Skip to main content

Understanding Sample Ratio Mismatch (SRM) in Your A/B Tests

What Sample Ratio Mismatch (SRM) means in a Shogun A/B test, why early confidence levels can swing, and what to do if you spot it.

Written by James Power

What is Sample Ratio Mismatch?

Sample Ratio Mismatch (SRM) happens when the number of visitors landing in each variant of your test doesn't match the traffic split you configured — for example, you set a 50/50 split but end up with something like 53/47 in sessions or pageviews. A small amount of variance is normal, but a persistent or growing mismatch can signal that something is skewing your results, and it's worth understanding before you trust a "winner."

Why does my confidence level swing early in a test?

Early in a test, small differences in a handful of conversions can produce a misleadingly high confidence score. Shogun requires a minimum of 20 successful conversions per variant before it begins reporting confidence — this reduces, but doesn't eliminate, the chance of an early result looking more decisive than it is. As more data comes in, that confidence can and often does move, sometimes significantly.

What to do if you see SRM or a large confidence swing

Reach out to our support team in-app with the test name — we can pull the result history and take a look.

Did this answer your question?