Skip to main content
Problem Solver I
April 22, 2024
Solved

Automatic selection of A/B test winner on flows

Hi,

I’ve had an A/B test running on a flow and I was under the impression that the test would end automatically once statistical significance was reached, which according to this article is defined as:

  • At least 500 recipients have received each variation.
  • A variation has at least 90% win probability.

As far as I can tell, this test (pictured below) satisfied both criteria, but it still says it’s not statistically significant. In the end I just manually chose the winner, but could someone explain why this wasn’t classed as statistically significant? 

 

This topic has been closed for replies.
Brian Turcotte
Klaviyo Alum
April 23, 2024

Hi @kfodor!

That is interesting - at first glance, it seems it should have deemed it statistically significant. 

 

Did you change any of the default definitions cited here?

- Brian

2 replies

Brian Turcotte
Klaviyo Alum
April 23, 2024

Hi @kfodor!

That is interesting - at first glance, it seems it should have deemed it statistically significant. 

 

Did you change any of the default definitions cited here?

- Brian

kfodorAuthor
Problem Solver I
April 23, 2024

Hi @Brian Turcotte 

Thanks, I’m glad I’m not the only one that thought so! I don’t think I changed any of the default settings when I set the test up, so it could be a bug...or I’m forgetting something. 

I have another test running on a different flow that only needs ~50 more recipients on one of the versions, so I’ll see what happens there. If it again doesn’t end automatically, I might report it as an issue!

Thanks,

Katherine