How PVOS outcomes are counted.
These show what members did, not that PVOS caused it.
Who is counted
Members are grouped by the month they signed up. Every member who started counts, including those who stopped: a commitment left open past its date counts as not kept, and a member who stopped checking in counts as not checking in. This is called intent-to-include.
Members under 18, members whose age is not confirmed, classroom accounts and the owner’s own accounts are left out entirely; they are never profiled.
The small-number rule
A figure built on fewer than 10 members is withheld, not shown. No number appears for a cohort until it passes that threshold. Totals add up published figures only, so a withheld one cannot be worked out from the rest.
What is counted
Keep rate
Of the commitments a cohort’s members made that came due within their first 1, 3 or 6 months, the share the nightly scorer had recorded kept by then. A commitment still open past its date, or corrected to not kept, counts as not kept. Released and undated commitments, entries written in crisis language and Circle parts a member marked kept themselves are left out.
Repair rate
Of a cohort’s missed commitments, the share made right within 30 days of the miss: a comeback whose commitment the scorer then recorded kept, the same rule Standing uses. A miss is counted once its 30 days have passed.
Check-in retention
Of every member who signed up in the month, the share who checked in on a commitment during week 4, 12 or 26 after signing up. Members who never made a commitment, or stopped, count as not checking in.
Say/do gap
For members with a weekly Value Stack reading in their first 30 days (month 0) or near day 182 (month 6, within 14 days), the average distance between the ranking they stated and the ranking their actions revealed: the mean, over the civic values, of how many places apart each value sits. Lower means closer. Readings are the weekly snapshots the nightly pass keeps; nothing is reconstructed.
Witness confirmation
Of a cohort’s commitments with at least one witness who answered, the share where a witness answered Yes. A witness who agreed but never answered is not a No: those commitments are shown as a separate count.
No cohort has passed the threshold yet, so no figure is shown.
What the data cannot show
- Deleted accounts cannot be counted. When a member deletes their account, their commitments, check-ins and readings are deleted with it, so they are not in any figure.
- Cohorts from before the weekly Value Stack snapshots began have no month-0 reading, so they have no say/do gap at month 0. Nothing is estimated in its place.
- Keep rate, repair rate and witness confirmation are shares of commitments, so they cover only members who made a commitment. Check-in retention is the one measure that counts every member of the cohort.
- These show what members did, not that PVOS caused it. There is no comparison group of people who did not use PVOS.
Figures are recomputed each night from records the platform already keeps. Nothing about any one member is published.