Fraud intelligence that flags, not bans

Affiliate fraud is quiet: self-referrals, coupon leakage, bot clicks, duplicate accounts farming a signup bonus. affixo watches for all of it automatically and surfaces high-confidence alerts — without ever auto-banning a legitimate partner.

Identity signals → a risk score

Every affiliate accrues an append-only stream of signals (signup, login, visit), each capturing the context that matters for fraud:

• IP, geo, ASN/ISP, and VPN / proxy / Tor detection
• device fingerprint, OS, screen, timezone, language
• normalized email (with disposable-address detection), payout email, phone

Those roll up into an affiliate identity with two scores, 0–100:

• Risk score — self-referral patterns, bot-flag rate, chargebacks, VPN use, disposable email, duplicate-cluster penalty and velocity.
• Quality score — the inverse, minus coupon-leakage penalties.

High-confidence duplicate detection

affixo clusters affiliates that look like the same person gaming the program, matched on normalized email, payout email, device fingerprint and phone:

• High — strong matches (email + fingerprint, or payout email).
• Medium — fingerprint or payout-email alone.
• Low — IP or name only.

It flags the cluster for review — it never bans on its own.

Control the channels, not just the people

Some traffic is against your rules before any fraud math runs. Channel policy lets you allow, hold or block specific channels — brand bidding, coupon/incentive sites, paid search, paid social, email and more — set per workspace with per-campaign overrides.

One program's fraud protects every other

Two shared, cross-program risk lists turn affixo into a fraud network rather than an isolated filter — both advisory-only, so you always make the final call:

• Network Risk List — confirmed-fraud signals (hashed email, IP, device fingerprint, phone) shared across every program on affixo. A known bad actor arrives in your program pre-flagged, even with zero history there.
• Buyer Risk List — the same idea, pointed at the real risk in a commission model: the buyer. When a buyer loses a chargeback dispute at one program, that buyer — matched by hashed email and card fingerprint, which survives them switching email addresses — is flagged at the moment they purchase through any other program, before you pay commission on revenue you'll never keep.

Neither list ever exposes another program's data: signals are stored as one-way hashes, and your program sees only an aggregate hit count and severity — never who reported it or any raw detail.

You're told at the moment of sale, in the email you already get

A flag is worthless if you find it a week later, in a console you had no reason to open. A buyer match lands in the same commission notification that tells you the sale happened — while the commission is still pending, and while denying it is still one click:

Notice what the alert does not do. The sale completed. The commission is untouched. The buyer wasn't blocked, and your affiliate wasn't accused of anything. affixo hands you the signal and the decision stays yours — that's what advisory-only means in practice.