Simple and Effective Autonomous Dynamic Pricing

What is it about?

One of the key technical challenges in dynamic pricing is due to its computational complexity. When the number of products becomes extremely large (which is not uncommon in e-commerce setting), even solving a deterministic price optimization (i.e., by ignoring demand uncertainties) once is already time-consuming. This paper addresses this issue by developing an autonomous dynamic pricing algorithm that only requires solving a deterministic optimization once and automatically adjusts the prices over time as demands are realized. The proposed adjustment does not require any re-optimization and only involves simple algebra (in fact, it can be implemented in a "blink of an eye").

Featured Image

Why is it important?

The result in this paper shows that, in some settings, dynamic pricing is surprisingly easy to implement. Although solving for the dynamic optimal policy is intractable, a heuristic that requires a minimal computational effort exists and can be theoretically shown to have a near-optimal performance. This has an important implication as it potentially opens up application of dynamic pricing to other large-scale problems.