How do I audit an existing affiliate program for fraud, leakage, and dead partners?

May 29

To audit an existing affiliate program for fraud, leakage, and dead partners, pull 12–24 months of partner-level data, segment by media type and vertical, and check three things: where conversions look suspicious, where revenue pays out twice or to the wrong source, and which partners haven't driven a real conversion in 90+ days. Most of what gets labeled "affiliate fraud" is actually measurement debt — last-click hides the truth, and partners that should have been cut a year ago still earn payouts. But sometimes it really is fraud, and I'll point to a live example from a program I run today later in this post.

Here's what I look for, in order.

What's the difference between fraud, leakage, and dead partners?

Three problems, three fixes. Separate them before you dig.

Fraud is intentional: conversions that aren't real customers — fake signups, click farms, brand-search hijacking, cookie stuffing, coupon abuse. The tell is high signup volume with almost zero activation or retention. You're paying for ghosts.

Leakage is structural: real conversions attributed to the wrong source, paid twice, or paid on traffic you'd have gotten anyway. The classic case is a partner running paid search on your brand terms — you pay commission on a sale you already owned. Impact, an affiliate network, has tooling for this, but it's off by default in most accounts I inherit.

Dead partners joined and stopped, or never started. Across the B2B SaaS programs I've audited, 60–80% of the inherited roster sits at zero conversions over the trailing 12 months. That's not fraud — it's neglect. POV holds on every audit I run: most affiliate programs are bloated because no one is actually running them.

What kinds of affiliate fraud should I look for first?

Four patterns surface fastest. None need a forensic tool — just pull the data and look.

Signup-only payouts with no activation tail. If you pay on raw signups, you attract junk. Cohort the last 12 months by source and check activation at 7, 30, and 90 days. Pay partners on activation, upgrade, or paying customer — not raw signups. Signup payouts create the fraud problem in the first place.

A real one from a B2B SaaS program I run right now. A partner that joined in November 2025 looked like my best new recruit on paper: 199 signups that month, scaling past 1,600 by March. The funnel underneath told a different story. Activation — people actually using the platform — fell from 11% to 2% as volume climbed. The metric that flagged it was the ratio of paying customers to activated users. In November, 36% of activated users converted to paid. By March, paying customers outnumbered active users more than six to one. People were paying without ever touching the product: conversions engineered to trip the CPA payout while skipping the part where someone actually adopts the software.

Brand-search hijacking. Check the landing-page paths on a sample of affiliate conversions. If clicks arrive through paid-search URLs containing your brand name, publishers are bidding on your trademark and intercepting demand you already had. Most networks let you ban this in the terms — enforcement is the job.

Coupon code abuse. Mostly DTC, but it appears in B2B SaaS wherever promo codes exist. A code tied to one publisher shows up on extension toolbars and aggregation sites, and that publisher collects commission on the whole converted-coupon population. On a DTC women's apparel brand I co-founded, the cohort math showed flat-fee creator partnerships beat coupon-tied ones on both first-order ROAS and repeat rate.

Cookie-stuffing. A "publisher" drops your tracking cookie on users who never see real content, then collects credit when they buy through any channel. It's less common than it used to be, but the check is simple: flag partners with abnormally high click-to-conversion ratios relative to their visible content footprint. Custom dashboards surface this where a default network report won't — the kind I built on Google Sheets fed by Salesforce for an enterprise B2B SaaS HR/IT/finance program.

How do I find revenue leakage in an existing program?

Leakage is harder than fraud because it doesn't look wrong on the dashboard. It looks like the program is "working." Four places I check:

Last-click masking the real source. Last-click revenue is a lagging metric, not a strategy. If most "affiliate-driven" revenue converts after five other touches, your affiliate is closing, not sourcing.
Double-attribution with paid search or paid social. If the same user converts in both, you're paying twice. Discount stacking — public coupon plus affiliate code plus a Klaviyo email discount — hides the same way until someone runs the margin math.
Partners paid on organic revenue. Trademark bidding is obvious; review-site SEO interception is subtler — and sometimes that placement genuinely drove the demand. The audit tells the difference.
Stale CPA structures. Sometimes you're paying 20% where 12% holds the top partners — or where the math has room to raise CPAs on partners sourcing real new revenue. Payout is a lever in both directions, not just down.

