Affiliate Attribution Models: Which One Is Costing You Money

An affiliate attribution model is the set of rules that determines which affiliate partner gets credit for a sale, and how much commission is paid as a result. Get the model right and you reward the partners genuinely driving growth. Get it wrong and you systematically overpay for traffic that was going to convert anyway, while underinvesting in the partners who are actually building demand.

Most affiliate programmes default to last-click attribution. It is the easiest model to implement and the hardest to defend commercially. If you are running a mature affiliate programme and still paying last-click across the board, there is a reasonable chance you are haemorrhaging margin on voucher sites and cashback platforms while your content and comparison partners go unrewarded.

Why Affiliate Attribution Is a Commercial Decision, Not Just a Technical One

When I was running iProspect, we managed affiliate programmes for clients across retail, travel, and financial services. One of the most common conversations I had with clients was about voucher code sites sitting at the end of the customer experience, claiming last-click credit on sales that had been in the basket for three days. The client was paying 8% commission on transactions that were essentially guaranteed. The attribution model was not measuring performance, it was subsidising interception.

That is the core problem with treating affiliate attribution as a tracking exercise rather than a commercial strategy. The model you choose determines who gets paid, and who gets paid determines who invests in promoting your brand. If your model rewards the wrong behaviour, your programme will gradually fill with partners who are good at exploiting that model rather than partners who are good at driving genuine growth.

If you want to go deeper on how attribution fits into a broader measurement framework, the Marketing Analytics hub at The Marketing Juice covers the full picture, from GA4 configuration through to marketing mix modelling and incrementality testing.

What Are the Main Affiliate Attribution Models?

There are six models worth understanding. Most affiliate networks support several of them, though the default is almost always last-click.

Last-Click Attribution

The final affiliate touchpoint before conversion gets 100% of the commission. Simple to implement, easy to audit, and commercially problematic for any programme with meaningful scale. Last-click creates a structural incentive for partners to position themselves at the bottom of the funnel, often by targeting branded search terms, offering discount codes, or sitting inside browser extensions that fire at checkout.

The model made sense in an era when affiliate journeys were simple and linear. Most customer journeys are neither of those things now. Paying a cashback site 8% commission to confirm a purchase the customer had already decided to make is not affiliate marketing, it is a loyalty tax you are paying to a third party.

First-Click Attribution

The first affiliate touchpoint in the customer experience gets full credit. This model over-rewards upper-funnel partners and ignores the contribution of anyone who helped close the sale. It is rarely used in practice because it creates the opposite distortion to last-click: partners who drive awareness get everything, partners who drive conversion get nothing. Neither extreme reflects how customers actually behave.

Linear Attribution

Commission is split equally across every affiliate touchpoint in the experience. If a customer interacted with a content site, a comparison site, and a cashback site before converting, each gets a third of the commission. Linear attribution is more equitable than last-click but it assumes every touchpoint contributed equally, which is rarely true. A content article that introduced a customer to your brand three weeks ago probably contributed more than a cashback notification sent 30 seconds before checkout.

Time-Decay Attribution

Touchpoints closer to conversion receive more credit than earlier ones. The logic is intuitive: the interactions that happened just before a customer converted were probably more influential than those that happened weeks ago. Time-decay attribution tends to favour lower-funnel partners, though less aggressively than last-click. It is a reasonable model for short consideration cycles, less so for high-value purchases where the research phase matters enormously.

Position-Based Attribution

Also called U-shaped attribution. The first and last touchpoints each receive a larger share of credit, typically 40% each, with the remaining 20% distributed across any middle touchpoints. This model acknowledges that both introduction and conversion matter, which is a more commercially honest position than either first-click or last-click alone. It is a pragmatic choice for programmes with multi-touch journeys and a mix of upper and lower funnel partners.

Data-Driven Attribution

Commission is distributed based on each touchpoint’s actual contribution to conversion, calculated using machine learning across your historical data. This is the most accurate model in theory and the most demanding in practice. You need sufficient conversion volume for the model to produce statistically meaningful outputs, clean and consistent tracking across all touchpoints, and a network or platform that supports it. Google has been pushing data-driven attribution in its own ad products for years, but the affiliate channel has been slower to adopt it, partly because the data infrastructure required is genuinely difficult to maintain at scale.

How Voucher and Cashback Sites Exploit Last-Click Models

I want to spend a moment on this because it is one of the most persistent sources of wasted affiliate spend I have seen across two decades of managing programmes.

Voucher code and cashback sites are not inherently bad partners. Some of them drive genuine incremental revenue by attracting price-sensitive customers who would not have purchased at full price. The problem is that last-click attribution makes it impossible to distinguish between a cashback site that genuinely influenced a purchase and one that simply intercepted a customer who was already at checkout.

Browser extensions are the most egregious version of this. A customer researches a product, reads three reviews, compares prices, and arrives at your checkout. At that moment, a browser extension fires a cashback notification. The customer clicks it, the cashback site claims last-click credit, and you pay commission on a sale that was already done. The extension contributed nothing to the customer’s decision. It just sat at the door and collected a toll.

When we audited affiliate programmes at iProspect, this pattern appeared in almost every mature retail programme we inherited. The fix was not to remove cashback and voucher partners entirely. It was to renegotiate commission rates based on incrementality data, move to a position-based or data-driven model, and in some cases exclude specific partners from receiving credit on transactions that originated from branded search or direct traffic.

What Does Incrementality Have to Do With Affiliate Attribution?

Attribution models tell you which touchpoints were present in a customer experience. Incrementality testing tells you which touchpoints actually changed the outcome. These are different questions, and conflating them is one of the most expensive mistakes in performance marketing.

