Hi everyone, We’re encountering an issue where ex...
# experimentation
h
Hi everyone, We’re encountering an issue where experiments we’ve set up and see running in the GB dashboard are not appearing under Experiments in the GrowthBook dev tool. Recently, we added a new boolean feature flag that we confirmed to be working. However, when we assigned this flag to a new experiment, it didn’t show up as expected when we started the experiment. To troubleshoot, we adjusted the split to 1% (variant) and 99% (control), but we’re still not seeing assignment to the control group, even with a 99% split. It’s worth noting that other experiments are running correctly but are also not visible in the GrowthBook dev tool. Does anyone have insights into what might be causing this issue?
s
Which SDK are you using?
h
the react one, I am running it on Nextjs15.
s
Thanks. If it's server rendered, then the dev tools won't work, as they only work with client-side implementations.
h
@strong-mouse-55694 we are only using
"@growthbook/growthbook-react": "^1.1.0"
and should be client-side in nextjs client components. The weird thing is that if I permanently turn the flag on it would work but not if it's in an experiment.
s
What do you mean by "permanently turn the flag on"?
h
@strong-mouse-55694 Oh sorry, what I meant was if I just turned the the flag on for everyone.
s
You mean remove all the rules? Are you able to share a screenshot of the config? Or a URL where it's live?
h
@strong-mouse-55694 here is the screenshot of the experiment setup/config that is having trouble. What I meant from my previous reply was if I turn the feature flag on for everyone, it will work, but when it's in an experiment I could never get assigned the
Variation 1
even when I get skewed the split to 99% for testing purposes. So I suspected that the experiment was somehow not loaded.
s
Gotcha. Couple things. The experiment assignment is persistent if you're using the same browser and have your
id
set up correctly. That means that you should always be in the same variation, every time. Also, I see that you have attribute targeting set up for country. Can you confirm that the country attribute is being passed in correctly? If it isn't set, then the experiment won't run.
h
@strong-mouse-55694 Now I did more testing, I think it was just not working with our unauthed users due to this experiment assignment table. Does that sound related to you? Also how do I edit that field I can’t seem to find the settings for it. Thanks.
s
Yes! That's likely it. To configure how that works, go to Metric and DataData Sources. Look for your data warehouse and check out Identifier Types. That'll be where the identifier types are set up. And, when you set up your experiment, you can choose which identifier to use. So, if you don't want to target logged-in users, then you can choose a different identifier like
userId
or whatever you've configured.
h
@strong-mouse-55694 is this you were talking about? I dont seem to find the Identifier Types here, maybe I am at the wrong place?
s
The mixpanel data source looks right, but there's something amiss here—I'm not sure what. In your experiment setup, you have
deviceid
and the logged-in user assignment table. Is your initial setup? Or were experiments working previously? I'd try setting up another ff/experiment and ensure that you're in the right project and that all the identifiers, etc. are aligned with your mixpanel datasource.
h
@strong-mouse-55694 yeah everything was working before, i think the Logged-in Users for the Experiment Assignment Table is the culprit to this not working now for our unauth flow. FYI we also use
id
when we are certain that the feature flag only triggers within our auth wall/to authed users only
s
gotcha. if you select all projects and look at your datasources, do you have others configured?
h
@strong-mouse-55694 yes I have this added very recently by another dev on the team. is this related?
s
I think so. Can you try creating a new experiment and see what's included in the analysis settings? Or, if you click edit on Analysis Settings, are you able to change this?
h
@strong-mouse-55694 I can choose two data source, but the other dev said that the Sample Data Source is another isolated demo project so should not affect this whatsoever.
@strong-mouse-55694 yeah so I set the variant 1 traffic split to 99% just to test and seems like when i am signed in I can get it turn true and when I am unauthed its false.
r
was this solved @happy-mechanic-46342 or @strong-mouse-55694? I have the same issue here I never used
Experiment Assignment Table
before and now this value is there, I have no idea where the table there comes from and what it is supposed to do
h
It appeared to be a visual bug on GB if you are using Mixpanel data source.