Affiliate fraud, coupon abuse and brand-bidding enforcement

On the programs we inherit, between four and eleven percent of paid commission fails a first test: the click and the order sit less than ninety seconds apart, or the click's referrer is a browser extension that fired while the customer was already on the checkout page. Most of that is not fraud in any criminal sense. It is a rule nobody wrote, and inside your current clawback window it is still recoverable. How the detection works: In your tracking database, not in a vendor's score. Click-to-conversion time distribution per partner against the program median, because bimodal partners are the ones to open first. Device and IP clustering across a partner's conversions. Referrer chains showing a direct-navigation session with an affiliate click stapled to the end of it. Coupon codes issued privately to one partner that are now indexed on Cuponomia, Méliuz or Pelando. Self-referral, found by matching partner registration against order data, including the CPF and delivery-address matches that a name comparison alone will miss. Brand bidding: Monitored on a schedule, on Google Ads and Bing, from Brazilian and target-market IP addresses rather than from wherever a tool happens to run. Your exact brand term, common misspellings, and brand-plus-coupon queries. Screenshots timestamped, ad copy and destination URL captured, matched to the partner ID sitting in the landing URL. A verbal policy does not survive a dispute; a dated evidence file does, and partners settle quickly when they see one. Sector patterns we look for: In gaming, first-deposit-only cohorts, chargeback concentration, and traffic converting at a rate that has no plausible explanation and then never returning. In education, refunds clustered inside the platform's guarantee window from a single source, which is a payout-timing problem before it is a fraud problem and should be fixed as one. In industrial lead programs, form fills carrying valid-format CNPJs that belong to no operating company. Enforcement that holds: Findings arrive as an evidence pack per partner: what happened, the underlying rows, the term it breaches, and a recommended action. We draft the notice. When a partner disputes, we handle the exchange and we tell you when they are right, which does happen. When the rule was missing rather than broken, we write the rule and you apply it going forward instead of retroactively. Clawing back against a term that did not exist at the time is how programs lose the partners they most need to keep. What we will not do: Ban on a score with no evidence behind it. Hold earned balances past your published terms as an investigative tactic. Run a blanket policy against coupon partners, because some of them are the reason your abandoned carts recover and the answer there is a rate class rather than a ban. Present a finding we have not reproduced ourselves. Who should not buy this: Programs paying under roughly ten thousand dollars a month in commission — recovery will not cover the work and an annual review is the right size. Also anyone who wants a number to justify a decision already taken. If the brief is to find grounds for removing one specific partner, we are not the practice for it and will say so on the first call.