We consider a retailer with an online channel. Items sold through the channel are stored either at the retailer's central warehouse, with fast delivery, or at its suppliers, with slower delivery. We use data from a European interior design and furniture retailer consisting of daily sales transactions and inventory data covering 18 months. We use a quasi-natural experiment—random transitions of products in and out of inventory—to estimate the causal effect of in-house storage on sales. Our results show a statistically significant increase in sales of, on average, 156% associated with storing the product in-house rather than at the supplier. This effect is negatively moderated by the price of the product. We use these insights to optimize the retailer's warehouse assortment. Our results suggest a possible 7.5% improvement in gross profits given the current inventory budget.