Frequently Asked Questions

Date of slack thread: 8/5/24

Anonymous: Curious if I am overlooking the ability to turn off a variant in an experiment and push the traffic allocation to the remaining variants? We have an experiment live with 3 variations. v3 is tanking but the only way we know how to turn it off is by disabling a group. If we update the experiment splits (or deleting a variant), we’ll need to reset the experiment which would mean restarting it completely and losing data.

Anonymous: My understanding is correct that we cannot push the available traffic freed up from v3 to v1 & v2? Only option is to have v3 traffic see the control experience?

Logan Bates (Statsig): When you say “available traffic” you mean users who were previously allocated into the v3 (tanking variant)?

Anonymous: Not previously allocated, but future allocation. So instead of 25/25/25/25, 25 control & 37.5 v1 & v2

Anonymous: I know not a huge difference, but just curious

Makris (Statsig): Hey Mayghen, we don’t support this for AB tests since the users would then change which variant they’re assigned to, inducing crossover.