Cross-channel Budget Coordination for Online Advertising System

TL;DR

This paper introduces AdCob, a novel framework for global budget optimization across multiple online advertising channels, accounting for advertiser competition, with proven effectiveness through experiments.

Contribution

It models competition among advertisers in cross-channel budget allocation and proposes a fast, scalable iterative algorithm with entropy constraints.

Findings

Effective in maximizing overall conversions

Converges quickly in large-scale systems

Proven successful in offline and online experiments

Abstract

In online advertising (Ad), advertisers are always eager to know how to globally optimize their budget allocation strategies across different channels for more conversions such as orders, payments, etc. Ignoring competition among different advertisers causes objective inconsistency, that is, a single advertiser locally optimizes the conversions only based on its own historical statistics, which is far behind the global conversions maximization. In this paper, we present a cross-channel Advertising Coordinated budget allocation framework (AdCob) to globally optimize the budget allocation strategy for overall conversions maximization. We are the first to provide deep insight into modeling the competition among different advertisers in cross-channel budget allocation problems. The proposed iterative algorithm combined with entropy constraint is fast to converge and easy to implement in large-scale online Ad systems. Both results from offline experiments and online A/B budget bucketing experiments demonstrate the effectiveness of AdCob.