Hello, we've set up a custom sticky bucketing serv...
# ask-questions
n
Hello, we've set up a custom sticky bucketing service for react native but are struggling to validate that it's setup correctly - do you have any advice on how to ensure our setup is correct and that our users won't be switching between variants
h
Hi Jo. Basically you'd need to: 1. create an experiment with SB enabled 2. observe the bucket value for a given user 3. make a change to your experiment (ex: setting variation weights to 100%/0% to force the opposite variant). Be sure to use a strategy that uses sticky bucketing (ex: "Same Phase, apply changes to new traffic only"). 4. observe that the user remains in their original bucket despite the change
n
thanks @happy-autumn-40938 - i've run a test where i set up an experiment with 50/50 split, got a user in the variant of the test, then updated the split to 0 control/100 variant. I didn't have the option to use sticky bucketing, is that intended behaviour? what are the constraints around when you can or can't use sticky bucketing for increasing roll-out or split percentages?
h
Changing the variation weights on a running experiment with linked feature flag (or VE / redirect implementation) should prompt you to choose a release strategy. SB-powered strategies should be in the dropdown. If you're not seeing them, verify that your org has sticky bucketing enabled in the settings menu.
n
Thanks for the clarification @happy-autumn-40938, we definitely have sticky bucketing enabled, and used a custom implementation for it due to the nature of our app. are there any additional docs I can use to check our implementation against?
h
This is basically it: https://docs.growthbook.io/app/sticky-bucketing Are you going through the "Make Changes" flow (button on the top-right of your experiment page)? Can you screenshot what options are available?
n
this is the only option i have available.
i followed the docs here for custom implementation https://docs.growthbook.io/lib/js#sticky-bucketing, would be really helpful if there were some examples or more documentation around that, and maybe the steps you outlined above on how to test the custom service implementation
for context, we definitely had the sticky bucketing option available when we first set it up, but it wasn't always an option and we haven't used it enough to figure out when it was/wasn't available
h
apologies, I forgot to mention that for changing variation weights you need to go through the "advanced" flow, otherwise only statistically safe options are shown in the release plan options.
👀 1
n
ahhh, that explains it! i have all the options i need now 🙏 thanks so much for your help with this!
🎉 1
@happy-autumn-40938 what are the implications of sticky bucketing not being a statistically safe option when selecting the advanced changes when sticky bucketing isn't an option in the other change types? anything we should be aware of when interpreting the results?
h
As I understand this is just a placeholder experiment to validate that sticky bucketing is working. Normally you would avoid the advanced options, and we would try to guide you towards sensible choices that do not bias your results.
👍 1
Think of advanced as “I know what I’m doing, give me an escape hatch” 🙂
n
got it, makes sense.thanks again for your help!
h
You’re welcome