We consider a retailer adopting a money-back-guaranteed (MBG) sales policy, which allows customers to return products that do not meet their expectations to the retailer for a full or partial refund. We assume that the market consists of infinitesimal customers with individual and uncertain valuations for the product.

Since customers have limited time and capacity to learn about the product’s true value, they optimally determine how much effort to spend to acquire and process value-revealing information while making their purchase decision. We model the choice behaviour of such rationally inattentive as generalized multinomial logit (GMNL), which depends not only on the product’s true value but also on the prior belief of the customer and the cost of information. In turn, we formulate the seller’s problem, facing such customers, using a stochastic optimization model to solve for the optimal product price and refund amount.

We characterize the customer’s optimal purchase strategy when the seller does not offer MBG as well as when he sells with MBG, and identify the structure of the optimal price and refund in both cases. We also run extensive numerical experiments to substantiate our analytical findings and generate additional insights on the impact of key problem parameters, such as information cost, level of uncertainty in product value, and customer’ prior beliefs.