Hi, a common scenario for ramping down a multi-variant test is to ramp down only one of the variants (say, ramping down treatment_2 from 33% to 0% in a control 34%/treatment_1 33%/treatment_2 33% setting). In our current abtest config, we can ramp down the treatment_2 to 0% and the remaining 67% of the traffic is still be in the experiment (i.e. 33% of the users will no longer be in the experiment, for those who were in treatment_2, they will experience the control but won't be registered to the control). This is the best practice IMO because it stops the damage caused by treatment_2 and doesn't reassign users in treatment_2 to control or treatment_1 to contaminate the buckets. However, it seems we can't make such config change in the growthbook. Could you please suggest the recommendation for this scenario in growthbook? Thank you!

Hi Tino- I'll discuss this scenario with the team. There are complications with sticky bucketing that would also need to be solved

Got it. Thank you!