Market Share Predictions: A New Model with Rating-Based Conjoint Analysis

Abstract

Conjoint Analysis (CA) is a technique heavily used by industry in support of product development, pricing and positioning, and market share predictions. This generic term CA encompasses a variety of experimental protocols and estimation models (e.g. rating-based or choice-based), as well as several probabilistic models for predicting market share. As for the rating conjoint, existing probabilistic models from the literature cannot be considered as reliable because they suffer from the Independence of Irrelevant Alternatives (IIA) property, in addition to depending on an arbitrary rating scale selected by the experimenter. In this article, after a brief overview of CA and of models used for market share predictions, we propose a new model for market share predictions, RFC-BOLSE, which avoids the IIA problem, yields convergent results for different rating scales, and outputs predictions that match regression reliability. The model is described in details and simulations and a case study on truck tyres will illustrate the reliability of RFC-BOLSE.