There’s a question I heard from a platform partner recently that has stayed with me. We were discussing campaign performance and they asked, completely sincerely: “Does the client even care about incremental revenue?”
I was stunned. Of course businesses cares about incremental revenue. That’s the whole point of running marketing at all. But the more I’ve turned the question over, the more I think it reveals something uncomfortable about how most companies actually operate. Not whether they care about incrementality in theory, but whether the systems they’ve built are capable of measuring it. And if a system can’t measure something, the business stops acting on it, whatever anyone says they care about.
The case that changed how we think about this
A brand brought us in to improve paid social performance. We did a full account restructure. Diversified the creative. Increased video investment, which we know opens up placements and resonates better with audiences. By every measure we could apply, things were moving in the right direction.
Then the data came back and reported conversions had dropped significantly.
This is the moment most agencies panic and start reversing course. We didn’t, because we’d run incrementality studies in parallel and they were telling a very different story: we were driving more revenue for the business. Platform performance was stronger. The incremental evidence was clear.
So we dug in. What we found was a shift in the composition of conversions that nobody had accounted for. Before the restructure, roughly 80% of the attributed conversions were post-click, meaning someone clicked an ad and then bought. After, it was closer to 20%. The reason was simple: we’d moved aggressively into video. Video influences people through view-through, where they see something, remember it, and come back days later, rather than clicking there and then. And the client’s data system ran on first-click attribution, so it simply couldn’t see what video was doing.
The revenue was there. The measurement infrastructure couldn’t account for it.
Why this matters beyond one account
If this went uncaught, the next conversation would have been about why performance had dropped and what to do about it. Reading from that data alone, the answer looks obvious: pull back on video, lean into static images, optimise for post-click conversions. Reasonable conclusions, but the wrong ones.
Even the “correct” short-term fix, increasing our proportion of static images to recapture attributed conversions, isn’t really a fix. It’s a concession to a measurement framework that can’t see the full picture. You might recover the numbers in the client’s reporting while simultaneously reducing the incremental value you’re generating. Incrementality tests run after that kind of adjustment often confirm exactly that.
The measurement model doesn’t just tell you how you’re doing. It determines what you optimise for. And when the model is wrong, everything downstream is wrong too.
The single-source-of-truth problem
Most businesses pick a measurement methodology, build their reporting infrastructure around it, and then treat that output as objective truth. The reasons are understandable. Attribution tools are expensive to build, complex to change, and deeply embedded in how performance is communicated internally. Moving from first-click to last-click, or to a data-driven model, isn’t a technical change you can make in an afternoon. It reshapes how every channel justifies its budget.
But the result is that companies make significant strategic decisions, budget allocations, channel mix choices, creative strategy, based on a model that was often chosen for historical or organisational reasons rather than empirical ones.
First-click attribution made sense when search dominated the funnel and every journey started with a direct query. It’s a relic in a world where paid social operates at the top and middle of the funnel, where video influences purchase decisions days before anyone types a search, where the customer journey runs across six touchpoints and none of them tell the whole story.
What a better approach looks like
The answer isn’t to find the “right” attribution model and switch to that. Single-model thinking is the problem, not any particular model. The answer is building a measurement architecture that doesn’t rely on any one signal as the ground truth. In practice that means four things working together.
- Run incrementality studies often enough to be directional, not as the occasional one-off.
- Use media mix modelling, which estimates each channel’s contribution from spend and outcomes rather than from tracking individual users, to get a view that doesn’t depend on cookies or clicks holding up.
- Treat platform data as a leading indicator, useful and fast, but not gospel.
- Then reconcile all three constantly: look for where they agree, and pay real attention when they disagree. The disagreement, as it was with this client, is usually where the most useful information is hiding.
There is a commercial point buried in here too. If the only shared language between an agency and a client is “attributed conversions in the reporting tool,” the conversation is constrained before it starts. Any tactic that doesn’t show up cleanly in that tool becomes very hard to defend, regardless of what it is actually worth. Good tactics get cut. Strategies that work get swapped for strategies that look good on a dashboard. The conversation either becomes a weekly negotiation, or the work slowly drifts toward optimising for what shows up in the reporting tool rather than what actually drives growth.
Bottom Line
The platform partner’s question, “does the client even care about incremental revenue?”, was naive. But the better question underneath it, “are we having the conversation about incrementality at all?”, is one worth asking more often.
Measurement isn’t a technical back-office problem. It’s a strategic decision that shapes everything else. The businesses that treat it as such, that invest in understanding what’s actually driving growth rather than what’s showing up in their reporting, are the ones making better decisions. Everyone else is optimising for a number that may have very little to do with the outcome they actually want.
