A Bayesian approach to the evaluation of stochastic signals of product quality

This paper explores consumers' use of stochastic signals of product quality from a Bayesian perspective. Stochastic signals are implicit in many of the analyses on signal usage in the marketing literature and yet formal analysis of their use has been relatively limited. The basis for our analysis is a seminal paper by Winkler on combining experts' forecasts into a single consensus distribution, together with the resulting literature to which this has given rise. This is a substantial body of literature and represents a readily available fund of results which can be used within the marketing context. Hence, it is shown that the weights on different signals will be inversely related to the variance of the signal and need not be convex. It is also shown that for repeatedly purchased goods, signals will play a decreasing role as the consumer gains experience of the good.