One lever worth knowing: Impact, an affiliate network, has a Credit Group setting — Preferred, Standard, Non-Preferred — that decides which partner gets credit first when several touch the same conversion, so a coupon site can't take last-click from a high-intent review partner. The catch is scope. Credit Group only orders credit within the affiliate ecosystem. It won't reconcile affiliate against paid, organic, or lifecycle — that happens only when affiliate data lives in the same infrastructure as the rest of your stack. When it's split across separate layers (a BI tool like Looker, product analytics like Amplitude, a Databricks warehouse), the affiliate numbers won't match the company's source of truth 1:1, and that gap is where cross-channel leakage hides.

How do I identify and handle dead partners?

The easiest part to run, and the part most programs avoid because it requires a decision.

My definition: a partner is "dead" with fewer than X conversions in the last 90 days, where X is set against average partner contribution — usually zero to three. Sixty to eighty percent of the roster falls below the line. What to do depends on why:

Never activated. One reactivation outreach. No response in 30 days, deactivate.
Activated and decayed. Worth a real conversation — economics, content freshness, what changed. Some recover with a calendar invite, not a campaign.
Activated but in a vertical or media type you're deprioritizing. Deactivate, document why, move on.

The cleanup mechanics — deactivating hundreds of partners without breaking tracking, handling pending commissions — are the subject of a forward post in this series, How do I clean up a bloated affiliate program with hundreds of inactive partners?

Can the affiliate network help detect this?

Partially. Impact has reasonable fraud tooling and can flag trademark-bidding violations. PartnerStack's monitoring is lighter, but its B2B SaaS partner mix is less fraud-prone anyway — it's harder to fake an enterprise SQO than a DTC signup. Off-platform setups catch patterns the network tooling misses, because you control what's measured.

What no platform solves is dead partners — the dashboards lean on last-click by default, so inactive partners look the same as active ones. That's the POV in practice: bloat is a neglect plus weak-measurement problem, not a network-rep problem.

How long does an audit take, and what's the output?

On that B2B SaaS HR and time-tracking audit — a few hundred partners, ~24 months of history — the audit phase ran about two week. What I deliver:

A partner-level scorecard segmented by media type AND vertical (review sites, comparison sites, creators, podcasts, newsletter operators, aggregators, niche industry blogs)
A fraud-and-leakage findings doc with named partners, dollar exposure, and recommended action
A dead-partner deactivation list with reactivation outreach copy for the salvageable cohort
A payout recommendation — where to raise CPAs on high-quality sources and where to move to down-funnel events
A 60–90 day restart plan if the findings justify one — the same plan behind the 16.7x revenue growth and 1,425% paid conversion lift, and the shape of the 30-day rebuild that hit 2.5x ROAS in month 1.

If your program isn't broken so much as plateaued, read about how to restart growth.

FREQUENTLY ASKED QUESTIONS

Is most affiliate fraud actually fraud? No. Most of what gets labeled "fraud" is measurement debt — last-click hiding the real source, signup payouts attracting junk, dead partners earning a trickle. Real fraud exists and matters, but it's the smaller share of what an audit finds.

What's the difference between affiliate fraud and affiliate leakage? Fraud is intentional — fake signups, click farms, brand-search hijacking, cookie stuffing. Leakage is structural — real conversions paid twice, attributed to the wrong source, or paid on traffic you'd get organically. Fraud needs removal and reversed commissions; leakage needs a policy or tracking fix.

Do I need to switch networks after an audit? Usually no. Audits point to operating-model fixes — payout structure, partner mix, measurement, recurring partner calls — that work on your current platform. A network switch is a separate decision driven by audience fit, not by audit findings alone.

About + how to work together

I've audited, rebuilt, and built affiliate programs across B2B SaaS, enterprise platforms, and DTC ecommerce. I work on Impact and Profound simultaneously — most affiliate consultants only think about revenue; I think about your program as both a revenue channel and an AI visibility channel.

If your program is stuck, get in touch. The first 30-minute call is free.