Prof. Dr. h.c. Hasso Plattner

Dynamic Pricing on Amazon

Dynamic pricing competition on modern market platforms such as Amazon Marketplace or eBay is constantly growing. Effective automated repricing systems become increasingly important for practitioners. This, however, is a highly challenging task, because offers are multi-dimensional, demand information is limited, and competitors’ price reactions are not known.

The Challenge

In this project, we partner with a top 10 seller for used books on Amazon in Germany. Our partner has an inventory of over 100,000 distinct books (ISBN), each with multiple items. The challenge is to extract as much profit as possible from a given number of books (inventory level) in a reasonable amount of time.


The pricing strategy of our project partner is characterized by a rule-based system that has been developed over the past years by carefully adjusting rules to lessons learned from selling books on Amazon. Our goal was to develop an improved data-driven pricing strategy that is based on a theoretical model.

Data-Driven Repricing Model

Process Overview - Prediction Pipeline
Process Overview - Prediction Pipeline
RecSys 2016 - Poster (How To Survive Dynamic Pricing Competition in E-commerce)
RecSys 2016 - Poster

To be able to set up a dynamic model in order to compute optimized prices, we estimated sales probabilities. We used a logistic regression approach to quantify how offer prices and specific market situations affect sales. We consider up to 10 offer dimensions (e.g., price, quality, ratings, feedback count, shipping time) per competitor for a particular market situation.


The data set that we used for the regression analysis contains both the requested market situations from Amazon as well as our seller's own data (offers, sales, and inventory). Our partner requests market situations for each offered book roughly every two hours (i.e., >20 M market situations per month which result in >140 M single competitor observations per month). We used thirty customized features, e.g., the price rank of our offer price within the competitors' prices. As a result, we were able to predict sales probabilities for any offer price and for any market situation. Using efficient solution techniques, we are able to compute optimized prices for current market situations.


The application of our dynamic pricing strategy works as follows: First, we observe current market situations for our products, then we calculate optimized prices according to the model, and finally adjust prices on the market platform. This procedure is repeated every two hours or in case of changing market situations. This way, our strategy is able to respond immediately to new situations as prices can be adjusted in milliseconds. Moreover, the new incoming sales observations are used to further improve the strategy by estimating demand more accurately.


Sales results over several weeks show that our strategy outperforms the established rule-based strategy of our experienced partner regarding profitability and speed of sales, resulting in a profit increase of more than 20%.


Contact: Rainer Schlosser, Martin Boissier


Related Publications

  • Schlosser, R., Boissier, M., Schober, A., Uflacker, M.: How To Survive Dynamic Pricing Competition in E-commerce. Proceedings of the Poster Track of the 10th ACM Conference on Recommender Systems (RecSys 2016), Boston, USA, September 17, 2016 (2016).