Optimal Auction Design with User Coupons in Advertising Systems

Xiaodong Liu, Zhikang Fan, Yiming Ding, Yuan Guo, Lihua Zhang, Changcheng Li, Dongying Kong, Han Li, Weiran ShenIJCAI 2024 · OpenReview

Introduction

Traditional online advertising auctions, such as those operated by Google or Meta, typically treat Click-Through Rate (CTR) and Conversion Rate (CVR) as static parameters. In these systems, the platform's role is passive: it selects winners based on submitted bids and pre-calculated relevance scores. However, this approach ignores the platform's potential to actively influence user behavior.

This paper introduces a novel auction mechanism that integrates the Law of Demand into auction theory. By offering coupons to users, the platform can lower the effective price of products, thereby increasing the likelihood of clicks and conversions. The core challenge lies in jointly optimizing coupon values and ad placement to maximize platform revenue while maintaining advertiser incentives for truthful bidding.

Mechanism Design

The researchers developed a new class of mechanisms that incorporate a coupon function alongside standard allocation and payment rules. This framework ensures that the auction remains Incentive Compatible (IC) and Individually Rational (IR), meaning advertisers are best served by bidding their true valuations and are guaranteed not to lose money by participating.

The Modified Virtual Value

Building on Myerson’s classic auction theory, the authors defined a modified virtual value function. Unlike traditional virtual values that only consider advertiser bids and valuation distributions, this new function accounts for:

  1. The advertiser's bid.
  2. The direct cost of the coupon to the platform.
  3. The expected boost in CTR and CVR resulting from the coupon.
The coupon function (left) and the resulting modified virtual value function (right) used to determine optimal ad placement.

In an optimal setup, the platform offers a coupon where the marginal cost of increasing the coupon value exactly equals the marginal revenue gain from the increased conversion probability. To handle real-world "irregular" data distributions, the authors adapted the "ironing" technique to ensure the mechanism remains robust and truthful.

Experimental Evaluation

The mechanism was evaluated using both synthetic and industrial datasets:

  • Synthetic Data: 100,000 simulated auctions with 8 advertisers using log-normal valuation distributions.
  • Industrial Data: A dataset of 300,000 records from Kuaishou, a major short-form video platform. A neural network was trained on this data to predict CVR responses to varying coupon values.

Key Results

The proposed mechanism significantly outperformed standard Myerson auctions and baselines using constant (CCV) or randomized (RCV) coupon values. On synthetic data, the mechanism increased revenue by 19.76% and advertiser utility by 32.58%.

Performance comparison on industrial data, showing significant improvements in revenue, social welfare, and advertiser utility.

As shown in the table above, the results on industrial data were even more pronounced. Compared to the Myerson baseline, the coupon-based mechanism increased revenue by 28%, social welfare by 67%, and advertiser utility by a staggering 313%.

Discussion and Future Work

This research demonstrates a rare "win-win-win" scenario in auction design. By treating CTR and CVR as dynamic variables, the platform creates value for users (lower prices) and increases volume for advertisers, which ultimately drives higher platform revenue.

While the results are promising, the authors identify several areas for future research. The mechanism currently assumes the platform can perfectly predict the "demand curve" (how users respond to coupons). Additionally, extending the model from single-slot auctions to multi-slot scenarios and incorporating advertiser budget constraints remain important challenges for real-world deployment.

Summary updated

Abstract

Online advertising is a major revenue source for most Internet companies. The advertising opportunities are usually sold to advertisers through auctions that take into account the bids of the advertisers and the click-through rates (CTRs) and the conversion rates (CVRs) of the users. Standard auction design theory perceives both the CTRs and the CVRs as constants. We consider a new auction mechanism that offers coupons to users when displaying the ads. Such coupons allow the user to buy the advertisers' products or services at a lower price, which increases both the CTRs and the CVRs of the ads. In this paper, we formulate the problem mathematically and perform a systematic analysis. We characterize the set of individually rational and incentive compatible mechanisms in our setting. Based on the characterization, we identify the optimal strategy of offering coupons that maximizes the platform's expected revenue. We also conduct extensive experiments on both synthetic data and industrial data. Our experiment results show that our mechanism significantly improves both the revenue and welfare of the platform, thereby creating a win-win situation for all parties including the platform, the advertisers, and the user.
BibTeX
@inproceedings{liu2024optimal,
  author = {Xiaodong Liu and Zhikang Fan and Yiming Ding and Yuan Guo and Lihua Zhang and Changcheng Li and Dongying Kong and Han Li and Weiran Shen},
  title = {{Optimal Auction Design with User Coupons in Advertising Systems}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2024},
  url = {https://openreview.net/forum?id=yYIlXyjQ7m},
}
Figure