Bidding Machine: Learning to Bid for Directly Optimizing Profits in Display Advertising

Abstract

Real-time bidding (RTB) based display advertising has become one of the key technological advances in computational advertising. RTB enables advertisers to buy individual ad impressions via an auction in real-time and facilitates the evaluation and the bidding of individual impressions across multiple advertisers. In RTB, the advertisers face three main challenges when optimizing their bidding strategies, namely (i) estimating the utility (e.g., conversions, clicks) of the ad impression, (ii) forecasting the market value (thus the cost) of the given ad impression, and (iii) deciding the optimal bid for the given auction based on the first two. Previous solutions assume the first two are solved before addressing the bid optimization problem. However, these challenges are strongly correlated and dealing with any individual problem independently may not be globally optimal. In this paper, we propose Bidding Machine, a comprehensive learning to bid framework, which consists of three optimizers dealing with each challenge above, and as a whole, jointly optimizes these three parts. We show that such a joint optimization would largely increase the campaign effectiveness and the profit. From the learning perspective, we show that the bidding machine can be updated smoothly with both offline periodical batch or online sequential training schemes. Our extensive offline empirical study and online A/B testing verify the high effectiveness of the proposed bidding machine.

An Effective Budget Management Framework for Real-Time Bidding in Online Advertising

“…But it does not support budget management, so the budget efficiency has not been guaranteed. Recently, the authors in [14] analyze that there are three main challenges when the advertisers optimizing their bidding strategies in RTB, namely (i) estimating pCTR of the ad impression, (ii) forecasting the market value of the given ad impression, and (iii) deciding the optimal bid for the given auction based on the first two. Furthermore, they point out the three challenges are strongly correlated and dealing with any individual problem independently may not be globally optimal.…”

“…The winning notice is then sent to the advertiser and the winner's ad will be fed back to the SSP, and displayed to the user on the page of the publisher. In practice, the entire process needs to be completed within 100 milliseconds [14]. Later, the DSP that the winner belongs to will track the user's behavior (e.g., click, conversion), and optimize the bidding strategy based on the user response.…”

Header Bidding as Smart Service for Selling Ads in the Digital Era

“…Research on real-time bidding investigates auctions in ad networks as the dominant form for allocating impressions. It analyzes how advertisers can optimize their bids (Ren et al, 2018), how publishers can allocate their advertising space and maximize revenue (Balseiro et al, 2014), and how equilibrium advertisers' and publishers' strategies together drive equilibrium profits (Sayedi, 2018). Finally, it lays out how and with what granularity demand-side platforms should segment real-time bidding advertising markets (Qin et al, 2017).…”

Impression Pacing for Jobs Marketplace at LinkedIn

“…There are also those bidding approaches that mention pacing in their methodology, but not directly address it. For example, the work presented in [11] jointly optimizes response prediction, bidding landscape forecasting, and bidding strategy. The paper mentions pacing as an important consideration for the bidding portion of their unified framework.…”