# Evaluate partner quality

Compare referral cohorts using exact currency amounts and observed coverage.

## What you need

Permission for the selected apps/programs and the intended reporting window. Financial and lifecycle source coverage must be inspected alongside results.

## Steps

1. Open **Affiliates → Partner quality** and choose the supported app, date and cohort filters.
2. Read accepted installs, 30-day collection conversion, 30/90-day installation retention, net collections, accrued referral commission and **Collections less referral commission** in the same currency and cohort.
3. Inspect eligibility, coverage/unknown reasons and optional usage observations before ranking partners.
4. Drill into the partner/referral records behind a surprising number. Compare equivalent maturity windows rather than new and mature cohorts without context.
5. For an integration, use [get_affiliate_partner_quality](/docs/mcp/get-affiliate-partner-quality) with the correct owner or scoped read credential.

## Expected result

A source-backed partner report with per-currency values and observable coverage. Collections less referral commission is a comparison, not full business profit; acquisition cost and other operating expenses are not invented.



## Troubleshooting

Missing usage data stays unknown; it does not prove no activation. Unallocated rewards, unmatched refunds, incomplete history and immature cohorts need their own review. Do not treat correlated retention or a message sent as causal uplift.

## Cohorts and denominators

The installation cohort uses an inclusive UTC installedFrom and exclusive installedTo. The supported input defaults to the last 90 days and allows at most 365 days; page limit defaults to 25 and allows 1–100. Cursors bind to the same filters and allowed app scope.

30-day collection conversion counts a positive subscription, usage or one-time collection within 30 days among mature installations with complete coverage. Later refunds do not erase the first collection. Retention measures installed state at the exact 30/90-day horizon, not product activation or paid retention. Samples below ten are labeled; immature, unknown and stale observations remain separate.

Accrued commission includes signed earning/refund corrections even while held. Payment reports/confirmations do not add commission cost a second time. Unallocated rewards remain separate and unlike currencies are never combined.