Adobe Target

LSchwanka · 12/1/25

Hello everyone,

I'm facing a problem with an A/B test, so here's some info:

I observed that shortly after being launched, a test setted 50/50 started to become 20/80.
Right after this, it ran unbalanced for approximately 10 more days, and then tried to return to 50/50, still stumbling.
Auto-allocate is turned off.
Most of the traffic is going to B, the challenger version.
It is implemented on a webpage.

Has anyone experienced this before? Could it be an implementation issue? Any tips?

Thanks already!

Perrin_Ennen · 12/2/25

That shouldn't be the case. It should actually be accurate when there is a lot of traffic – see: Acceptable Traffic Split Deviation in Adobe Target A/B Testing

A slight deviation is normal, but a 20/80 deviation is unusual, unless the test has just started and/or has very little traffic.

If I remember correctly, Adobe Target uses window.crypto to generate the A/B split with random values. I used to generate this myself with Math.random to move users into different groups via profile scripts.

In your case, I would simply double-check with another A/B test to see if the split itself is working or if something else, such as the implementation, is not quite right. In other words, a test activity that does a 50:50 split and does nothing itself except, for example, create a test log, so that you can test this yourself.

Hope that helps.