Pricing with Bandits in the Long-tail: The Role of Competitive Monitoring

Most e-commerce retailers offer a long-tail of very low demand products. Individ- ually, these items may have low sales but collectively they are critical to the overall e-commerce business model. Because of their minimal sales, pricing is a constant chal- lenge. The academic literature has considered price exploration as a primary source of information for price adjustments, but this approach may be insufficient in low demand situations. In this paper, we propose a bandit algorithm for long-tail products that is informed by both monitoring competitor prices and price exploration. We show that monitoring a larger competitor can inform pricing of long-tail products. Our bandit model is motivated by a unique dataset from a large e-commerce firm that regularly monitors competitor prices. We illustrate consistency between the bandit assumptions and our empirical evidence. We then show that three predictions from the bandit model are consistent with our empirical data.