Hi! I'm running an A/B test with traffic allocation set to 33%, and the traffic distribution across the variants isn't equal. Could the 33% traffic allocation be causing this, or should I be looking for another explanation?

Hi Mario, good morning! Are you seeing any SRM errors in the Health tab of the affected experiment? If so, please can you confirm which SDK you're using and also what type of experiment this is? A redirect URL test, Visual Editor test or feature flag test?

I don't have the Health panel available. We're using the JS SDK v1.6.5, and it is connected correctly. I found that one of our development teams launched an experiment without any targeting rules. As a result, the experiment was exposed to all traffic, including pages where other experiments were already running. We read

experiment.variationId

and send it to our internal analytics system. In some experiments, the reported

variationId

came from this untargeted experiment, while in others it appears that the correct

variationId

from the intended experiment was sent. How does the GrowthBook SDK behave when multiple experiments are active for the same user? Which experiment's

experiment.variationId

is returned by the SDK in that case?

Ah got it, glad you were able to isolate the issue and get it resolved. Thank you for the update 🌻