Is your affiliate program adding revenue, or just capturing it? Incrementality testing answers this question with data rather than assumption — and for programs with high coupon/cashback concentration, the answer is often surprising.
The Incrementality Problem in Affiliate Marketing
Affiliate GMV is not the same as affiliate-incremental GMV. The difference: affiliate GMV = all revenue attributed to affiliate channel by last-click tracking; affiliate-incremental GMV = revenue that would NOT have occurred without the affiliate channel. A program with $2M affiliate GMV might have only $1.2M in truly incremental revenue if 40% of its attributed revenue is from buyers who were already going to purchase without the affiliate's influence.
Why this gap exists: last-click attribution methodology assigns credit to the last affiliate touch before purchase — regardless of whether that touch actually influenced the decision.
The three main sources of non-incremental affiliate attribution:
(1) Coupon/cashback cannibalization: a buyer decides to purchase, searches for a coupon code, clicks through a coupon publisher, and uses a code — the coupon publisher gets last-click credit for a sale that was already decided.
(2) Brand search hijacking: a publisher appears in paid search for your brand name, captures clicks from buyers who were already searching for your brand, and earns commission on conversions they didn't influence.
(3) Direct navigation redirect: some publishers use tracking links that fire even when a buyer navigates directly to your site (cookie stuffing variant) — the publisher earns commission with zero influence.
The business cost of non-incremental affiliate attribution: brands paying commission on non-incremental sales effectively reduce their marketing ROI — the affiliate channel appears to perform because GMV is high, but contribution margin is lower than it should be because commission is paid on captured-not-created revenue.
Incremental Lift Testing Methods
There are three practical methods for measuring affiliate channel incrementality, each with different precision and complexity:
(1) Ghost bidding / holdout test: run your affiliate program normally for a control group while deliberately excluding a randomly selected holdout group from affiliate publisher exposure; compare conversion rates between the two groups; the difference in conversion rate is your incrementality lift.
Practical implementation: possible for brands with large enough customer files to create meaningful holdout groups; requires clean audience segmentation and working with publishers to exclude the holdout segment (complex); most rigorous method when properly implemented.
(2) Publisher-level incrementality test: pause a specific publisher for 30 days and measure whether total brand revenue declines by the publisher's attributed GMV; if revenue holds steady, the publisher was not incremental; if revenue declines by approximately the publisher's GMV, the publisher was incremental.
This is the most practical method for most programs — easily implemented on any publisher, actionable results within 30-60 days. Note: only valid when the publisher has no close substitutes in your program (if you pause one coupon publisher and buyers simply use another coupon publisher, the test doesn't tell you what you think it does).
(3) New Customer Rate (NCR) proxy: NCR is the most accessible incrementality proxy for most programs. Assumption: if a publisher is driving a high percentage of first-time customers (70%+ NCR), their conversions are likely incremental — it's difficult to be non-incremental when your audience has never purchased from the brand before. Coupon publishers with 20% NCR are almost certainly non-incremental for 80% of their attributed sales — those buyers were already customers or were going to convert anyway.
NCR is not a perfect incrementality measure, but it's a practical first filter that requires no additional testing infrastructure.
Publisher-Level Incrementality Scoring
Build a publisher incrementality score that combines available signals into a single metric for each publisher:
Signal 1 — New Customer Rate (40% weight): 80%+ NCR = high incrementality signal; 50-80% = moderate; below 50% = low.
Signal 2 — Publisher Type (30% weight): educational content publishers = high incrementality; comparison publishers = moderate; coupon/cashback = low.
Signal 3 — Traffic source (20% weight): direct content site traffic = high; paid search on non-brand terms = moderate; coupon aggregator or brand paid search traffic = low.
Signal 4 — Cookie window analysis (10% weight): conversions clustering within 24 hours of click = potential non-incrementality (coupon capture); conversions distributed throughout 30-day window = more incremental (publisher content drove consideration over time).
Composite score (0-100): sum weighted signal scores. Publishers scoring above 70 are likely incremental; 40-70 ambiguous (test before changing commission); below 40 likely non-incremental (consider commission reduction, publisher type limits, or testing before termination).
This scoring doesn't replace formal incrementality testing, but it prioritizes which publishers to test and guides commission differentiation decisions.
Acting on Incrementality Data
Incrementality data should change program management decisions across three dimensions:
(1) Commission differentiation: publishers with demonstrated high incrementality (>70 score or positive holdout test result) justify premium commission rates; publishers with demonstrated low incrementality (coupon publishers with 20% NCR, failed holdout tests) should receive base rate or below-base commission; this aligns commission investment with actual value creation.
(2) Publisher mix management: if incrementality analysis reveals that 50% of program GMV is concentrated in low-incrementality publishers, gradually shift mix toward high-incrementality publisher types; don't eliminate coupon publishers abruptly (GMV impact is immediate), but cap their GMV concentration and invest recruiting budget in content publishers.
(3) C-suite reporting: present affiliate channel contribution margin based on estimated-incremental GMV rather than total attributed GMV; a program claiming $2M GMV with 40% estimated non-incrementality is actually contributing $1.2M in new revenue — report both figures to give leadership an accurate picture.
The honest conversation: most programs have some non-incremental attribution; the goal is not to eliminate it (some coupon/cashback use is genuinely helpful for conversion) but to price it correctly (lower commission for lower-incrementality publishers) and maintain a publisher mix where the majority of GMV is demonstrably incremental.
