Pull up your affiliate dashboard and the number looks great: revenue is up, the channel reports a healthy return, and the partner roster keeps growing. Now ask a harder question — how much of that revenue would have happened anyway? For most programs at the $10M–$100M stage, the honest answer is uncomfortable. A meaningful share of what the channel claims is not new business at all. It is sales you had already won, credited to a partner who appeared in the final second before checkout.
This is not an argument against affiliate marketing. Done well, it is one of the few channels where you pay for outcomes rather than impressions, and where genuine content and creator partners can build demand you could never buy. It is an argument against measuring it the way almost everyone still does. If you cannot say which partners create demand and which merely intercept it, you are not running a performance channel — you are running a rebate program with extra steps.
What incrementality actually means.
Incrementality is a deceptively simple idea: it is the revenue that would not have happened without the partner. Not the revenue the partner touched. Not the revenue that carried the partner's cookie at checkout. The revenue that exists because of them and would vanish if they did not.
The gap between revenue the partner touched and revenue the partner caused is where the money leaks. Affiliate platforms almost universally report the former, because it is what their tracking can see. They credit the last partner whose link or cookie was present before the order. That is fine for partners genuinely introducing new customers. It is a disaster when the last touch is structurally rigged to land on partners who add nothing.
“The question is never did this partner appear before the sale? It is would this sale exist if the partner did not? Almost no dashboard answers the second question.”
Where the inflation comes from.
Three partner types account for most of the non-incremental revenue in a typical program. None of them are necessarily acting in bad faith — they are responding rationally to a measurement system that rewards proximity to the sale rather than contribution to it.
Cashback and loyalty partners. A customer has already decided to buy. On the way to checkout they pass through a cashback or loyalty site to claim a few points or a small rebate. That click drops the last cookie, and the partner collects a commission on a sale that was already yours. Some of these partners do drive genuine discovery; many simply sit at the bottom of the funnel and tax conversions you had already earned.
Coupon browser extensions. This is the sharpest example. A customer with a full cart sees the promo-code field and searches "[your brand] coupon." A browser extension — or the search result it feeds — injects itself, drops a cookie, and claims last-click credit. The customer was already buying; the extension contributed nothing but a code (often one of your own). This coupon-extension hijacking is one of the most efficient ways to bill a brand for demand it generated itself.
Brand-bidding affiliates. Here a partner buys your own trademark in paid search — brand bidding — and intercepts people already typing your name into Google. They were coming to you anyway. The affiliate inserts a paid link, collects the click, and bills you a commission, often while also driving up the cost of your own brand keywords. You pay twice to reach a customer who was already at your door.
Deal and discount-aggregator sites round out the picture. A subset send genuinely new buyers; many mostly harvest existing intent from people already hunting for a price on a product they had decided to purchase.
What it costs you.
The damage is not limited to a line item. When non-incremental sales are counted as wins, three things follow, and each compounds the others.
- You overpay for sales you already had. Every commission on a non-incremental order is pure margin erosion — you are buying a customer who was already buying.
- Your reported ROAS is a fiction. Blended channel numbers look strong precisely because free-riders inflate them, which makes the channel impossible to manage on the metrics you are looking at.
- Budget flows to the parasites. Because intercepting partners post the best apparent returns, naive optimization sends them more share — starving the content and creator partners who actually create new demand.
- You misjudge the whole channel. Leadership either over-invests in a channel that is quietly bleeding margin, or kills a channel that has a genuinely strong incremental core buried under noise.
How to measure incrementality.
You do not need a data-science team to start. You need a willingness to run tests that can prove a partner wrong — and to act on the results. Here are the methods that work, roughly in order of rigor.
New-vs-existing customer reporting. The single highest-leverage move. Split every affiliate order by new-vs-existing customer status. A partner whose revenue is overwhelmingly existing customers is, almost by definition, not creating new demand — it is being paid to be present at repeat purchases.
Holdout, geo, and time-based tests. The cleanest causal read. In a holdout/geo test, you pause a partner (or suppress it in selected regions or windows) and watch what happens to baseline revenue. If total sales barely move, the partner was riding demand that existed without it. If baseline drops, the partner was genuinely incremental. Matched-market tests — pairing similar regions and running the partner in one but not the other — sharpen this further.
Multi-touch versus last-click comparison. Rerun the same period under a multi-touch view and a last-click view. Partners whose credit collapses when earlier touches are counted are the ones living off the final click. The size of that collapse is a direct map of where your inflation sits.
Coupon-code and on-site behavior analysis. Look at how the partner's buyers actually behave. Orders that cluster on the checkout page after a coupon search, with near-zero on-site browsing beforehand, are a behavioral signature of interception rather than discovery.
Active compliance monitoring. Run automated checks for affiliates bidding on your trademark in paid search, and audit for toolbar and extension cookie injection in the final click. You cannot manage abuse you are not watching for, and the offending partners will not flag themselves.
