Sales promotions and product enhancements are commonly expected to increase a brand's sales, when they do not negatively impact its utility and cost. That is, the purchase probability of consumers who find the promotion or additional feature attractive will increase, whereas the purchase likelihood of other consumers will not be affected. In contrast, we propose that consumers, who perceive a new feature or promotion as providing little or no value, will be less likely to purchase the enhanced brand even when the added feature clearly does not diminish the value of the brand. Thus, a new product feature or promotion may decrease a brand's overall choice probability when the segment of consumers who perceive it as providing little or no value is large compared to the segment that finds the feature attractive. This prediction was supported in three studies using actual promotions that have been employed in the marketplace (e.g., a Doughboy Collector's Plate that buyers of Pillsbury cake mix had the option to purchase for $6.19). We examined five alternative explanations for this effect. The results suggest that, when consumers are uncertain about the values of products and about their preferences, such features and premiums provide reasons against buying the promoted brands and are seen as susceptible to criticism. We discuss the theoretical and practical implications of the findings for segmentation, product, promotional, and pricing strategies.brand choice, buyer behavior, product policy, promotion

No abstract

Weillustrate the application of BUGS, a Bayesian computer program, with two examples. The algorithm used in the program is a popular Markov chain Monte Carlo procedure called Gibbs sampling. Bayesian analysis based on simulation has been applied to a wide range of complex problems (Gilks, Richardson, & Spiegelhalter, 1996).The availability of a general purpose program like BUGSshould facilitate important applications of Bayesian inference in psychological research.More than three decades ago, Edwards, Lindman, and Savage (1963) published a ground-breaking paper on Bayesian statistical inference in psychological research. Bayesian data analysis, however, rarely appears in mainstream journals in the behavioral sciences.One of the major limiting factors in applying Bayesian methods is computational difficulty. Given data, Bayesian inference proceeds from prior distributions to posterior distributions ofthe model parameters. In real applications involving many parameters, Bayesian computation requires the evaluation of complex, high-dimensional integrals. Because commonly used statistical packages cannot accommodate these difficult calculations, practitioners of Bayesian inference must rely on their ability to adapt specialized routines to the problem at hand. In our view, this lack of user-friendly software has hampered the adoption of Bayesian statistics by researchers in the social and behavioral sciences.Inrecent years, the Bayesian computational problem has been addressed by taking a sampling approach. References on a number of simulation tools for drawing samples from the posterior distribution are found in Tanner (1993). So remarkably simple are these computational algorithms, they can be described injust a few lines in any computer language. With the wide availability of highspeed computing, simulation-based methods have become increasingly important in data analysis.The purpose of this paper is to give an accessible account of the sampling-based Bayesian computation through a language called BUGS. Unlike other computer programs compiled by Press (1989), BUGS provides an integrated computing platform for Bayesian inference. Programming BUGS is rather similar to implementing statistical routines in an object-oriented language like S (Chambers

No abstract

No abstract