Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Ads in Conversations

Martino Banchio, Aranyak Mehta, Andres Perlroth

TL;DR

The paper studies how conversational platforms should time ad auctions when ad quality evolves over a user conversation. By modeling the platform as a real-options problem that can commit to auction format but not timing, it derives starkly different timing equilibria for first-price and second-price auctions: first-price auctions delay to improve allocation quality, while second-price auctions expeditiously run to preserve market thickness, potentially sacrificing efficiency. Introducing Myerson reserves can shift the revenue ranking, with first-price auctions with optimal reserves capable of implementing the optimal mechanism, sometimes outperforming second-price auctions despite their inherent inefficiencies. The work highlights that endogenous information flow and market thickness are central to auction design on next-generation interactive platforms and offers insights applicable to generative AI monetization, dynamic procurement, and M&A-style decision processes.

Abstract

We study the optimal placement of advertisements for interactive platforms like conversational AI assistants. Importantly, conversations add a feature absent in canonical search markets -- time. The evolution of a conversation is informative about ad qualities, thus a platform could delay ad delivery to improve selection. However, delay endogenously shapes the supply of quality ads, possibly affecting revenue. We characterize the equilibria of first- and second-price auctions where the platform can commit to the auction format but not to its timing. We document sharp differences in the mechanisms' outcomes: first-price auctions are efficient but delay ad delivery, while second-price auctions avoid delay but allocate inefficiently. Revenue may be arbitrarily larger in a second-price auction than in a first-price auction. Optimal reserve prices alleviate these differences but flip the revenue ordering.

Ads in Conversations

TL;DR

The paper studies how conversational platforms should time ad auctions when ad quality evolves over a user conversation. By modeling the platform as a real-options problem that can commit to auction format but not timing, it derives starkly different timing equilibria for first-price and second-price auctions: first-price auctions delay to improve allocation quality, while second-price auctions expeditiously run to preserve market thickness, potentially sacrificing efficiency. Introducing Myerson reserves can shift the revenue ranking, with first-price auctions with optimal reserves capable of implementing the optimal mechanism, sometimes outperforming second-price auctions despite their inherent inefficiencies. The work highlights that endogenous information flow and market thickness are central to auction design on next-generation interactive platforms and offers insights applicable to generative AI monetization, dynamic procurement, and M&A-style decision processes.

Abstract

We study the optimal placement of advertisements for interactive platforms like conversational AI assistants. Importantly, conversations add a feature absent in canonical search markets -- time. The evolution of a conversation is informative about ad qualities, thus a platform could delay ad delivery to improve selection. However, delay endogenously shapes the supply of quality ads, possibly affecting revenue. We characterize the equilibria of first- and second-price auctions where the platform can commit to the auction format but not to its timing. We document sharp differences in the mechanisms' outcomes: first-price auctions are efficient but delay ad delivery, while second-price auctions avoid delay but allocate inefficiently. Revenue may be arbitrarily larger in a second-price auction than in a first-price auction. Optimal reserve prices alleviate these differences but flip the revenue ordering.
Paper Structure (16 sections, 12 theorems, 45 equations, 1 figure)

This paper contains 16 sections, 12 theorems, 45 equations, 1 figure.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Model
  4. The Auctioneer's Problem
  5. Auction formats, market thickness, and information
  6. Simple auctions
  7. Optimal Auctions
  8. Discussion
  9. Number of advertisers.
  10. Explicit cost of information.
  11. Conclusion
  12. Mergers & Acquisitions.
  13. Public Procurement.
  14. Extensions
  15. Costly Information Acquisition ($r>0$)
  16. ...and 1 more sections

Key Result

Proposition 1

If assumption:identifiability is satisfied, then $P( \lim_{t \to \infty} \mu_t = \theta \ |\ \theta ) = 1$ almost surely in $\theta$.

Figures (1)

  • Figure 1: This figure depicts a sample path taken by the belief process $\mu_t$ when the conversation follows the Poisson "bad news" model. The beliefs drift upwards in the absence of news (as is the case for the process $\mu^1_t$). Instead, news about advertiser $2$ arrived at time $\tau_2$, and thus $\mu^2_t = 0$ from $\tau_2$ onwards.

Theorems & Definitions (28)

  • Proposition 1
  • Example 1
  • Lemma 1
  • proof
  • Lemma 2
  • proof
  • Theorem 1
  • proof
  • Lemma 3
  • proof
  • ...and 18 more