Optimal Revenue Guarantees for Pricing in Large Markets

TL;DR

This paper investigates revenue guarantees of posted price mechanisms in large markets, showing improved bounds for fixed-price policies and analyzing the impact of market size and multiple units on revenue performance.

Contribution

It proves that fixed-price posted price mechanisms achieve higher revenue guarantees in large markets and analyzes how these guarantees change with multiple units.

Findings

Fixed-price PPM guarantee improves from 0.632 to 0.712 in large markets.

Large market advantage diminishes as the number of units increases.

Revenue gap approaches 1 - 1/√(2kπ) for selling k units.

Abstract

Posted price mechanisms (PPM) constitute one of the predominant practices to price goods in online marketplaces and their revenue guarantees have been a central object of study in the last decade. We consider a basic setting where the buyers' valuations are independent and identically distributed and there is a single unit on sale. It is well-known that this setting is equivalent to the so-called i.i.d. prophet inequality, for which optimal guarantees are known and evaluate to 0.745 in general (equivalent to a PPM with dynamic prices) and $1 - 1/e \approx 0.632$ in the fixed threshold case (equivalent to a fixed price PPM). In this paper we consider an additional assumption, namely, that the underlying market is very large. This is modeled by first fixing a valuation distribution F and then making the number of buyers grow large, rather than considering the worst distribution for each possible market size. In this setting Kennedy and Kertz [Ann. Probab. 1991] breaks the 0.745 fraction achievable in general with a dynamic threshold policy. We prove that this large market benefit continue to hold when using fixed price PPMs, and show that the guarantee of 0.632 actually improves to 0.712. We then move to study the case of selling k identical units and we prove that the revenue gap of the fixed price PPM approaches $1-1/\sqrt{ 2k\pi}$. As this bound is achievable without the large market assumption, we obtain the somewhat surprising result that the large market advantage vanishes as $k$ grows.