Affiliate Fraud and Traffic Quality: How to Spot Bad Traffic Early

On a UK home-services lead generation campaign paying £8 a lead, one publisher's volume went from about 250 leads to 500 in a single week. The advertiser's feedback said quality was slipping, so we held off on scaling the source and looked closer.

Contactability, the share of leads the call team could actually reach, had fallen from around 60% to 35%. This guide covers how we found 200 invalid leads, what happened to the publisher, and the controls I now put on every programme.

The case: 200 leads that didn't add up

First, I made sure the drop was real. We compared leads after a similar follow-up period and the same number of contact attempts, so we weren't comparing fresh leads with old ones. The gap held.

Then I broke the data down by publisher and by sub-publisher. One sub-source stood out:

• the same contact details appearing again and again
• submissions bunched into unusually tight time windows
• leads from outside the locations approved for the campaign

We checked those findings against the advertiser's validation feedback and stopped taking traffic from that sub-source while we investigated. In the end, 200 leads were confirmed invalid against the agreed campaign criteria. At £8 each, that's about £1,600 in invalid payouts that were never made.

I'm careful with the word "fraud" here. These leads were confirmed invalid under the campaign rules. Calling every one of them deliberate fraud would need more evidence than we had, and it's not a claim I'd make to a partner without it.

What happened to the publisher

We shared the findings with the publisher. They removed the sub-source and explained the checks they'd be adding on their side. We then allowed a controlled restart, capped at 25 leads a day for two weeks.

Contactability came back to about 58%, close to where it had been before the spike. So we kept working with them, under closer monitoring, with the problem sub-source still excluded. We also added duplicate checks on our side and closer monitoring for any new source in its first weeks.

Where the partner co-operates, I'd rather fix the source and restart slowly than cut the publisher off, because cutting them off loses their good traffic too.

What I'd do differently

At first we paid too much attention to the jump in volume. With hindsight, I'd have brought the advertiser's contactability and validation feedback into the review earlier. When any source doubles in a week, that's now the first thing I ask for.

Six kinds of bad traffic and how each one shows up

Fake or low-quality leads. Made-up details, recycled data, or people pushed into signing up for an incentive. They show up as repeated or patterned details, leads bunched at odd hours or from outside your area, and falling contactability. This is the biggest risk on CPL campaigns, and it's what the case above looked like.

Cookie stuffing. A partner drops their tracking cookie on people who never saw their content, then takes the commission when those people buy another way. Look for very short click-to-sale times, again and again, from one partner.

Brand bidding. A partner runs paid search ads on your brand name against your rules, so you pay commission on customers who were already looking for you. Search your own brand name on a few devices and see whose ads appear.

Voucher code leakage. An exclusive code ends up on public coupon sites, and a partner gets paid for the last click on a sale you'd have made anyway. It usually shows when an exclusive code starts appearing in sales from partners it was never given to.

Bots and click fraud. Automated traffic that clicks, installs or fills forms to trigger payments, mostly on CPC and CPI deals. Expect unexplained spikes and many actions from the same device or IP address.

Undisclosed incentivised traffic. People paid in points or rewards to complete an action. Fine when it's agreed and priced in; a problem when it isn't. Watch for conversion rates far above the programme average from a new partner, followed by high returns.

Controls to put in place from day one

1. Write clear programme terms. Spell out which traffic is allowed, including brand bidding, voucher sites, email, incentives and paid social, and what happens if the rules are broken. There's a checklist of what to agree in our guide to choosing a pricing model.

2. Approve partners, don't auto-accept them. Look at every applicant's website or channel. If you can't see where their traffic will come from, ask.

3. Validate before you pay. Use a validation window that covers your returns or cancellation period, and decline what doesn't meet your rules. The major networks all support this, but it only works if someone does it.

4. Protect your brand and your codes. Check your own brand search regularly, use unique or single-use codes for exclusive offers, and look for codes on coupon sites that you never published.

5. Add basic lead checks. Validate emails and phone numbers on submission, block duplicates and add a simple bot check.

6. Report by sub-source and get the advertiser's quality data early. In our case, contactability and validation feedback showed the problem before the volume numbers did. Ask for them weekly, and daily when a source is scaling fast.

When you find a problem

• Make sure the drop is real by comparing like with like, with the same follow-up period and contact attempts.
• Pause the affected source while you investigate.
• Share the evidence with the partner and ask them to explain the traffic. Some issues are honest mistakes.
• Decline what's invalid under your terms. If the partner fixes the problem, restart with a daily cap and watch quality closely before lifting it.
• Keep records and act the same way every time, whoever the partner is.

We pay on validated results only in our affiliate and publisher marketing work, so this kind of checking is routine for us. If you'd like a second opinion on the traffic you're paying for, get in touch.