Sample Pollution

Published by

March 15, 2017

When we conduct A/B testing, sample pollution happens when the same user is subjected to experiment, but due to deleting the cookie, they see both the original and challenger variation. This violates the terms of controlled experiment, as we can no longer be sure whether the visitor was influenced in his choice by the variations we made. The best way to eliminate the sample pollution is to limit the duration of the experiment, so the likelihood of cookie deletion on client browsers will be lower.

Back to Glossary Index

Published by