More and more brands are using marketing mix modeling (MMM) to judge marketing effectiveness and allocate budget. And that’s not just anecdotal: over the summer, the APMA surveyed 67 brands and advertisers to understand how widely they’re using MMM, how they’re representing affiliate within those models, and how the results are impacting investment decisions. Nearly 60% already use it, and more than half said MMM has a moderate or high influence on budget decisions.
But here’s the challenge: as those adoption and influence numbers climb, the data also shows that affiliate marketing isn’t always represented in a way that reflects how complex it actually is. About 70% of respondents said they measure affiliate as a distinct channel within their MMM; based on those measurements, most respondents judged affiliate as having either moderate or weak contributions to their output.
For affiliate marketing, this data shows a risk, and an opportunity. To uncover those opportunities, we wanted to get to the bottom of what’s causing this measurement gap, and why MMMs aren’t representing the channel in a way that reflects its complexity. Those critical questions formed the focus of a recent webinar that Acceleration Partners hosted alongside the APMA featuring Helen Southgate (Acceleration Partners), Amy Johnson (TUI), James Little (TopCashback) and Kevin O’Farrell (Analytic Partners), bringing together advertiser, publisher and measurement perspectives. Three genuinely different seats at the table, based on the belief that if this group can’t find a straight answer between them, it’s hard to know who could.
The measurement gap is real
First off, what’s marketing mix modeling? Kevin O’Farrell outlined a helpful analogy during the webinar to answer this question. He thinks of it as statistical detective work, untangling the complex web of what’s really driving sales (TV campaigns, price cuts, heat waves, affiliate sites, the state of the economy) from historical data. MMM takes that history and changes in sales to unpack what drives incremental impacts and separate those drivers from other factors.
When it comes to MMM, the APMA research showed that affiliate has a representation and trust problem. The channel is often visible in MMM, but brands and advertisers don’t trust it. While 70.1% measure affiliate as a distinct channel within their MMM, 35.8% still represent affiliate as a single undifferentiated line. Only 29.8% split by partner type and/or individual partners. Plus, only 9% believe their MMM measures affiliate accurately.
Those beliefs are having a real impact on budgets: in the last 12 months, MMM output cut affiliate budget for nearly 18% of respondents. AP and APMA even conducted a live poll during the webinar, which showed that 43% of attendees named proving incrementality as the biggest barrier to demonstrating affiliate’s value. James Little told a story of an affiliate team that discovered their whole program was closing based on MMM output they were never involved in producing. Those issues stem from how affiliate is incorrectly represented in the MMM space.
Affiliate marketing has a data problem
The point isn’t that MMM is inherently bad for affiliate marketing. The challenge is that measurement decisions are becoming investment decisions. If the inputs don’t properly reflect what affiliate actually does, the consequences can be significant.
The truth is that most MMMs are working from a very limited data set, typically with last-click data across most programs. In the past, affiliate measurement has largely revolved around clicks and last-click sales. At the same time, publishers hold plenty of data that never makes it into that picture, including onsite placement performance, email performance, and paid social activity. None of this information is currently fed back into the MMM, but not because publishers are withholding it on purpose. It’s just that no one has asked for it before.
Other channels beyond affiliate are also realizing that clicks aren’t the whole picture. For example, in online searches, AI overviews are translating to fewer clicks but still lead to real exposure. That tension leads to a shift away from measuring success according to clicks and focusing more on capturing impressions and exposure data.
So, should publishers do more to feed in richer data? Even if they wanted to provide this data, there are real unresolved questions, like how to consistently define an impression, and how different networks would ingest it in a common way. The root cause is actually a shared data gap, meaning that MMM is not the enemy of affiliates. Poor data is.
Three common MMM setup mistakes
The data gap comes down to three specific mistakes in how affiliate gets fed into the model.
- Mistake one: MMM weights on impressions. The model expects reach and frequency data, but affiliate doesn’t generate that. Feed it click volume instead and ask publishers for whatever additional data they can provide. The more data the model gets, the better it reads the channel.
- Mistake two: every partner lands in one bucket. Content and voucher partners behave differently, but a model that can’t tell them apart has no way of representing that difference.
- Mistake three: affiliate is performance-based. Commissions pay on the sale, so the model reads affiliate activity as riding along with demand rather than causing.
We can solve this problem, together
We had three seats at the table during this webinar and quickly realized that each seat at the table sees a different piece of the data measurement puzzle. That means each has its own role to play in fixing it. Waiting for MMM to improve on its own isn’t an option.
Let’s start with what the APMA survey results state the industry actually wants: best-practice guidance on how to represent affiliate in MMM. Respondents asked for attribution as the most requested form of support, ahead of incrementality testing help and better data provision from networks and publishers.
Now, let’s break down what each group can do to provide this guidance and close the measurement gap.
Brands
- Don’t assume your MMM provider understands affiliate’s nuances: brief them proactively on partner types and how the channel actually operates.
- Push for partner-level or partner-type splits in your model: don’t accept a single undifferentiated affiliate line.
- Document every MMM cycle: data sources used, who pulled what, what assumptions were made. This protects institutional knowledge as teams and providers change.
- Involve publishers in the conversation before decisions are made off the back of MMM output, not after: this way, teams aren’t blindsided by a decision they had no part in.
Publishers
- Be more open to sharing data, such as impressions and exposure figures: much of it hasn’t been shared before, largely because it hasn’t been requested.
- Push back with evidence, not assumption: when a channel is being undervalued.
- Talk proactively to advertisers about measurement: don’t wait until budgets are challenged.
Industry
- Create a common measurement language. Support the case for a shared industry framework or standard for how affiliate should be represented in MMM and attribution.
MMM isn’t the enemy of affiliate marketing. Poor data is, and that’s a problem the industry can fix. Brands, publishers, and the wider industry each hold a piece of it, and the steps above are where to start.
