Learning From Editorial Decisions: Optimizing Audience-Wide Content Promotions With Causal Machine Learning
MCML Authors
Stefan Feuerriegel
Prof. Dr.
Principal Investigator
Abstract
Stefan Feuerriegel
Prof. Dr.
Principal Investigator
Abstract
Marketing decisions that once relied on expert judgment are increasingly informed by data and models. One prominent application is the audience-wide promotion problem faced by content publishers (e.g., newspapers, magazines, independent producers), who sequentially over time select a few items out of many to promote uniformly to their audience based on rich past information. In this paper, we consider the objective of using data from such expert decisions to learn new decision policies that improve upon them. To this end, we collaborate with a leading Swiss newspaper and use six weeks of high-frequency panel data that record all editorial choices, all information shown to editors at decision time, and all outcomes from their decisions. Informed by the setting, we frame the editorial decision process as a contextual bandit with selection bias but strong variation across contexts and treatments. We combine this framing with causal machine learning methods to learn conditional average treatment effects (CATEs) of promotion and evaluate counterfactual policies. In our application, we find that editors act 'as if' they optimize for the CATE of their promotion decisions and associated uncertainty, yet implementing counterfactual policies that actually optimize for such objectives could have increased annual revenue by USD 0.28–1.41 million. Our study shows how causal machine learning can be applied to observational data on expert decisions to test what latent reward signals experts optimize for and, in turn, evaluate what gains machine-learned policies that optimize for such signals could bring.
BibTeXKey: PFK25