50/50 Split or not?

Mark Eltsefon
Analytics Vidhya
Published in
3 min readFeb 13, 2022

--

Whenever we hear about A/B testing , we inherently assume that we separate our traffic equally. Let’s try to find out what are the underlying reasons for not splitting just by 50/50.

There are two main linked reasons for trying this:

  1. We mitigate the risk if there are some problems with the implementation or if significant bugs emerge.
  2. We save money in case our experiment is unsuccessful.

These two reasons are aligned with the overarching goal of A/B testing , in general. Faster and with more confidence to get which…

--

--

Mark Eltsefon
Analytics Vidhya

A data scientist focusing on experimentation and constantly seeking new knowledge.