Regularity in individual shopping trips: implications for duration models in marketing

Most models for purchase-timing behavior of households do not take into account that many households have regular and non-shopping days. We propose a statistical model for purchase timing that exploits information on the shopping days of households. The model is formulated in a counting process framework that counts the recurrent purchases for each household over (calendar) time. In our empirical application of yogurt and detergent purchases from the ERIM1 database, we show that calendar time effects and regular and non-shopping days are important features to include in models for purchase-timing behavior. We find, for instance, that for these product categories the probability of purchasing is 50-60% higher on Saturdays and 70% higher on regular shopping days. We highlight the managerial implications of these model features by simulating some promotional actions.