2017 IEEE International Conference on Big Data (Big Data)
Download PDF

Abstract

A/B testing is commonly used to evaluate new features of websites and mobile apps. Before the A/B phase of the experiment, one best practice is to do A/A validation, where all experiment groups receive the control experience. A/A validation requires time and effort to carry out. It serves to ensure that there are no pre-existing differences between the control and test groups as well as to verify the data is flowing in as expected. We propose a method of assigning users to buckets such that the likelihood of observing such a pre-existing difference is vanishingly small. The proposed framework allows us to skip A/A validation, run experiments more quickly and develop our products more swiftly.
Like what you’re reading?
Already a member?Sign In
Member Price
$11
Non-Member Price
$21
Add to CartSign In
Get this article FREE with a new membership!

Related Articles