Proof Over Assumption: A Practical Guide to Incrementality Testing in Affiliate & Performance Marketing

Every affiliate manager has sat across from a finance director who is quietly suspicious that affiliate spend is just paying for sales that were already going to happen. In many programmes, that suspicion is at least partially correct.

Traditional last-click attribution - still the default model in a large number of affiliate networks - gives 100% of conversion credit to the final touchpoint before purchase. The result is a measurement system that systematically favours lower-funnel partners who insert themselves late in the customer journey: cashback browser extensions, coupon aggregators, and toolbar plug-ins that activate precisely when a shopper is about to complete a checkout they had already committed to.

The affiliate industry is enormous. Global affiliate marketing advertiser spending reached $15.7 billion in 2024, up from $14 billion the prior year, and the total ecosystem is valued at roughly $18.5 billion worldwide. With that kind of spend, the cost of misattribution is not academic - it is a direct drag on programme profitability and a barrier to securing budget increases for partners who genuinely grow the top of the funnel.

Incrementality testing exists to solve this problem. Not by eliminating partner types, but by ensuring that commissions and investment levels reflect the actual value each partner creates.

Incrementality measures the additional impact - commonly called "lift" - that a specific marketing activity produces beyond what would have occurred without it.

The clearest definition used across the industry comes from Rakuten Advertising: incrementality is "the additive business value driven by a singular marketing channel or partner." Impact.com frames it as the difference between partners "who drive genuine growth and those who simply take credit for sales that would have happened anyway."

In practice, the question is simple: if this partner, campaign, or channel disappeared tomorrow, how many of today's conversions would we lose?

If the honest answer is "not many," that partner is not highly incremental. If removing them would cause a meaningful drop in net-new customers or revenue, their contribution is real.

Important nuance: incrementality is not a binary judgment. A cashback partner with low incremental lift might still justify its commission rate through volume, customer loyalty, or competitive defensive value. The goal of measurement is to make those trade-offs explicit and conscious, not to trigger wholesale partner culls.

To understand why incrementality testing matters, it helps to see precisely where conventional models fail.

Consider a shopper who discovers your brand through a content review, compares products on a niche editorial site, receives an email from you, and then - as they navigate to checkout - is prompted by a browser extension to activate cashback. Under last-click attribution, the extension receives full credit. The content site receives nothing. The editorial comparison piece receives nothing. Your email receives nothing.

This creates a compounding distortion: you reward the last touchpoint, so you invest more in last-touchpoint partners, which trains your data to show those partners as your top performers, which justifies further investment. Meanwhile, the upper-funnel partners who introduced the customer to your brand in the first place are underfunded or dropped.

Research has consistently shown that traditional attribution methods produce inflated lift measurements due to selection bias, because exposed and unexposed groups are not genuinely comparable without experimental design. The only way to establish causality - not just correlation - is through controlled experimentation.

Not all affiliate partners carry equal incremental value, and the variation follows a consistent pattern across programs. Understanding this hierarchy is essential for building a commission structure that reflects true ROI.

High incrementality partners
Content publishers, review sites, editorial comparison platforms, and influencers typically exhibit the strongest incremental signals. They operate at the top and middle of the funnel, introducing products to audiences who have not previously considered the brand. Research from DMi Partners found that orders with affiliate touchpoints achieve a 27% higher average order value on average - a signal consistent with upper-funnel discovery driving more considered, higher-value purchases. Creative Market's 2024 tracking of new versus returning customer rates through affiliate content showed a clear picture of genuine acquisition being driven through editorial placements.

Mixed incrementality partners
Coupon partners occupy a more complex position. They can drive incremental volume by attracting deal-seeking audiences who would not convert at full price, and upper-funnel coupon placements (paid search, editorial deal roundups) behave more like discovery channels. However, coupon partners whose primary mechanism is intercepting users already in checkout function more like last-click credit claimants than genuine demand creators. The channel's incrementality is therefore highly dependent on how and where the partner promotes - which is why exclusive promotional codes, rather than generic discount codes, are a best practice for tracking coupon-specific lift.

Lower incrementality partners
Cashback partners and toolbar extensions are, as a category, the most likely to exhibit low incrementality. Their mechanism typically activates at the point of conversion for users who had already decided to buy. This does not make them valueless - they can serve legitimate defensive purposes in competitive categories, retain customer loyalty, and contribute volume - but it does mean their commission rates should be calibrated accordingly rather than assigned on par with top-of-funnel publishers.

As one affiliate programme director put it: "Understanding incrementality is essential for optimising affiliate programmes. While maximising incrementality is important, it shouldn't come at the expense of programme growth and sustainability. Many successful programmes maintain a mix of both high and low incrementality partners, with commission structures and resource allocation reflecting their relative value."

Running tests during seasonal peaks. Q4, major sale events, and other high-traffic periods introduce too many confounding variables. Run baseline tests during stable periods first.

Testing too small a holdout group. A holdout that represents less than 10% of your audience is unlikely to reach statistical significance in a reasonable timeframe. A 20–30% holdout is generally recommended for most programme sizes.

Conflating high revenue with high incrementality. A partner can generate substantial attributed revenue while delivering low incremental lift. Volume and incrementality are separate dimensions. Both matter; neither tells the full story alone.

Acting on a single test. One test establishes a hypothesis. Two or three consistent tests establish a pattern. Build an incrementality testing cadence into your programme governance rather than treating it as a one-off exercise.

Ignoring spillover effects. When you pause a partner in a geo test, their users may shift to a competing channel. Measure halo effects across your full channel mix - not just the paused channel - before drawing conclusions.

One of the most significant structural advantages of incrementality testing - particularly geo experimentation - is its compatibility with a post-cookie measurement landscape. As third-party cookie signal loss accelerates and platform-reported attribution becomes less reliable due to iOS restrictions and GDPR compliance requirements, holdout and geo experiments offer a causal measurement methodology that does not depend on individual-level tracking.

This positions incrementality testing not just as a tool for optimising today's affiliate programmes, but as a foundational pillar of future-proof measurement strategy alongside Marketing Mix Modelling (MMM) and multi-touch attribution (MTA). Leading measurement practitioners increasingly describe the combination of all three as the industry standard for marketing teams that need both strategic budget planning and tactical campaign optimisation.

Incrementality testing is not a sophisticated academic exercise reserved for enterprise brands with data science teams. It is a practical, implementable framework that any affiliate programme manager can begin applying - starting with commission decrement testing at the partner level, moving toward holdout experiments at the programme level, and eventually integrating geo experimentation for cross-channel validation.

The payoff is significant. Programmes that measure true incrementality consistently report better budget allocation decisions, stronger relationships with high-performing partners, and more credible ROI reporting to leadership. More importantly, they stop paying commissions for sales that were always going to happen.

In a performance marketing ecosystem where every dollar of spend is scrutinised, the ability to prove causality - not just claim correlation - is the clearest competitive advantage a programme manager can build.