How do you prove to a client that their media money is working? Stop leading with ROAS screenshots. The proof clients now trust has three parts: a model that shows what each channel added on top of the baseline, an experiment that confirms it, and evidence that explains why the numbers moved. Marketing mix modelling is back because it supplies the first part well.
The latest proof point comes from the Gulf. Campaign Middle East reported that performance agency TRKKN used Google's open-source Meridian marketing mix model to analyse three years of data for beauty retailer Faces. Across 11 channels, the model projected an 8.5% incremental revenue lift and a 10.8-percentage-point ROI improvement, simply from reallocating existing budget.
Why is marketing mix modelling back?
Because the alternatives stopped convincing anyone. Platform dashboards report the conversions they believe they caused, so the platform selling the media is also the one measuring it. Last-click attribution hands the credit to whichever touchpoint came last, which flatters search and retargeting and starves the activity that built the demand in the first place.
MMM takes a different route. It looks at spend and sales over time, across every channel including offline, and estimates what each one contributed beyond the baseline. It does not need to track individual users. And with open-source options such as Meridian, the barrier is now data engineering and expertise rather than a hefty licence fee.
Harry Lonergan, digital marketing director at Faces, told Campaign Middle East: "Meridian has helped us move on from what we see in last click performance." That is the sentence most clients would like to be able to say.
What did the Faces and TRKKN case actually show?
The detail is worth reading, because it doubles as a template. TRKKN led the data engineering, model building and deployment inside Faces' existing Google Cloud environment. The model covered online and offline marketing and revenue data in Saudi Arabia from 2023 to 2025.
Three outputs mattered. First, it showed which channels were driving incremental revenue, identifying 19% of revenue as campaign-driven. Second, it estimated each channel's response and saturation curve, showing where extra spend would yield diminishing returns. Third, it modelled budget scenarios, which is where the 8.5% figure came from. Performance Max came out as the strongest channel in the analysis.
Note what the agency did not ask for: more money. The recommendation was to move the money already there. And the framework is ongoing, reassessing ROI as new data arrives, rather than a one-off report. As Vimal Badiani, managing director of TRKKN, put it: "It's about giving marketers the evidence and confidence to make better decisions."
What evidence do clients need to believe media ROI?
No single method is enough. Each layer answers a different question and carries a different weakness.
| Evidence layer | Question it answers | Typical method | Weakness |
|---|---|---|---|
| Platform-reported ROAS | How efficient was this channel, by its own measure? | Ads manager reporting | Marks its own homework; ignores incrementality |
| Marketing mix modelling | What did each channel add beyond the baseline? | Statistical model of spend and sales over several years | Explains what, not why; needs consistent data |
| Incrementality tests | Did this channel cause the lift? | Geo holdouts, conversion lift studies | Slow; usually one channel at a time |
| Saturation curves | Where does extra spend stop paying back? | Output of the MMM | Only as good as the model behind it |
| Audience and competitor evidence | Why did results move? | Social listening, competitor ad analysis | Anecdotal unless traced to real data |
Most agencies present the first row and hope. The ones that keep budgets present all five.
Why do media budgets end up in the wrong places?
Economists have a name for this: market failure. Markets allocate resources well when buyers and sellers have good information. When information is missing, or held by one side, money flows to the wrong places and stays there.
Media buying is a textbook case. The client cannot see what the platforms see. The agency sees more than the client but is often paid, at least partly, in relation to what it spends. Platforms report on their own performance. Everyone acts rationally on partial information, and the result is budget piled into channels that look good on last click while the channels that build demand go hungry. A projected 8.5% gain from simply moving budget Faces was already spending is, in effect, a measure of that misallocation.
The fix is not more data. It is shared clarity. In the realm of commerce, clarity breeds connection: when client and agency look at the same evidence and read it the same way, the conversation moves from defending spend to deciding where it goes next.
How do you build a media ROI story clients will trust?
- Agree the business outcome and the baseline. Revenue, new customers or margin, and an honest view of what would have happened with no marketing.
- Assemble a consistent data history. Spend and results by channel, online and offline, plus pricing, promotions and seasonality. Faces used three years across 11 channels.
- Run the model. Use an open-source MMM such as Meridian or a specialist partner. Ask for contribution by channel and saturation curves, not a single ROI figure.
- Validate one finding with an experiment. A geo holdout on the channel the model says is overfunded is the most persuasive test you can run.
- Explain the why. Pair each channel finding with what was actually said and seen: the creative that ran, what competitors launched and how the audience conversation shifted.
- Present a reallocation, not a request. Show the scenario that moves existing budget and the outcome it projects.
- Keep the model alive. Re-run it as new data arrives, so ROI becomes a standing conversation rather than an annual argument.
Where do creative and local context fit into media ROI?
MMM tells you which channel worked. It says little about what the message did, and that gap matters. Campaign reported new research from Publicis London, System1 and Effie comparing global campaigns with local-market work, which found that global campaigns deliver fewer commercial outcomes. If the fit between a message and its market affects results that much, a channel-only explanation is half an answer.
This is the missing half of most ROI decks. When a model says paid social over-delivered in the third quarter, the client's next question is why. Was it the creative? A competitor going quiet? A shift in what the audience was talking about? Defy Monitor answers that by tracing every AI conclusion back to the real published posts behind it, with strategy reverse-engineered from what audiences and competitors actually publish rather than from surveys. It sits alongside your measurement, not instead of it, as part of the Defy platform.
