A Multi-Product Individual-Level Model for New Product Sales: Forecasting and Insights

We develop a novel approach for modeling new product trial and early repeat purchase behavior, and we apply this approach in the context of consumer packaged goods. Our approach takes advantage of the cross-individual, cross-product, cross-time data that is increasingly available from retail customer relationship management programs as well as research panels. It enables us to account for differences in consumers’ intrinsic preferences for new products as well as for differences in their responsiveness to marketing variables, during both trial and early repeat purchases. By leveraging these uncovered individual differences, we attempt to achieve three goals. First, we aim to improve the accuracy of post-launch sales forecasts based on data from a period that can be as short as two to three months, tapping into the fact that each individual trial or early repeat purchase observed during the post-launch period sends a different signal about the new product’s sales potential. Second, we aim to provide more informative diagnostics for managers to act upon. (e.g., how to re-position the new product or target it for a particular consumer segment). Finally, we aim to investigate potential empirical regularities regarding consumer responses to new products (e.g., potential differences in responsiveness to marketing variables between early and late triers).