with Mihir Tendulkar, Simon Ejdemyr, Dhevi Rajendran, David Hubbard, Arushi Tomar, Steve Beckett, Judit Lantos, Cody Chapman, Apoorva Lal, Ekrem Kocaguneli, and Kyoko Shimada. Netflix Tech Blog, Jun 2024.

An automated approach to estimate annualized incremental business impact from A/B test results, replacing a previously manual Finance/Strategy process. The method uses a surrogate index approach and transportability assumptions to project treatment effects beyond the observed time window.

Presented as part of Netflix's internal Causal Inference and Experimentation Summit, alongside four other talks covering survey weighting, synthetic control for Netflix Games, metric tradeoffs via double machine learning, and experimentation platform design. See the full survey: A Survey of Causal Inference Applications at Netflix.