You are paying commission on sales that were already yours
Affiliate and partner programmes pay on the last click before a sale, which rewards whoever stands closest to the till rather than whoever found the customer. The cookie was always a weak witness. Here is how to check a partner's claims against your own closed sales, and what to pay for instead.
Contents
The affiliate report arrives on the third of the month and it is the happiest document in the marketing department. Every line is a sale. Every sale has a partner's name beside it and a commission that was only paid because the sale happened. There is no wasted spend, no impressions that went nowhere, no clicks that bounced. It looks like the one channel where you only pay for results, which is exactly how it is sold.
The question the report does not ask is whether those results would have happened anyway. A customer who had already decided to buy, typed your name into a search engine, clicked a coupon site's listing to look for a discount code and then completed the order is, to the affiliate platform, a sale the coupon site produced. To you, it is a sale you already had, on which you have now paid eight per cent to somebody who met the customer at the door of the shop and walked them to the till.
Paying only when a sale happens is not the same as paying only for sales you would not otherwise have made. Affiliate reporting measures the first and is silent on the second.
How a last-click commission decides who gets paid
An affiliate programme credits a sale to whichever partner's tracking link was clicked last before purchase, within a cookie window that commonly runs for thirty days. The rule measures proximity to the sale. It carries no information about who introduced the customer or whether anyone needed to.
The mechanism is simple enough to describe in a sentence, and the simplicity is the problem. A visitor clicks a partner's link. A cookie is set naming that partner. If a purchase happens before the cookie expires, and no other partner's link has been clicked in the meantime, that partner is paid. Last one to touch it wins.
Consider what that rule rewards. It does not reward the blogger whose review persuaded somebody three weeks ago; her cookie was overwritten. It rewards the last click, and the last click before a purchase is overwhelmingly made by somebody who has already decided. So the rational strategy for a partner who wants to maximise commission is not to find new customers. It is to stand wherever people who have already decided will pass.
- Coupon and voucher sites, which rank for your brand name plus "discount code" and catch buyers between basket and checkout.
- Cashback and browser-extension tools, which activate at the checkout itself and set a cookie at the last possible moment.
- Partners bidding on your brand terms in paid search, intercepting people who were typing your name.
- Comparison and review pages built to rank for your brand plus "review", which is a search made by somebody late in the decision.
Every one of these is a legitimate business and some of them do introduce new customers some of the time. The point is structural: the payment rule cannot tell introduction from interception, so it pays for both at the same rate, and interception is far cheaper to produce.
The cookie was always a weak witness
An affiliate cookie records that a browser was tagged with a partner's identifier. It does not record that a person saw a recommendation, was influenced by it, or even clicked anything. Everything a last-click programme pays out rests on that one weak piece of evidence.
Most of the time the weakness shows up as mild distortion: a partner credited with a sale they had little to do with. At the extreme it is fraud, and the most instructive case is nearly twenty years old and involved one of the most sophisticated online businesses in the world.
Fraud on that scale is rare. The lesson generalises anyway. A programme's top performers are, by construction, the partners who are best at being the last cookie before a sale. Whether that is because they send you the most new customers or because they are best positioned between your existing customers and your checkout is a question the affiliate report is not equipped to answer.
Your best affiliate is whoever is best at holding the last cookie. Whether that is the same as your most valuable partner is a separate question, and the programme's own report cannot answer it.
The same company measured what interception is worth
Incrementality is the share of sales credited to a channel that would not have happened without it. It can only be measured properly by withholding the channel from a comparable group. Where that has been done for bottom-of-funnel placements, the incremental share has turned out far smaller than the reported one.
Affiliate marketing has few public, rigorous incrementality studies, which tells you something in itself. But there is a well-known experiment on a close cousin: paying for clicks from people who were already looking for you by name.
A partner sitting on your brand name, or appearing at your checkout, is in the same position as that brand-keyword ad. The customer was on their way to you. A reported conversion rate that looks miraculous is what it looks like when you measure a toll booth instead of a road.
Offline sales make it worse, not better
When the sale closes on a call, in a contract or at a counter, an affiliate or referral partner's claim is usually a registered lead or a declared introduction rather than a tracked purchase. There is no cookie to argue about, and nothing independent to check the claim against except your own records.
Everything above describes online checkout. For a business that sells by quote, consultation or contract, the affiliate relationship looks different and the problem is the same in a less technical costume. A lead-generation partner sends you enquiries and is paid per lead or per sale. A referral partner — an adjacent trade, a consultant, a reseller — registers a deal and claims commission when it closes.
What these arrangements share is that the partner's claim is the primary record. They say the customer came from them. And frequently the customer did also talk to them. But the same customer may have rung you directly a month earlier from a search ad, or be an existing account buying a second product, or have been in your CRM as a nurtured lead for a year. In a B2B partner programme this is the perennial deal-registration argument: the partner registered the opportunity on the fifth, and your own rep logged the first call on the second.
| The partner's report says | Your own records can show | What it means for the commission |
|---|---|---|
| Sale credited to us | Customer first appeared as a lead from another source weeks earlier | Interception; the introduction was somebody else's |
| Sale credited to us | Customer had bought from you before | An existing customer using a discount or a portal |
| Sale credited to us | No earlier trace of this person anywhere | A genuine introduction; this is what the programme is for |
| Deal registered on a date | First contact with the account predates registration | A registration race, not a referral |
| Lead delivered | Same phone number already delivered by another paid source | You paid twice for one enquiry |
The right-hand column is a commercial decision and yours to make. The middle column is a matter of fact, and it is sitting in files you already have.
Auditing partner claims against your own closed sales
A commission audit matches every sale a partner was paid for against your full lead and sales history, on normalised phone number and email, and asks one question of each: when did this person first appear in your records, and from where? The answer sorts commissioned sales into introductions, interceptions and existing customers.
