Statistical Learning from Attribution Sets

TL;DR

This paper develops a privacy-preserving method for training conversion prediction models using attribution sets, providing unbiased loss estimation and robust guarantees, outperforming industry heuristics especially with large or overlapping attribution sets.

Contribution

It introduces a novel unbiased estimator for loss from attribution sets and demonstrates its effectiveness and robustness in privacy-constrained advertising scenarios.

Findings

Unbiased estimator enables effective learning from attribution sets.

Method outperforms industry heuristics in experiments.

Robustness to prior estimation errors.

Abstract

We address the problem of training conversion prediction models in advertising domains under privacy constraints, where direct links between ad clicks and conversions are unavailable. Motivated by privacy-preserving browser APIs and the deprecation of third-party cookies, we study a setting where the learner observes a sequence of clicks and a sequence of conversions, but can only link a conversion to a set of candidate clicks (an attribution set) rather than a unique source. We formalize this as learning from attribution sets generated by an oblivious adversary equipped with a prior distribution over the candidates. Despite the lack of explicit labels, we construct an unbiased estimator of the population loss from these coarse signals via a novel approach. Leveraging this estimator, we show that Empirical Risk Minimization achieves generalization guarantees that scale with the informativeness of the prior and is also robust against estimation errors in the prior, despite complex dependencies among attribution sets. Simple empirical evaluations on standard datasets suggest our unbiased approach significantly outperforms common industry heuristics, particularly in regimes where attribution sets are large or overlapping.