Abstract
Expired products from retail stores account for a high proportion of waste in the supply chain. While some retail stores offer discounts to attract sales on the nearly expired but healthy products, others discard them to prevent damage to their brand image as sellers of nearly expired products. Sometimes, despite this discount, these products remain unsold and eventually expire. This could mean that there is a disproportionate value assigned by the store and the expectations of consumer or shopper. This unaligned pricing discount culminates in avoidable waste and revenue loss in the retail store. This research aims at investigating factors for predicting consumer preferred discounts of nearly expired products in retail stores. In this study, with 17 questions in the questionnaire, 205 valid responses from 217 random retail store shoppers were analysed descriptively, by correlation and regression. The Chi-squared test was used to identify seven uncorrelated predictor variables for the predictive modelling (preferred brands, use-rate, brand loyalty, preferred discounts, perceived discounts, perceived value, perceived safe period and safety bias). Binary logistic regression was employed for the purpose of forecasting and categorizing shoppers within the scope of the dependent variable category, with the goal of comprehending the factors that bear the greatest likelihood of exerting influence on the choices made by consumers when procuring these products at a marked down rate that is deemed desirable. The resulting model is statistically significant, indicating that it can distinguish between the categories of the dependent variable. The specificity and sensitivity of the model were 72.0% and 84.4% respectively, representing an overall prediction percentage of 79.4%, a 32.8% improvement on the model without predictor variables. The model also met Hosmer and Lemeshow’s goodness of fit criteria and explained between 32.8% (Cox-Snell R-squared) and 44.3% (Nagelkerke R-squared) of variance of the dependent variable. It predicted a higher likelihood of more shoppers buying the nearly expired products at a substantial discount than those unwilling to buy. Meanwhile, five predictor variables (preferred brands, use-rate, brand loyalty, perceived value, perceived safe period) significantly contributed to the model and about 4% or 8 entries in the dataset returned as outliers but were left in the model.