Data driven supply allocation to individual customers considering forecast bias

We propose a data-driven allocation planning approach, which is designed for use in advanced planning systems as they are widely used in industrial environments. The approach exploits increasingly available data on individual customers and products by allocating supply on a highly granular level at high planning frequencies. It counteracts rationing gaming by customers, which we assume to be the reason for demand forecast biases. We create an incentive for truthful forecasting by not only allocating supply based on customer profitability but also based on forecast bias. In the long term, this approach gives access to a profit potential and an on-time service level increase. In the short term, however, setting such an incentive does not only have a positive impact on service levels but also leads to a decline in profits. Our methodology quantifies this trade off providing decision support for determining the extent to which the forecast bias should affect the allocation. In a numerical study based on the semiconductor industry, we demonstrate that the approach has a large long-term profit potential while having limited effect on short-term profits for significant service level incentives. The analysis further shows that the allocation efficiency increases with the granularity level and the predictive quality of the available data.