Source: arXiv — 2026-08-06
Summary
Itai Ashlagi (Stanford), Shahar Dobzinski (Weizmann Institute), Jacob D. Leshno (University of Chicago Booth), and Sigal Oren (Ben-Gurion University) study how a revenue-maximizing seller facing two bidders, given access to only a single binary signal about the bidders' joint value, should design that signal. They compare three signal classes — demand (overall value level), ranking (relative strength between bidders), and competitiveness (spread of values while keeping bidders anonymous) — and characterize the revenue-optimal signal and matching auction mechanism for each.
Key Takeaways
- The setup deliberately restricts the seller to one bit of information about two bidders' joint values, rather than assuming full knowledge — modeling real-world settings where a platform only has a coarse signal to work with.
- Three fundamentally different signal types are compared head-to-head: whether overall demand is high or low, which bidder is relatively stronger, and how spread out (competitive) the two bidders' values are without revealing who's who.
- The paper's central result: under standard regularity conditions, the optimal competitiveness signal earns at least as much revenue as the optimal demand or ranking signal — and often strictly more.
- When restricted to signals that induce a monotone allocation rule (a natural fairness-adjacent constraint), the optimal competitiveness signal beats every other possible binary signal, not just the other two named classes.
- This gives platform and marketplace designers a concrete, provable answer to "what one piece of information is worth revealing" — competitiveness, not raw demand level or a head-to-head ranking, is the revenue-maximizing choice.