Referrals & fraud review

The referral lifecycle

Every commission is a referral record moving through states: pending (created when a referred order lands, waiting out the maturity window), held (flagged by fraud review, waiting for your decision), approved / payable (matured and clean — counted in the partner's balance), rejected (refunded order, failed review, or your manual decision), and paid (included in a completed payout).

Affiliates → Referrals lists them all — filter by status or affiliate, open any referral to see the order it came from, the originating visit, and the amount explanation. Held referrals show a badge count in the menu so they never rot unseen.

Explainable fraud review

Automated fraud filters usually fail in one of two ways: silently rejecting good referrals (partners notice, trust dies) or approving everything (you pay for fraud). This plugin takes a third path: suspicious referrals are held, not rejected, and every hold comes with a plain-language reason — for example a self-referral pattern, an implausible click-to-order gap, or a burst of referrals from one source.

You review the held queue and decide: approve (reason was innocent) or reject (it wasn't). Your decision is recorded in the activity log with the reason attached, which is exactly the paper trail you want if a partner disputes it — and it's why holds beat silent rejection for partner trust, a point argued at length in preventing affiliate fraud.

Common patterns to know

Self-referrals — a partner buying through their own link to discount their purchase. Most programs ban it in their terms; the review reasons make it visible. Coupon leaks — a personal code escapes to a deal aggregator and starts claiming full-price organic sales; watch for a partner whose code volume suddenly dwarfs their link volume. Refund cycling — referred orders that reliably refund after maturity; the refund handling in commissions plus a longer maturity window contains it. The fraud glossary catalogs the full zoo.

Tuning strictness

Settings → Fraud review controls how aggressive the holds are. Start with defaults; if your held queue stays empty for months, tighten a notch — if it's full of false positives, loosen one. Either way, nothing is paid without passing review and maturity, so tuning is about your review workload, not about risk of silent loss.