Q: One simple fraud rule that paid for itself?
A velocity check on new accounts. For a program with a free-trial-to-paid flow, we were paying commission on signups, and one cluster of partners was farming fake trials.
What we built:
— Flagged any partner whose conversions came from 5+ signups sharing one IP range in 24 hours
— Held those commissions for manual review instead of auto-approving
— Cross-checked flagged signups against trial-to-paid conversion (real users convert ~22%; the flagged cohort converted at 1%)
Result: we caught and withheld about $7,300 in fraudulent commissions in the first month, then blocked two partner accounts driving most of it. Legitimate partner payouts were untouched.
The caveat: IP velocity catches lazy fraud, not sophisticated rings using residential proxies. It's a first filter, not a wall. Pair it with a conversion-quality check (do these signups ever pay?), which is much harder to fake.
Got a question? Send it in.
Program Desk
@ProgramDesk
Q: One simple fraud rule that paid for itself?
Этот пост опубликован в Telegram-канале Program Desk. Подписаться можно по ссылке: @ProgramDesk.