Hi Team,
We have a use case where we want the ability to duplicate an experiment along with it's feature flags.
Context: If we run an experiment with 5 variations and 3 feature flags on scale in production. Now if we want to add a new variation along with adding a new feature flag to the experiment, in the current setup we would have to clone the experiment, add variation and then we have to link all the feature flags and their respective states again along with the new feature. The problem with this approach is that it makes the system very high touch and hence error prone in case of duplicating and modifying the experiment.
Is there a solution for this currently or can this be added as an enhancement to the system.
Thanks!
f
flaky-noon-11399
02/06/2025, 10:04 AM
Hi Gaurav, correct, when you duplicate an experiment the features flags are not saved on the copies. However, when you duplicate a feature, the experiment rules saved on the feature will also be saved on the copies. Would this be more helpful for your usecase? It will just mean duplicating the features but considerably less room for human error
c
creamy-sugar-90600
02/06/2025, 7:51 PM
It will raise the issue of altering feature flags in the application consuming the growthbook api in order to accommodate for this, we cannot cleanly just alter the experiment in this scenario as well.
f
flaky-noon-11399
02/07/2025, 9:28 AM
Ah I see, there does not seem to be a way to do this in the tool currently. For this, I will log a product gap in your honour. In parallel, I'll also log it with the team internally to put it on their radar as we recently launched the experiment templates feature back in version 3.4 as this may be a feature that would complement that. Hope this helps and sorry I do not have better news for you today on this feature being currently available 🙏
flaky-noon-11399
02/07/2025, 9:38 AM
Here is the public-facing GitHub issue. You may wish to upvote and subscribe to the issue so that you are alerted on any updates directly from the Engineering team. There's also an internal note flagging this too.