Method

Double machine learning

ML nuisance models with orthogonalization and cross-fitting for treatment effects.

9 cases from 8 companies apply Double machine learning. 1 was checked in an online A/B test; 5 report results in production.

Typical tasks: Measure impact without an A/B test, Decide who gets an offer, Experiment when users affect each other, Optimize decisions and 3 more. Companies: T-Bank, DoorDash, Lyft, Netflix and 4 more. Often combined with: Causal inference (general), Doubly robust estimation, Inverse propensity weighting, Random forest, Logistic regression and 23 more.