Estimating quantile treatments efficiently in experiments.

When we run experiments, we look for ways to make them find results faster. After all, it’s a business and speed matters. Usually, we use a CUPED-like method. But these methods only work for average treatment effects. In this short blog post, I show how to do this for quantile treatment effects (spoiler: it’s not quantile regression, that does not work in this context).

Blog post.