Основні поняття
The author introduces the MEBS method, applying bid shading to multi-slot display advertising. The approach optimizes shading ratio estimation to maximize expected surplus.
Анотація
MEBS introduces bid shading to multi-slot display advertising, optimizing shading ratio estimation through multi-task learning. Extensive experiments demonstrate its effectiveness and efficiency, leading to significant improvements in Gross Merchandise Volume, Return on Investment, and ad buy count.
Online advertising is crucial for Internet companies, with Real-Time Bidding enabling automated ad matching. Single-slot and multi-slot display advertising differ in cost-effectiveness due to varying user attention levels. Bid shading adjusts bid prices to win more economical positions.
Advertisers are incentivized to cut bid prices for lower but more cost-effective ad slots. Traditional bid shading methods focus on optimal bid price distribution but may not suit multi-slot display advertising due to changing ad positions affecting Click-Through Rate (CTR).
MEBS employs a multi-task end-to-end approach for bid shading in multi-slot display advertising. It optimizes the shading ratio based on win rate and calibrated predicted CTR, maximizing cost savings for advertisers.
Challenges in multi-slot display advertising include proving the optimality of bidding strategy with bid shading and addressing data sparsity issues in estimating optimal bid prices. MEBS addresses these challenges through an end-to-end paradigm and extensive offline and online experiments.
Статистика
We obtain a 7.01% lift in Gross Merchandise Volume.
A 7.42% lift in Return on Investment is achieved.
There is a 3.26% lift in ad buy count.