An incrementality audit you can run this quarter.
How to fix the program.
Measurement tells you where the leaks are. Program design closes them. The goal is an economic structure where the only way for a partner to make money is to do something genuinely incremental.
| Inflation source | Why it is not incremental | How to test | How to fix |
|---|---|---|---|
| Cashback / loyalty | Customer had already decided to buy; partner sits at the bottom of the funnel | Holdout test; new-vs-existing split | Reduced rate for existing customers; weight toward genuine new-customer acquisition |
| Coupon browser extensions | Injects a cookie at checkout on a sale already in motion | Audit last-click for extension injection; coupon-page behavior analysis | Terms banning toolbar/extension last-click injection; de-duplicate against the session |
| Brand bidding | Intercepts people already searching for you | Paid-search monitoring on trademark terms | Explicit trademark-bidding ban; enforce with monitoring and removal |
| Deal / aggregator sites | Mostly captures existing price-shopping intent | New-vs-existing split; multi-touch comparison | Pay full rate only for new customers; cap or tier existing-customer credit |
| Generic last-click free-riding | Credit lands on whoever is nearest the sale | Multi-touch vs last-click; geo holdout | Attribution de-duplication; commission weighting toward upper-funnel partners |
The fixes themselves are concrete and within your control. The most effective ones are structural, not cosmetic.
- Tighten program terms. Explicitly ban brand bidding and trademark abuse, and prohibit toolbar and extension last-click injection. Make enforcement a condition of payment, not a polite request.
- Introduce new-customer commission tiers. Pay the full rate for a genuinely new customer and a reduced rate for an existing one. This single change re-prices most of the free-rider problem automatically.
- De-duplicate attribution across channels. Use de-duplication so an affiliate cannot double-bill on a sale your paid search or email already drove. One sale, one credit.
- Weight commissions toward upper-funnel partners. Reward the content and creator partners who introduce your brand more richly than the ones who appear at checkout — commission should track contribution.
- Make compliance monitoring continuous. Trademark bidding and extension injection are not solved once. Audit on a recurring cadence and remove repeat offenders.
None of this is exotic. It is the same discipline you already apply to paid media, extended to a channel that has been allowed to grade its own homework. If you are building or rebuilding the program from the ground up, the time to bake in new-customer tiers and clean attribution is at the start — see our guidance on launching a program and on partner quality, and weigh it against your program ROI targets.
Isn't some non-incremental credit just the cost of doing business?
A little, yes — no attribution model is perfect, and chasing the last fraction of a percent is not worth the effort. The problem is scale. When a large share of a channel's reported revenue is non-incremental, it is no longer rounding error; it is a structural mispricing that quietly redirects budget away from the partners actually growing the business. The goal is not perfection. It is making sure your biggest commission dollars are going to genuine lift.
How is incrementality different from attribution?
Attribution divides credit for sales that happened. Incrementality asks whether those sales would have happened at all without the partner. A coupon extension can win 100% of the attribution credit for an order and contribute 0% of its incrementality. Attribution is bookkeeping; incrementality is causality. You need the second to manage the channel, and only experiments — not attribution models — can measure it cleanly.
Are cashback and coupon partners always bad?
No. Some genuinely introduce new customers, drive trial, or reach audiences you cannot reach otherwise. The point is not to ban a category wholesale — it is to measure each partner on incremental lift rather than last-click revenue, and to pay accordingly. A cashback partner that consistently brings new customers earns its rate. One that mostly rebates your existing buyers does not.
What is the single fastest way to expose a free-rider?
A holdout test. Pause the partner, or suppress it in a few matched regions, for a clean window and watch baseline revenue. If total sales barely move, the partner was riding demand that existed without it. No model, survey, or vendor dashboard gives you a cleaner causal read, and you can run a basic version with the tools you already have.
Will fixing this make my affiliate numbers look worse?
In the short term, almost certainly — and that is the point. Reported revenue will drop as you strip out sales that were never incremental, but your true return improves because you stop paying for demand you already owned. The headline number was the illusion. What you are left with is a smaller figure you can actually trust and scale against.
How often should we re-run an incrementality audit?
Treat it as a standing discipline, not a one-time project. Partner mix shifts, new extensions and tactics emerge, and compliance erodes if no one is watching. A full audit each quarter, with continuous monitoring for trademark bidding and extension injection in between, keeps the program honest as it grows.
The bottom line.
Affiliate marketing is one of the most accountable channels you can run — but only if you measure what the partners actually caused, not what they happened to be near. The uncomfortable truth is that a real share of most programs' reported revenue is sales the brand would have made anyway, credited to whoever arrived last. Find those partners, re-price them, and redirect the budget toward the ones building genuine demand, and the channel becomes exactly the growth engine it claims to be. If you want a second set of eyes on whether your program is creating revenue or just taking credit for it, tell us what you're working on.