A partner can appear in every converting experience and still contribute nothing incrementally if those customers would have converted anyway. Last-click attribution is particularly blind to this distinction because it assigns full credit to the final touchpoint regardless of whether that touchpoint influenced anything.

The practical way to test incrementality in affiliate programmes is to run holdout experiments: temporarily suppress a partner’s activity for a segment of your audience and measure whether conversion rates change. If they do not, that partner is not driving incremental revenue. If they do, you have evidence of genuine contribution. This is more operationally demanding than switching attribution models, but it gives you something attribution models cannot: actual causal evidence.

I judged the Effie Awards for several years. One of the things that struck me consistently was how few entries could demonstrate genuine incrementality. Most were showing correlation, not causation. The same problem exists in affiliate measurement. The model shows you who was there. It does not show you who mattered.

Understanding how marketing metrics connect to business outcomes is the foundation of any honest measurement approach, and affiliate attribution is no exception.

How to Choose the Right Affiliate Attribution Model for Your Programme

There is no universally correct answer here. The right model depends on your programme structure, your customer experience, and what behaviour you want to reward. But there are some practical principles that hold across most situations.

Start with your partner mix. If your programme is dominated by content and comparison sites that do genuine upper-funnel work, last-click will undervalue them and they will eventually reduce their investment in your brand. If your programme is dominated by voucher and cashback sites, last-click will over-reward them and your programme will gradually become a margin drain rather than a growth channel.

Map your customer experience before you choose a model. How many affiliate touchpoints appear in a typical converting experience? How long is the consideration cycle? Are there clear patterns in which partner types appear at which stages? Most affiliate networks provide experience-level reporting that lets you answer these questions. If yours does not, that is a problem worth solving before you optimise the attribution model.

Consider a tiered commission structure rather than a single model applied uniformly. Pay content partners a flat fee or a higher CPA for introducing new customers. Pay comparison sites a standard CPA for converting customers already in the market. Pay cashback and voucher sites a lower CPA and exclude them from credit on transactions where they were not genuinely the last meaningful touchpoint. This is more complex to administer but far more commercially honest than applying last-click to everyone.

If you have the volume and the infrastructure for data-driven attribution, it is worth the investment. The challenge is that most affiliate networks are not built for it natively. You may need to layer your own analytics on top of network reporting to get there, which brings me to the tracking question.

The Tracking Infrastructure That Makes Attribution Models Work

Attribution models are only as good as the data feeding them. If your affiliate tracking is leaky, inconsistent, or missing touchpoints, no model will give you accurate outputs.

The basics matter more than most programme managers acknowledge. Consistent UTM parameters across all affiliate links, so you can cross-reference network data with your own analytics. Server-side tracking where possible, to reduce the impact of ad blockers and browser privacy changes on cookie-based attribution. Deduplication logic that prevents the same sale being claimed by multiple partners or multiple channels simultaneously.

I have seen programmes where the affiliate network was reporting 40% more conversions than the client’s own analytics. That gap is not a rounding error. It means either the network is overclaiming, the client’s analytics are undercounting, or both. Until you understand and close that gap, any attribution model you apply is working with unreliable inputs.

GA4 has changed some of the mechanics here. GA4 is better understood as a directional reporting tool than a precise source of truth, and that framing applies equally to affiliate tracking. The goal is not perfect measurement. It is honest approximation that is consistent enough to make good commercial decisions.

If you are evaluating whether GA4 is the right analytics foundation for your programme or considering alternatives, there are credible GA4 alternatives worth assessing depending on your data governance requirements and technical setup.

The Conversation Your Affiliate Partners Will Have With You

If you move away from last-click attribution, expect pushback. Voucher and cashback partners in particular will argue that they drive incremental revenue, that their audiences would not have converted without the incentive, and that changing the attribution model undervalues their contribution. Some of that is true. A lot of it is not.

The right response is to bring data to the conversation. If you have run incrementality tests, share the results. If you have experience-level data showing where partners appear in the funnel, use it. Partners who are genuinely driving value will be able to demonstrate it. Partners who are not will struggle to make the case once you move beyond anecdote.

Early in my career, I learned that the best commercial conversations are the ones where both sides have data and neither side is pretending. The worst are the ones where one side has data and the other side has a loud voice. Attribution model changes are a good test of which kind of relationship you have with your affiliate partners.

Visualisation tools can help here. Bringing partner performance data into a clear dashboard, whether through your network’s reporting or a tool like Tableau for marketing analytics, makes these conversations more productive and less adversarial.

What a Commercially Honest Affiliate Attribution Framework Looks Like

Pulling this together, a commercially honest affiliate attribution framework has four components.

First, a model that reflects your actual customer experience rather than the default your network offers. That might be position-based, time-decay, or data-driven depending on your programme structure and volume. It is unlikely to be last-click across the board.

Second, a tiered commission structure that rewards different partner types differently based on their role in the funnel. Content partners who introduce new customers should be compensated differently from cashback sites who appear at checkout. The model should incentivise the behaviour you want, not just the behaviour that is easiest to track.

Third, an incrementality testing programme that runs alongside your attribution model. Attribution tells you who was present. Incrementality tells you who mattered. You need both.

Fourth, clean tracking infrastructure that gives you consistent, cross-referenced data across your network and your own analytics. Understanding how traffic sources appear in your analytics is a prerequisite for any attribution work that is going to hold up commercially.

None of this is technically exotic. Most of it is commercially disciplined programme management applied to a channel that has historically been allowed to run on autopilot. The affiliate channel can drive genuine, efficient growth. But only if the attribution model is honest about what growth actually looks like.

For more on building measurement frameworks that connect channel performance to business outcomes, the Marketing Analytics section of The Marketing Juice covers attribution, incrementality, GA4, and the metrics that actually matter at a commercial level.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions