Wednesday, January 21, 2015

A/B testing, multiple comparisons, imaginary uplifts

Yesterday gave a talk on Web Analytics Wednesdays Singapore (analyticsmaturity.com).

If you made A/B test, found significant result, implemented to production and see no real improvement - you just suffered from "imaginary uplift". In this presentation I explain why this is happening (and so often!) and what do to about it.

Post with source code for all experiments is coming.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.