if you have your own bucketing/assignment code, then you can still use GrowthBook for the analysis - but you’ll need some way to pull which user saw which experiment and variation from the db.