Hi! We’ve decided to use GrowthBook for our project, but have some questions on the Experiment Analysis part. We are using GA4+GBQ data for analysis and the first iteration we did with GrowthBook was that in API we construct the user object with related feature flags and their values and send it for the FE clients (web/mobile). Then FE uses those flags to determine the related feature/experiment logic and add the feature flag props to already existing GA4 events (eg search) as event_params (eg

improved_search_feature: true

). But to determine the variation groups, this approach relies on the actual feature flag values, so the user would be divided into variation group by the `true`/`false` feature value. Therefore I'm not able to run proper A/A tests as feature flag values are same for all users aka I have no data on the user variation group on the GA4 event. So I'm now thinking of a solution where I trigger a

experiment_assigned

event from BE (in

trackingCallback()

) whenever user (feature) flags are requested and user is included into the experiment. It does not mean that experiment is actually being "conducted" for the user as flags are just part of user object and user might not travel through the flow where experiment is defined. So in GrowthBook I would have

experiment_assigned

events for purely user experiment assignment purposes. The actual experiment event is based on already existing GA4 events (search, item bought etc) My question: Is this solution overall viable for GrowthBook experiment analysis: to send feature flags as part of user object from API -> FE and having the experiment user assignment event decoupled from the actual experiment in FE? Or should I take another approach on this topic? Asking because I did not find a good material from the docs how to actually construct and link the experiment events together. Appreciate for any help or thoughts on this!

I think that will work. You can use activation metrics should you want to continue using this method of tracking assignments outside of experiment exposures.

Thank you for the response! Currently we've decided to take a step back and move the GrowthBook implementation to web where the actual experiment takes place