Retail Promotion Optimization Problem (POP)


Retail Promotion Optimization Problem (POP)

Principal Investigator

Professor Georgia Perakis

MIT, Sloan School

Oracle Fellowship Recipient

Jeremy Kalas, Lennart Baardman, Maxime Cohen, Zachary Leung

Oracle Principal Investigator

Kiran Panchamgam
Sajith Vijayan


An important and challenging problem for the retail industry is to determine the optimal number, depth and timing of price promotions for a set of items. For instance, if we understand the demand for a set of items, then what price-promotions, offered when and how often, will result in the most profit for those items? The solution to POP is known to be computationally difficult and expensive. We have proposed a fast, efficient and near-optimal solution. At this stage, we propose to build a scalable and near-optimal model for multiple items that will be evaluated for (1) how well the model scales for a tier-1 retailer, (2) solution quality compared to the optimal solutions, (3) performance and results of the model on real customer datasets. We plan to incorporate the customer dimension to POP – this will yield promotions targeted to specific consumers, which is the holy grail of retail promotions. With the availability of many new structured and unstructured data systems, customer information can provide valuable information for retailer in devising promotional offers. We will also be conducting pilot with an Oracle RGBU client to demonstrate the value of recommendations from our models by going live in this phase.