Abstract
For e-commerce companies to assess how best to invest in improving delivery times, it is important to understand how improving delivery times affects customer demand. In collaboration with a business-to-business (B2B) e-commerce company, we study how the promised delivery time in a quote affects the customer's purchase probability. We use observational and experimental data from our partner, with quote-level variation in promised delivery times. This allows us to estimate demand as a function of promised delivery time after flexibly controlling for customer, product, and vendor differences. We find that there is a large, robust effect of promised delivery time on demand: a 1-day improvement in promised delivery time increases demand by 1.82\%, equivalent to a 2.21\% discount, comparable to prior findings in business-to-consumer (B2C) retail contexts. Interestingly, using semiparametric analysis, we find that this effect is non-linear: demand is not sensitive to promised delivery times of under a week, but drops quickly when delivery is expected to take over a week. We show that the largest improvements in demand are to be gained from investing in measures that can reduce the ``long tail'' of slow deliveries (e.g., avoiding stockouts and processing delays, ensuring geographic coverage of fulfillment centers) rather than reducing the delivery time of products that are already relatively fast to deliver.