https://www.growthbook.io/ logo
#ask-questions
Title
# ask-questions
l

limited-ghost-77841

10/16/2023, 11:29 PM
Hey everyone, I started an experiment in June, added a variant in July, and just tried adding two more variants (JSON) through the Features tab (6 variants in total now). I'm getting an error "*Warning:* Expected 4 variation ids (
0
,
1
,
2
,
3
), but database returned 6 (
0
,
1
,
2
,
3
,
4
,
5
)." It seems like there's no way to refresh the full set of 6 variants and have them show on the experiment screen. I'm wondering if I should just remove the old feature flag and experiment and try recreating them. Does anyone know if there are any limitations when deleting and recreating a feature flag under the same name? Our client app is already deployed with active users but we have sensible fallback values in case of network errors. Right now it seems like we've got feature flagging with the correct % distribution but no experiment reporting. Thanks
I found the "Edit variation" option by switching to the old experiments UX. I'm guessing the best option is to keep the current feature flag / experiments for now and remove and rebuild the feature and a new experiment for the next round/phase of testing.
f

fresh-football-47124

10/17/2023, 1:08 AM
the new UI has it, but only when the experiment is in draft mode
You added variations to the same experiment?
did you set up new phases?
if you recreate the same experiment and use the same key, it will show the same results - the biggest thing that a new experiment will do is change the start dates
which you can achieve through phases
l

limited-ghost-77841

10/17/2023, 1:11 AM
Okay thanks, yes I added new variants to same experiment and set up a new phase.
I tried removing some of the old variants but it appeared to mess with the UI. Thanks for confirming, maybe I'll recreate the feature flag (using same key) and create a new experiment to have cleaner reporting in the new UI
f

fresh-football-47124

10/17/2023, 1:13 AM
ya, typically we don’t recommend adding variants to an existing experiment
l

limited-ghost-77841

10/17/2023, 1:14 AM
Good to know, we have a mobile app with a lot of existing deployments. I guess with the change of managing variants at the experiment level, instead of feature flag, that might be a bit easier in future. Thanks for the quick help
16 Views