Hi Team, I want to ask about percentage rollout lo...
# sdk-flutter
a
Hi Team, I want to ask about percentage rollout logic from Growthbook. Assume currently I have split between control and variant with same value 50%. Then, my team want to update the variant from 50% to 70% and control from 50% to 30%. Users that existing get control will be changed into variant and vice versa? Or, just changed from control to variant only? Without any changes from existing variant?
c
Hi @acoustic-wolf-53513. Please let us check that and we will write to you.
a
Ok @calm-dog-24239 thanks
🙌 1
c
If you change the rollout split from 50%/50% to 70%/30%, GrowthBook will recalculate the assignment for all users based on their hash and the new percentage ranges. This means: • Some Control users will move to Variant. • Some Variant users can also move to Control (if their hash now falls outside the new Variant range). • GrowthBook does not “remember” previous assignments; it always evaluates based on the current configuration.
a
So, Growthbook didn't have any approach how to "remember" this?
c
To avoid users switching between Control and Variant when changing rollout percentages, you should enable sticky bucketing in GrowthBook. This feature keeps users assigned to their original variation even if the percentages change.
a
Any article related how to enable sticky bucketing in GrowthBook?
But, when the variant increased, still need changes on several users at control.
c
I recommend trying the "Same Phase, apply changes to everyone" option when updating the rollout percentages. This keeps the same randomization seed, so most users maintain their original variant assignment, minimizing users switching back from Variant to Control.
a
This feature is available for Growthbook self hosted? How I can access this feature @calm-dog-24239?
c
If you’re referring to sticky bucketing, yes — it is a paid feature. If you’re referring to changing the rollout, that is also possible. When making changes to the experiment in the admin panel, please select the appropriate plan, as shown in the screenshot.