Foundations · Lesson 6 of 7
Affiliate fraud and traffic quality
About 11 minutes · Yogender Kumar, Managing Partner
By the end of this lesson you'll be able to
- Name the main kinds of bad affiliate traffic
- Spot the data patterns that point to each one
- Respond in proportion, from a question to a stop
Why this matters
In our experience most partners are honest. A few aren't, and some honest ones have a sub-source that isn't. Because you pay on results, bad traffic costs you twice: once in commission and again in the decisions you make from false numbers.
The core idea
Bad traffic comes in a handful of familiar forms:
- Fake or low-quality leads: made-up or recycled details, or people pushed into signing up.
- Cookie stuffing: a partner sets its tracking cookie on people who never saw its content, then collects commission when they buy another way. It shows up as lots of clicks with almost no time on the partner's page, and sales long after.
- Brand bidding: a partner runs paid search ads on your brand name against your rules.
- Code leakage: a code meant for one audience published on voucher sites.
- Unapproved browser extensions that appear at checkout and take the last click.
- Bots: automated clicks or form fills.
Each leaves a pattern in your data. The skill is knowing what normal looks like, so the odd pattern stands out.
Worked example
Hartwell Outdoor, illustrative figures. A new content partner joins and within a month sends 600 sales, more than the next three partners together. At £8 a sale, that's £4,800 in commission.
Hartwell's manager looks at the time from click to sale. For most partners, about 5% of sales happen within a minute of the click. For the new partner it's 70%: 420 of the 600 sales.
People rarely read a review, click and complete a full order inside a minute. A pattern like this usually means the partner's link fired just before an order the shopper was already placing. That fits a browser extension or coupon pop-up taking the last click at checkout.
Hartwell doesn't accuse anyone. It asks how the traffic is generated, and holds the partner's pending commission while it checks. It checks the partner's site on several devices. The partner can't explain the pattern, so Hartwell declines the 420 sales under its programme terms, which ban extensions it hasn't approved, and removes the partner. That saves 420 × £8 = £3,360, and stops the false numbers from making the partner look like Hartwell's best.
Share of sales within a minute of the click
How we'd respond, in order
- Ask: share the pattern and ask the partner to explain it.
- Hold: pause pending commission while you check.
- Cap: limit volume from a source you're unsure about.
- Decline: reject sales that break your written terms.
- Remove: end the partnership when the explanation doesn't hold.
The case study in the next lesson shows this process on a real lead generation campaign.
Responding in proportion
From lightest to strongest
In practice
- Check click-to-sale times by partner every month.
- Search your own brand name on a few devices each week, and see whose ads appear.
- Give each partner its own codes, and watch where they turn up.
- Look below partner level, at sub-IDs, whenever volume jumps.
- Put each rule in your programme terms, so declines are backed by something written down.
Common mistakes
- Celebrating a spike before checking it. A partner that doubles in a week deserves a look before a bigger cap.
- Calling it fraud without evidence. Say what the data shows and what the terms say. Let the partner explain.
- Rules that only exist in someone's head. You can only decline sales for breaking terms you've written down.
Try it yourself
A partner sends 300 sales in a month at £6 commission each. You find that 60% of them happened within a minute of the click, against 5% for your other partners. The partner can't explain it, and your terms allow you to decline sales taken by unapproved checkout extensions. How many sales would you decline, and how much commission does that save?
Show answer
60% of 300 is 180 sales. Declining them saves 180 × £6 = £1,080. The other 120 sales are less clear. Many brands would hold them too while deciding whether to keep the partner.
Quick check
Not scored. Pick an answer to see the explanation.
Which pattern most suggests an extension taking the last click at checkout?
What's the right first step when a partner's data looks wrong?
Key takeaways
- Each kind of bad traffic leaves a recognisable pattern in your data.
- Know what normal looks like for your programme, so odd patterns stand out.
- Respond in order: ask, hold, cap, decline, remove.
- Write every rule into your terms before you need it.
Further reading: the full guide on our blog

