Resource Allocation Heuristics for Unknown Sales Response Functions with Additive Disturbances

We develop an exploration-exploitation algorithm which solves the allocation of a fixed resource (e.g., a budget, a sales force size, etc.) to several units (e.g., sales districts, customer groups, etc.) with the objective to attain maximum sales. This algorithm does not require knowledge of the form of the sales response function and is also able cope with additive random disturbances. The latter as a rule are a component of sales response functions estimated by econometric methods. We compare the algorithm to three rules of thumb which in practice are often used for this allocation problem. The comparison is based on a Monte Carlo simulation for five replications of 192 experimental constellations, which are obtained from four function types, four procedures (i.e., the three rules of thumb and the algorithm), similar/varied elasticities, similar/varied saturations, and three error levels. A statistical analysis of the simulation results shows that the algorithm performs better than the three rules of thumb if the objective consists in maximizing sales across several periods. We also mention several more general marketing decision problems which could be solved by appropriate modifications of the algorithm presented.