Predicting the effects of physician-directed promotion on prescription yield and sales uptake using neural networks

Abstract

In this paper, the potential of neural networks to predict the effects of promotion on prescription yield and sales uptake is explored. The motivation for this work derives from the notion that marketing strategies guided by predictions of the effects of promotion help in identifying marketing expenditure that needs augmentation, thereby generating a healthier return on investment. Five promotional spending descriptors for an antibiotic product are selected as descriptive data (network inputs) to neural networks with prescription yield and sales uptake volumes as the outcomes (network outputs). A back-propagation neural network with five hidden elements (commonly called neurons) compares favourably with the benchmark obtained by multiple linear regressions in terms of r 2 and mean absolute percentage error (MAPE). Further, a 'leave-one-out' cross-validation revealed that the optimal neural network is able to correctly predict the effects of promotion more than 95 per cent of the time (r² &<0.825, MAPE<4.3 per cent), suggesting the general utility of this approach as a potential tool for predicting the effects of physician-directed promotion. [ABSTRACT FROM AUTHOR]