response to ad exposure with econometric research on market share response to ad expenditures.
Threefold: (1) to provide insights on whether repetitive advertising is effective; (2) to help identify those consumer segments most responsive to repetitive advertising; (3) to determine how ad response functions differ between established brands and market newcomers.
Reviews behavioral research on repetitive advertising exposure in order to develop relevant hypotheses; uses choice modeling to develop a test of these hypotheses; tests the hypotheses with scanner data; the test relies on a two-stage model of consumer choice. Both within-and-between-subject analysis and within-and-between-brand analysis are performed, minimizing problems of collinearity.
Advertising decision-makers, advertising researchers, and strategic marketers.
Tests of the hypotheses indicate that, within the limitations dictated by the research design:
- Advertising exposure seems to reinforce preferences more than motivate brand choices
- The effect of advertising appears non-linear, with an optimum between two and three exposures per week.
- This response is significantly mediated by consumer brand familiarity. Ads for familiar brands are more effective than those for unfamiliar brands. Ads for unused brands require much higher levels of exposure to affect response.
- Overall, the effect of advertising, relative to other influences measured in the study, is small. Brand choice is most strongly influenced by brand loyalty (brand preference), followed by coupons, features, displays, and price. In this study, advertising has no effect on brand choice. Quantity purchased is most strongly influenced by volume loyalty, followed by brand preference, price, and advertising. Coupons, features, and displays have no impact on quantity purchased.
- Controls for the households’ television viewing, inventory level, and exposure to various ad-flights did not invalidate the basic results. Checks indicated that multicollinearity was not a serious problem. Exclusion of outliers did not affect the non-linearity of the ad-response function.
Interpretation and Implications
The importance of loyalty as a significant moderator of the effects of ad-exposure was re-emphasized by this research: Buyers respond more strongly and quickly to brands to which they are loyal.
More interestingly, the difference in results across stages of the analysis is unexpected and appears counter-intuitive: Advertising seems to affect the quantity purchased more than the brand chosen. There is some earlier research which indicates that advertising exposure is more effective in increasing volume purchased than in promoting brand switching. Additionally, the findings in the present study may not be inconsistent with theory: If advertising affects loyal buyers more than non-loyal buyers, then it should be expected to affect the quantity purchased of the preferred brand rather than which brand is chosen.
The differential effects of the other marketing mix variables are more intuitive. Price affects the quantity purchased more than brand choice, but coupons, displays, and features do primarily affect brand choice, which may be their intended function. Furthermore, the more visible of these variables should have a greater impact on brand choice. Thus, features in local media are likely to be the most effective, coupons in select media the next most effective; in-store display are likely to attract only in-store buyers, and price changes by themselves are probably the least noticed.
Overall, advertising is not the strongest determinant of purchase behavior. Without question, loyalty is the strongest determinant: brand loyalty at the brand-choice stage, and volume loyalty at the quantity purchase stage. All other marketing mix variables are also more effective than advertising. The effects of the loyalty variables are not merely definitional all the loyalty variables are carefully defined on behavior that occurs prior to that being predicted here. The strong effect of loyalty indicates that the bulk of purchase behavior is characterized by inertia or predetermined preferences.
Advertising is effective in increasing the volume purchased by loyal buyers, but not effective in winning new buyers. For loyal buyers, high levels of exposure per week may be unproductive due to a leveling off in effectiveness. Given the high effectiveness of the other marketing variables, especially in brand choice, a reasonable strategy would be to promote trial with displays, features, and coupons, and then motivate heavier purchases with advertising.
Professor Tellis indicates:
“For several decades, researchers have tried to analyze the effects of advertising, both because of the huge expenditures involved and because of the controversy surrounding advertising’s role. While analyses of market data have provided some answers, they have been often criticized for their aggregate nature. Laboratory studies have been very insightful, but they too have been criticized. By using cable-scanner data, this study suggests a new, powerful approach that combines the disaggregate nature of laboratory studies with the relevance of market studies. The analysis suggests that advertising effects are generally weak. It offers specific insights about repetitive advertising exposure, and it throws light on (1) the relationship between advertising and brand loyalty and (2) advertising’s role as a barrier to market entry.
“While the latter issue will probably never be wholly resolved, this research suggests that advertising does not hinder entry. Advertising is one of the least-effective determinants of purchase behavior, with consequently limited power either to deter or facilitate entry.
“Even though the results of the study are consistent with the theory and literature, conclusions must be drawn cautiously. The sample is limited to one product category in one market. While such scanner data is not known to have serious biases, confidence in the results must await replications.
“Nor does the study account for the content of advertising. If it did, it would revolutionize testing of advertising effects.
“Finally, I was unable to test how the model would perform with products in the early phases of the life cycle. These are all promising avenues for future research.”
About the Authors
Gerard J. Tellis is Assistant Professor of Marketing at the University of Iowa.
This resource is not available online.