MMM, MTA & Incrementality: Building your measurement system

Every tool answers a different question. Here's how to combine them.

We've already covered what modern MMMs do and why daily cadence matters. But an MMM isn't the right tool for every question. This lesson: the three foundational approaches to marketing measurement, and how to combine them.

The three approaches

Every measurement tool you've used is built on one of three foundations:

1. Deterministic attribution (MTA)

Tracks individual channel paths and assigns credit based on rules. Last-click, first-click, MTA, and data-driven attribution all fall here. These are late signals in the channel path: they pick up the click, the search, the visit that happens close to the sale. That makes them the right tool for the day-to-day demand-capture questions: which keyword is closing, which creative is winning, where bottom-funnel performance shifts hour by hour.

The issue: deterministic tools sit close to the sale, so they over-claim credit for the demand they were closest to. The closer a signal is to conversion, the less likely it caused the demand and the more likely it's harvesting something already there. Cookie loss, cross-device behaviour, and walled gardens compound the problem. Rely on deterministic attribution alone and you'll see every platform claiming credit for the same conversions.

2. Statistical attribution (MMM)

MMMs work from aggregated data (daily impressions, clicks, revenue) rather than individual paths. They capture the full funnel without depending on user-level tracking. These are early signals: they pick up the demand-creation work that happens before anyone has clicked or signalled intent. That makes them the right tool for the strategic questions: where to split budget across the funnel, what brand and upper-funnel spend is delivering, which channels create demand versus capture it.

The trade-off: traditional MMMs are slow, need years of data, and can struggle to separate correlation from causation without external validation. Rely on a traditional MMM alone and you're operating without visibility between quarterly refreshes.

3. Incrementality testing

Uses controlled experiments (test group vs. control group) to prove whether a specific channel or tactic actually caused a result. Geo-lift tests, A/B tests, and holdout tests all live here. The results are causal and precise, which makes them the right tool for the highest-stakes questions: did this channel really drive the lift we're seeing, or would we have got it anyway?

However, tests are expensive, narrow, and quickly out of date. Expensive because the holdout means giving up revenue while the test runs. Narrow because most tests only look at one or two channels at a time, not the full mix. Out of date fast because the result describes one moment, and the channels you tested rarely look the same by the time you act on it. Brands using tests alone end up with high-confidence answers for a handful of channels and no picture of how everything fits together.

None of these approaches works well on its own. The question is how to combine them.

Triangulation: using them together

Triangulation is one of those terms that gets thrown around in measurement but rarely explained in practical terms. In practice, it's a workflow:

1. Model with your MMM. This is the top-down view: modelling how impressions, views and clicks drove sales downstream to identify which channels are driving results across the full funnel, and where you might be over- or under-investing.

2. Activate with attribution data. This is the bottom-up view: understanding which channels are most effective at capturing demand in the short term, then using MTA or platform signals to optimise campaigns and creative based on what the MMM is telling you at a strategic level.

3. Validate with incrementality tests. When your MMM and attribution data disagree, or when you want to prove the effect of scaling a channel up or down, run a targeted experiment to establish causal ground truth.

Each tool does what it's best at. Over time, the test results calibrate your MMM, which gives you better strategic direction, which you execute on with attribution tools. The loop tightens. And because every output is cross-checked rather than taken on faith, the numbers carry weight with finance and leadership, not just the marketing team.

Which tool for which question?

Start from the decision you need to make, then work out what kind of measurement it needs:

- "Which creative or keyword drove this conversion?" You need deterministic measurement that updates frequently, because creative and keyword performance shifts quickly. MTA, data-driven attribution, ad platform reporting and last-click all fit here.

- "What's the optimal budget split across channels next quarter?" You need probabilistic measurement that captures the full funnel and reveals the long-term picture. That's MMM territory.

- "Is this new channel incremental before I scale it?" You need measurement that proves cause and effect. Run a geo-lift or holdout test.

- "Our MMM and last-click disagree on Meta. Who's right?" A modern MMM should already reconcile the two via calibration. If you want extra confidence, run an incrementality test for causal ground truth.

Two more tools that show up in most brands' measurement systems:

- Ad platform reporting gives you an hourly read on creative and targeting performance. The limit: each platform only sees inside its own ecosystem, so credit is naturally inflated.

- Last-click attribution tells you which channels close the sale. Useful for tactical decisions, misleading as the only measure of channel value.

The most useful questions are the ones that combine approaches. Your MMM flags an opportunity, your attribution data guides the execution, and an incrementality test confirms you got it right.

If managing all these signals sounds like a lot, that's because it is. Platforms like Fospha combine the MMM, incrementality, and daily layers into one system, so triangulation happens by default rather than as a manual process.

Finding the balance

Marketing measurement has to be both accurate and practical to use.

Lean too far on incrementality tests and standalone MMMs, and you get high-confidence answers but weeks without usable signal in between. Lean too far on last-click, ad platform reporting and MTA, and you see a constant stream of detail but only ever a partial picture, which keeps budget locked in short-term wins.

What you need is the balance: measurement that's full-funnel, at least daily, trusted by everyone who reads it, and clear enough that the next decision is obvious.

In the next lesson, we'll go deeper on incrementality testing: how to design tests, when to run them, and how to use the results to sharpen your always-on measurement.

‍