Saved Groups in GrowthBook lets us define groups once and reuse that combination elsewhere to target across experiments. We would use condition groups to have pre-set groups that our users can select.
Goal / scope
- Produce a catalog of saved groups that teams need
- Determine what's needed for v1 vs. later (i.e. we anticipate teams may want X in the future, but we don't need to do that today)
- For each group, define the condition and the source for any metric-derived membership (e.g. top 10 wikis by pageviews -- what table/query/dashboard is that coming from?)
- Establish naming protocol
- Document an annual refresh plan
Product questions
- What groups are our user teams looking for? (collect requests)
- By what metrics? (e.g. Top 50 wikis by... pageviews? active editors? articles?)
- How often do we revise these groups? IIUC, editing a saved group changes membership immediately, so we don't want to revise groups while a dependent experiment is running. We should time this carefully either around late Dec (before holiday deployment freeze) or early July (new FY)
Engineering questions
- Would we deprecate groups that aren't being used? (do groups cost us anything?)
- How might we test that this is working as intended?
Acceptance criteria
- A catalog exists listing each group with: type, definition, computation source + cadence, owner.
- Naming conventions are documented.
- An revision plan including timing of changing groups is agreed upon.
- v1 groups are created in GrowthBook