OPTIMIZING THE DELIVERY PROMISE IN E-COMMERCE WITH DETERMINISTIC BINARY PROGRAMMING METHODS

The definition of the delivery promise plays an important role in e-commerce performance, as customers expect faster and faster deliveries. Delays can hinder the shopping experience, while longer promised delivery times can reduce sales competitiveness. This work presents an optimization model based on historical data to determine the most appropriate delivery promise for different routes, considering a balance between conservative policies (which can lead to lost sales) and bold policies (which can cause a bad experience for buyers). The problem is modeled as binary linear programming and solved with the HiGHS and CBC algorithms, implemented in Python through the SciPy and PuLP libraries. Both methods achieved optimal solutions for instances with more than 8,000 variables in a few seconds. Compared to a trivial approach (without promise protections) and a simple heuristic (with windows on the routes with the highest percentage of late deliveries until the limit imposed by the restriction is reached), the proposed method offered relevant practical gains for real logistics and sales operations.