Learning Sensitive Combinations of A/B Test Metrics

Eugene Kharitonov, Alexey Valerievich Drutsa, Pavel Serdyukov · 2017

Online search evaluation, and A/B testing in particular, is an irreplaceable tool for modern search engines. Typically, online experiments last for several days or weeks and require a considerable portion of the search traffic. This restricts their usefulness and applicability.

Read the paper · More papers on PaperTik