Make your investment count
It’s no secret that in 2026, businesses are laser-focused on making sure that every pound they spend on traditional & digital marketing is contributing to increasing their return on investment.
The real question: How can we make the most of our data to boost the bottom line?
There are more than a couple of ways, including customer analysis and competitor analysis, but as the data pool grows, there can be blind spots.
That’s where a structured, three-layer measurement stack (MMM, experiments and attribution) helps cut through the noise and find real drivers of performance and confidently decide where your next pound should go.
What is MMM?
Put simply, Media Mix Modelling (MMM) is a statistical analysis of aggregated data to determine the impact of a wide range of marketing activities. It takes a large amount of historical data, such as advertising spend, media impressions, and external factors like seasonality, to calculate the return on investment for each channel, giving a top-down view of performance. There are 4 main insights to come out of it:
- Contribution: How much each channel has contributed to sales
- ROI Calculation: The return on investment for each marketing channel
- Baseline Sales: The sales that occur without marketing activities
- Forecasting & Optimisation: Predicting future outcomes based on budget shifts
A great advantage to using MMM alongside user-level tracking is that it is not only more privacy-friendly, but it also operates on aggregate data, which can make it easier to identify trends and patterns that are otherwise hidden in raw data.
MMM, forecasting
Futurology has been around for years, and it was commonplace to give a finger-in-the-air forecast; with MMM, you can be bold in your strategy with the confidence that there is a huge amount of data backing your decisions.
Although there are a huge number of external factors that can challenge businesses, processing years of data into one condensed overview can really help support and build confidence in moving budget around for maximum impact on your digital marketing campaigns.
Experiments for truth
If Media Mix Modelling is where you are deciding “how much and where”, experiments are where you can find out what is actually true. They cut through the assumptions that models can make and give you exceptional, clean treated vs control comparison. At the core, they answer a simple question: “Compared to doing nothing, what really changed?”
There are three groups we’re going to explore here: geo experiments, user-level A/B tests, and pseudo-experiments. Geo experiments work by tuning spend in a set region while holding others as a control group.
When each geography contains a natural mix of users, channels and bahaviours, you get a more realistic and holistic view of total lift with the combination of online and offline effects, brand and performance.
User-level A/B tests are closer to what most digital marketing agencies are familiar with whereby we randomly assign users to see an ad, a campaign or a site change then compare the outcomes.
Pseudo-experiments sit in the awkward middle space between proper experiments and pure observation. The basic move is to treat real-world variation as if it were the experiment: a staggered rollout, a budget shock in one region, a platform outage, or even a sudden promo change can all be carefully leveraged to business advantage.
In a hybrid stack, experiments are the reality check on everything else. When MMM says “Bing prospecting has strong marginal ROI”, you validate that with a holdout test before you reassign your entire budget. This is where “truth” appears in the data; if the experiment says the incremental lift is tiny, you absolutely have the backing of clear data to ignore platform dashboards that try to persuade you otherwise.
Attribution for accountability
If experiments give you truth, and MMM gives you strategy, attribution is all about the day-to-day accountability in your campaigns. It’s here so that if someone on your team asks “which campaigns are pulling their weight and what should we do before the next pacing check?”, you don’t just have a crude last-click data sheet, you have a consistent scoreboard for operational decisions.
As always, we’re trying to empower marketing teams to make better decisions. With the accountability of spending, we can ask direct questions like “Can we see where the budget is going?” and “What did this spending bring in?” to a team, and know that they have access to clear, concise data to answer them.
Importantly, spending accountability is incredibly important when it comes to reporting back to clients on what their investment is doing for them.
Alongside spending, there is also accountability of the model itself. Your attribution view shouldn’t be treated as the unquestionable truth but as a working hypothesis that gets checked against experiments and MMM on a regular basis.
When platform ROAS, tests and modelled contribution disagree, that’s a feature, not a bug; it is an indication to revisit the assumptions that have been made, tweak optimisation rules and sometimes even retire a favourite campaign that isn’t as effective as it looks on an operational dashboard.
How it comes together
These three layers are not siloed; they are complementary. When you put all three layers together, you get a measurement stack that is far more robust than using a singular tool or model. MMM is your strategic steering wheel, experiments are your tour guide and attribution becomes your everyday accountability. It’s the scoreboard that turns strategy and evidence into planned optimisations guiding which campaigns to paud, which audiences to grow and, sometimes, which creatives to retire.
The trick in 2026 isn’t to put one above the other, it’s to ask the right questions to each layer and keep communication flowing between them all. When you do that, you give your team a shared language for decisions instead of three competing stories about performance.
If you want support building a three-layer measurement stack like this for your own business, get in touch!
Further reading
https://funnel.io/blog/marketing-mix-modeling-explained
https://facebookexperimental.github.io/Robyn/docs/analysts-guide-to-MMM/
