Daily activity and retention answer different questions. Retention asks whether members of a defined cohort returned; a count of everyone active today cannot establish that.
Validate the cohort and counts before rendering: each returning account must belong to that cohort, be counted once per day and satisfy returned_accounts <= eligible_accounts. Supply a fixed 0–100 colour scale and translated text identifying cohort, return event and reporting window. For an empty cohort, the summary explains why there is no rate.
Dates must be complete and validated because absent cells render as zero; a missing observation is not confirmed churn. Weekday captions currently remain fixed English. Test the denominator, duplicate events and reporting cutoff, and provide exact cohort counts. The gallery has independent data.