Short answer: based on the docs I have, the Retention → Frequency report gets you very close conceptually (it measures “unique days in a week/month” etc.), but it is not documented as a way to:
restrict to an arbitrary saved cohort, and
then output a single “average unique days per user” number for that cohort.
So I can’t confirm a fully native, point‑and‑click way to get exactly “avg unique days per user for event + property filter + cohort” from the available documentation.
What Frequency Retention actually does
Frequency Retention:
Buckets users by how many unique time buckets they were active:
Lets you choose Cumulative (“at least X unique intervals”) or Non‑cumulative (“exactly X unique intervals”). [Frequency View]
So conceptually, if you:
- 1.
Use Retention → Frequency.
- 2.
Set the event to your Viewed Widget event.
- 3.
Filter by widget_id = 5443.
- 4.
Choose an interval such that the “unique intervals” correspond to unique days (e.g. “unique days in a week” or “unique days in a month”). [Frequency View]
…you will see, for each cohort of “users who did the event in week X”, how many of them did it on 1, 2, 3, … unique days that week.
However, the docs don’t show:
A way to plug in a saved cohort as the population for Frequency Retention.
A built‑in “average” over those frequency buckets (e.g. turn the distribution into a single “4.1 unique days per user” metric).
Because of that, I cannot give a precise, documented step‑by‑step that ends with “and here is the average unique days per user for your saved cohort”.
Is there a “correct native way” documented?
From the docs I have:
“Frequency per User” → counts events per user, not unique days. [Measurements]
Computed / Behavioral properties:
“Frequency per User” and “Aggregate Property per User” again operate on event counts or numeric properties, not “unique days active”. [Computed Properties]
None of these are documented as “unique days per user” metrics.
Given that:
I don’t see a documented native metric that is exactly “avg unique days per user” for:
restricted to a saved cohort.
I can’t confirm that Retention → Frequency plus some UI option will output that exact number.
What is documented as flexible enough
The only documented mechanism flexible enough to compute arbitrary per‑user metrics is JQL with groupByUser, which lets you:
Inspect their timestamps,
Build a set of unique dates per user,
But the docs I have don’t show:
How to directly plug a saved cohort into JQL, or
A ready‑made snippet for “unique days per user”.
So I can only say: JQL is the documented path that can express this calculation, but I can’t provide an official, copy‑paste‑ready solution from the sources.
-------------------------
Because of these documentation limits, I can’t honestly say “yes, Retention → Frequency is the correct native way and here is the exact setup” nor “here is the official native recipe for avg unique days per user for event + property + cohort.” If you need this metric strictly within the UI, you may need to approximate it via Frequency Retention distributions, or compute it exactly via JQL and then post‑process the results.