You do not need the partner's cooperation or access to the affiliate network's internals. You need three exports: the list of sales on which commission was paid, with a customer contact detail and a date; your closed sales; and your lead history across every source, as far back as your sales cycle runs.
- Normalise phone numbers and email addresses on all three files to one format. Most apparent mismatches are formatting, not different people.
- For each commissioned sale, find the earliest record of that person anywhere in your lead and sales history.
- If the earliest record is the partner's own lead or click, mark it as an introduction.
- If an earlier lead exists from another source, mark it as an interception and note the source that actually found them.
- If an earlier purchase exists, mark it as an existing customer.
- Total commission paid under each heading, by partner. The partners separate into groups very quickly.
CloseRev is built to do this kind of match. It reconciles closed sales against lead and call exports on normalised phone and email, with every match auditable and nothing counted unless the match is high-confidence — so the partner file can go in as one more source, and you can see, sale by sale, which lead record each closed sale matched and when that lead was created. Sales it cannot match stay in Direct / Unknown.
A word on fairness, because this exercise can be run as a witch-hunt and should not be. Finding that a customer was already in your records does not prove the partner added nothing. A lead that had gone cold for eight months and was revived by a partner's recommendation is worth paying for. The audit does not make that judgement for you. It puts the dates in front of you so the judgement can be made on evidence, by partner, instead of by the programme's default rule for everybody.
What to pay for instead
Commission terms can pay for what a business actually wants — new customers — by setting different rates for new and existing customers, excluding sales another channel already produced, and barring brand-term bidding. Partners who introduce customers earn more under such terms, and those who intercept earn less.
Most programmes run on default terms because nobody had the data to justify anything else. Once an audit exists, the terms can follow it.
- Pay full commission for customers with no prior record, and a reduced rate or none for existing customers. Most networks support new-versus-returning rates; few advertisers switch them on.
- Exclude or discount sales where your own records show an earlier lead from another source within a defined period. State the rule in the terms so nobody is surprised by it.
- Prohibit bidding on your brand terms, and check. It is the cheapest interception available and the hardest to justify.
- Shorten the cookie window for checkout-stage partners and lengthen it for content partners. A reviewer's influence lasts weeks; a coupon's lasts minutes.
- For referral and reseller programmes, define first contact by your own records, not by the registration date, and say so up front.
- Require clear disclosure of the commercial relationship in your programme terms, and spot-check it.
Expect some partners to leave when the terms change. The ones who leave are telling you which business they were in. The ones who stay, and the content partners who suddenly find their three-week-old recommendation is being paid for, are the programme you thought you had.
What an audit cannot settle
A match against your own records shows who reached a customer first and whether they were already yours. It cannot prove what would have happened without the partner; only a holdout experiment can, and most programmes are too small to run one. The audit is evidence for a commercial decision, not a verdict.
It is worth being straight about the limit. A customer with no prior record who arrived through a coupon site might have found you anyway the following week. A customer with a year-old lead record might never have come back without the partner's nudge. Your records cannot see the counterfactual, and nobody else's can either without an experiment.
But compare where that leaves you with where you started. The programme's own report said every commissioned sale was the partner's doing, at one flat rate, on the evidence of a cookie. After the audit you know, for every sale, whether that person was new to you, already a lead from a channel you were also paying, or already a customer — and you know it from records a sceptical finance director can open and check. That is not certainty about incrementality. It is a great deal better than paying eight per cent to whoever was standing nearest the till.
You cannot prove what a partner caused. You can prove who reached the customer first, and that is enough to stop paying commission on sales that were already yours.
Questions people actually ask
- What is wrong with last-click affiliate attribution?
- It pays whoever held the final click before the sale, regardless of whether they had anything to do with the customer deciding to buy. That rewards being near the checkout — coupon pages, cashback tools, brand-name search ads — over the harder work of introducing somebody who had never heard of you. The commission is real money, so the incentive it creates is real too.
- How do I know whether an affiliate sale was incremental?
- Strictly, only by experiment: withhold the partner from a comparable group and see whether sales fall. Short of that, your own records give strong evidence. If the customer was already in your CRM as a lead from another source before the affiliate click, or was an existing customer, the partner did not find them. Matching commissioned sales against lead history on phone and email answers that for every sale, not a sample.
- What is cookie stuffing?
- It is the practice of planting an affiliate tracking cookie on a visitor's browser without them clicking an affiliate link, so that any later purchase pays the affiliate a commission they did nothing to earn. It has been prosecuted as wire fraud. It is the extreme case of a general weakness: a cookie records that a browser was tagged, not that a person was persuaded.
- Does this apply to B2B partner and referral programmes too?
- More so, because the deals are larger and the evidence is thinner. A reseller or referral partner registers a deal and claims the introduction; often the account was already talking to your sales team, or had come in through a campaign months earlier. The check is the same: compare the partner's claimed date against the first date that contact appears anywhere in your own lead and sales records.
- Should I stop paying affiliates who fail the check?
- Not automatically. Change what you pay for. Pay full commission for customers who are new to your records, a reduced rate or nothing for existing customers and for leads another channel already produced, and nothing for sales through your own brand terms. Partners who genuinely introduce people do better under those terms, which is how you tell them apart from the ones who do not.
- Do affiliates have to disclose that they are paid?
- In the United States the Federal Trade Commission's guidance is that they should: a relationship that pays commission on sales is a material connection and ought to be disclosed clearly and conspicuously near the recommendation. That is the affiliate's obligation and, in the regulator's view, something the advertiser should be monitoring. It is worth a line in your programme terms. This is not legal advice, and the rules differ by country.