Forecasting the forecastability quotient for inventory management

Abstract

This research develops and empirically tests a model for estimating the economic advantage of using a time phased order point system (TPOP) with time series forecasting rather than a simple reorder point system in an independent demand inventory management context. We define the forecastability quotient (Q) to support this economic analysis. We implement TPOP in our empirical analysis via double exponential smoothing with a damped trend, and implement ROP through a simple moving average.Our empirical study of a large dataset of time series from a Fortune 100 firm found that Q in the holdout sample can be predicted using just three variables from the estimation sample. Surprisingly, many highly touted time series metrics (e.g., the coefficient of variation and approximate entropy) and forecast accuracy metrics (e.g., the mean absolute percentage error) were not good predictors of Q. We then validated this model on four additional datasets. This research contributes both to the research literature and to managers who need to decide whether an independent demand item should be managed with a TPOP or reorder point system.