Using online data for offline decisions

Even though the online channel is generally not bounded by space, much insight can be derived from examining the variation in online patronage across retail space. In this paper, we propose the use of a Bayesian approach utilising the integrated nested Laplace approximation (INLA) within its framework to estimate the risk ratios as well as probabilities of online patronage. To this end, we used publicly available data (such as household access to broadband connection, population figures and business location data) in combination with online transaction data from a focal retailer to estimate and identify which locations (i.e., postcodes) have the highest probabilities of patronage from online customers of the retailer. The three covariates adjusted for were presence of a competitor, average income at the postcode level and internet access at the postcode level; of these, internet access was found to be the most significant influencing factor on online patronage of the retailer. Overall, the resultant spatial maps provide useful predictions of how the estimated risk ratios and exceedance probabilities vary across the geographic market and the implication on opening a new offline outlet and targeting customers are discussed.