Setting Quality Expectations When Entering a Market: What Should the Promise Be?

Successful new products generate both trial and repeat purchases. While trial is based on expectations about product quality, repeat depends largely on satisfaction as assessed by the gap between observed and expected quality. As a consequence, companies make a tradeoff between setting expectations too high, resulting in strong trial but little repeat business, and too low, which enhances repeat business but leads to low initial trial rates. Here we focus on determining the “just right” level of advertised quality to maximize profits. The contribution of this paper is three fold. First, we explore a firm’s optimal advertised quality as well as average actual quality and price. Second, we delineate conditions under which firms have an incentive to overstate quality, i.e., deceptively advertise, in an application using data from a field study involving a firm entering a new geographical market. Third, we empirically test some of our model results using data from deceptive advertising cases at the Federal Trade Commission during the period mid-1996-2002 as well as from an experiment with MBAs (future managers). In our model, first period sales are driven by quality expectations and price. Quality expectations in turn depend on both the advertised and actual levels of quality. Subsequent period sales then depend on price, revised expectations, and first period satisfaction. The model allows for (1) a non-linear relation between actual and expected quality gap and satisfaction, (2) decisions concerning advertised quality as well as (3) average actual quality and price. Several results are derived from the model. Based on FTC cases, unknown companies were more likely to overstate of quality and the level of deceptive advertising was higher among companies whose customers were more satisfied. In an experiment, future managers (MBAs) demonstrate a slight tendency to overstate quality and were directionally consistent with most of the model results. With one exception, the experiment also provided additional support for the other model results.