Market share forecasting: An empirical comparison of artificial neural networks and multinomial logit model

We empirically compare the forecasting ability of artificial neural network (ANN) with multinomial logit model (MNL) in the context of frequently purchased grocery products for a retailer. Using scanner data on three grocery product categories we find that performance of ANN compares favorably to MNL in forecasting brand shares. We test the sensitivity of the forecasting error in the two approaches to the length of the estimation period and the clustering of households which is used to define homogenous segments of households. We find the results to be robust to these variations. We also derive a few empirical propositions regarding performance of ANN and MNL from our analysis. The results are consistent with those in Kumar, Rao and Soni (1995) and suggest that although neural networks suffer from interpretability problem, they are a useful method to forecast brand shares in grocery product categories where large amounts of scanner data are readily available.