Large-scale heterogeneous treatment effect estimation for advertising experiments and targeting

Instead of assuming every user reacts the same way to an ad or intervention, this use case estimates which users are actually more likely to change their behavior because of the treatment.