Performance of judgmental - statistical forecast combination strategies under product-market configurations

Combining independent judgmental and statistical forecasts is one of the prevalent methods to increase forecast accuracy in the supply chain. Several strategies for combining them have been proposed in the literature. However, little guidance is available on selecting the proper strategy given various forecasting conditions. This study investigated the performance of four forecast combination strategies, namely (i) judgmental adjustment, (ii) statistical correction, (iii) mechanical combination, and (iv) judgmental input, under four product-market configurations as in Ansoff's matrix. We used archival data from a global pharmaceutical company as the basis for evaluation. The result suggests that combining independent statistical and judgmental forecasts can be beneficial for organizations. However, not all combinations strategies are efficient across the product-market configurations. Amongst the strategies, only the judgmental input and mechanical combination strategies can give consistent accuracy improvement over the baseline. In contrast, judgmental adjustment and statistical correction strategies only work in the environment where the forecasters have high experience with the products and markets. The empirical findings from the study provide some managerial implications for operational forecasting.