Improving Prediction in Conjoint Measurement

Conjoint measurement generally is used to assess utilities for a few discrete attribute levels. An important practical application of such utility assessments is the prediction of the overall value of alternative new product concepts. If these product concepts are defined as combinations of attribute levels other than the original discrete levels presented to the respondents, linear interpolation commonly is used to predict utilities for the new attribute levels. However, linear interpolation may lead to poor predictions and the consequent choice of the wrong new product to introduce in the market. Therefore, the authors present an alternative to linear interpolation: the estimation of a utility function. An analytical procedure and some numerical results are provided to indicate the superiority of the utility function approach for the prediction of attribute utilities. The market share implications of improved utility prediction are discussed.