Postprint version. Published in Journal of Retailing, Volume 71, Issue 2, Summer July 1, 1995, pages 153-171. Copyright © 1995 Published by Elsevier Science Inc. All rights reserved. The definitive version can be found online at http://dx.doi.org/10.1016/0022-4359(95)90005-5.
A shelf management model was developed to assist retailers with the decision of which products to stock and how much space to allocate to those products. Due to the non-linearities in the formulation a closed form solution is not possible. Borin, et al. develop a search heuristic based on simulated annealing and compare the solution against a known optimum. A barrier to the use of such models is the fact that managers typically do not have access to error-free estimates of the parameters required for the model construction (shelf elasticities, search loyalty, and consumer preferences). In this article we analyze the degree of error that may be introduced into estimates of the parameters before the model yields assortments and shelf allocations that are inferior to those produced by the merchandising rule of thumb, SHARE-OF-SHELF = share-of-sales. The results indicate that judgmental estimates of parameters can vary by as much as 50 percent and still make application of the model useful.