Predicting new product adoption using Bayesian truth serum

This article explores whether a modified version of recently proposed survey mechanism called a Bayesian Truth Serum (BTS), which is designed to provide more accurate subjective survey results, can be used to improve new product adoption forecasts. To validate the test, a survey is administered before the introduction of a new product, in which respondents are asked about their adoption of a product that is not yet available, and the survey results are compared to actual adoption once these products are available. The results indicate that prediction performance of the modified BTS model is improved by up to 36 per cent over the standard reference model, although both models show opportunities for additional improvement.