Forecasting Product Life Cycle Curves: Practical Approach and Empirical Analysis

Abstract

We present an approach to fit product life cycle (PLC) curves from historical customer order data and use them to forecast customer orders of ready-to-launch new products that are similar to past products. We propose three families of curves to fit the PLC: the BASS diffusion curves, polynomial curves and piecewise- linear curves. Using a large data set (133 products) of customer orders for short lifecycle products, we compare goodness-of-fit and complexity for these families of curves. Our key empirical findings from PLC fitting are that simple, piecewise-linear curves are very effective at fitting the PLC in our data set, and the products in our data rarely have a “mature” or “sustain” phase often represented in traditional PLC curves. Using time-series clustering techniques, we cluster the fitted PLC curves into several representative curves and use these curves to generate forecasts for the products in our data set. Our forecasts result in absolute errors approximately 9% lower than the company forecasts.