Determine whether weak historical coupling reflects merchandising practice or shopper behavior

Determine whether weak joint upsell–cross-sell coupling observed in historical retail receipts reflects the retailer’s existing merchandising practices rather than the absence of an underlying customer-response opportunity, using randomized joint-offer experiments to identify the causal response to displaying coupled offers.

Background

The paper argues that receipt data record customer behavior under an existing offer regime. If a retailer rarely presents premium products together with complementary add-ons, the resulting historical data may show little coupling even when customers would respond positively to jointly displayed offers. The paper’s observational analyses cannot distinguish whether weak measured coupling reflects weak underlying customer response or merely limited exposure to joint offers.

Randomized joint-offer experiments are proposed as the means to refresh the data with logged display and acceptance events and to determine whether coupling emerges when upsell and cross-sell offers are actually presented together. This question is unresolved in the paper because its empirical results are based primarily on calibrated simulations and observational receipt associations rather than a live grocery field experiment.