Marketing Mix Modeling for Brands Under €10M: MMM is free and fashionable in 2026, but it was built for enterprise data. Here is what actually happens below that scale, and what delivers the same budget decision for far less.
Read the full article below for detailed insights and actionable strategies.
Channel comparison
Platform-reported vs. causal contribution
Platform-reported numbers double-count assists; causal inference reveals reality
Marketing mix modeling is having its best year in a decade, and most brands under €10M should still not run one. Google's Meridian made the software free when it went generally available in January 2025, and its 2026 releases added real enterprise muscle. What has not changed is the appetite of the method: practitioners generally need 2 to 3 years of clean weekly data and enterprise-scale spend before an MMM stabilizes. Below that, coefficients wobble between runs and confidence intervals swallow the answer you paid for. For a sub-€10M brand, a causal read on your GA4 export delivers the budget decision MMM promises, at a fraction of the data requirement. Here is the honest math.
What did Google Meridian actually ship in 2026?
Credit where due: Meridian is the most serious free measurement tool Google has released. It went generally available in January 2025 as an open-source marketing mix modeling framework. On February 19, 2026 it added Scenario Planner, which lets teams simulate budget allocations before committing spend. May 2026 brought GeoX for designing and reading geo experiments, plus Meridian Studio, a managed interface for teams without Python.
At Google Marketing Live 2026, Google announced Analytics 360 integration, wiring MMM outputs closer to the GA4 reporting most teams live in. Notice the direction of travel: scenario planning, experiment design, enterprise analytics integration. This is a tool built for organizations with an analytics engineering function and the history to feed it.
Free software is not the same as a free answer. The license costs nothing. The data, the cleaning, and the analyst weeks are where the bill lands, and that bill scales with how messy your history is.
Why does marketing mix modeling need so much data?
MMM is regression on aggregated history. You feed it weekly spend per channel, weekly revenue, seasonality, promotions and pricing, and it estimates how much each input moved the total. The math is respectable. The problem is arithmetic: 2 years of weekly data is about 104 observations, and you are asking the model to split credit across 6 to 10 channels that mostly went up and down together.
When channels move together, the model cannot tell which one did the work. Statisticians call it multicollinearity. Practitioners call it "the Meta and Google coefficients swapped sign again between refreshes." Enterprise advertisers escape this because their spend is large and varied enough, across enough regions and channels, to create separation in the data. At €40K a month across three channels, that variation simply is not there.
Clean is doing heavy lifting in that sentence, too. Clean means the same channels tracked the same way for the whole period: no store re-platforming, no re-tagging of campaigns, no six-week gap where the pixel was down. Most sub-€10M brands changed something material in the last two years, a new site, a new agency, a new channel mix. Every one of those resets the clock on usable history.
The model tells you so honestly: confidence intervals so wide they include both "scale this channel" and "cut it." That is not a Meridian flaw. It is what the method costs, and it is why every MMM case study quietly features a brand spending eight figures.
What are the real minimum requirements per method?
Here is the table we wish someone had shown us before the first MMM pitch. The requirements are practitioner estimates as of July 2026, not vendor guarantees, and your mileage will vary with data quality.
| Method | Minimum data history | Minimum spend to stabilize | Cash cost | Time to first answer |
|---|---|---|---|---|
| Open-source MMM (Meridian) | 2 to 3 years of clean weekly data | Enterprise scale, think €10M+ annual revenue | Software free; analyst weeks per build and refresh | Weeks to months |
| Geo holdout experiment | None beyond clean tracking | Enough conversions per region to detect lift, often €50K+/month in the tested channel | Five-figure vendor contracts | 4 to 8 weeks |
| Platform attribution (GA4, ad managers) | Immediate | None | Free | Instant, but correlational |
| Causal read on GA4 export | Months of daily transaction data | Ordinary DTC spend | €99 per read, €299/mo Pro | 5 to 10 minutes |
Two rows deserve a second look. Platform attribution is instant and free, but an attribution model redistributes credit among touches; it never asks whether the sale would have happened anyway. And the causal read row is not a lite MMM. It answers a different question that fits your stage better: not "allocate €5M across 12 channels" but "was the €40K I spent last month incremental, and which channel earned it?"
Is a causal read a stand-in for MMM?
For the decisions a sub-€10M brand actually faces, mostly yes. You are not optimizing a national media plan across TV, retail and sponsorship. You are deciding whether to move €10K from Meta to Google, whether branded search deserves its budget, whether last month's reported ROAS was real. Those are incrementality questions about individual channels, and causal inference on your GA4 export answers them directly, with assumptions stated instead of buried in a model you cannot inspect. The methodology piece on proving causation without an experiment covers how the counterfactual estimation works.
Where MMM still wins: true budget allocation across many channels including offline, saturation curves for annual planning, and scenario simulation once you have years of history. If that is your problem and you have the data, Meridian is a genuinely good free option. Just do not mistake "free to install" for "ready to decide."
The cadence difference matters as well. An MMM refresh is a project: data collection, validation, a review meeting. Teams run them quarterly at best, so the answer often arrives after the budget has already moved. A causal read is cheap enough to run every month, which changes what it is. It stops being an annual verdict and becomes a steering wheel.
When does MMM become the right call?
Three signals say you are ready. You have 2 to 3 years of consistent weekly tracking you actually trust. Your media spend is enterprise-scale, varied across enough channels and regions to create separation. And the question on the table is allocating a budget big enough that a 10% improvement pays for the analytics team. Most brands reading this are not there yet, and that is fine. Until then, MMM gives you the aesthetic of rigor without the rigor.
When should you revisit MMM as you grow?
Treat MMM readiness as a scheduled checkpoint, not a vibe. Revisit the question once a year, and any time one of these shifts: annual revenue crosses into eight figures, you add channels GA4 cannot see, like retail or offline, your spend spreads across enough regions to give a model real variation, or leadership starts asking for annual scenario planning rather than monthly channel reads. Between checkpoints, protect your future options: keep weekly spend and revenue history tidy, annotate re-platforming and re-tagging events so a future analyst can see the seams, and avoid switching attribution windows mid-year. Brands that do this arrive at their first MMM with usable history. Brands that do not arrive with two years of expensive noise.
There is also a sequencing argument nobody makes: causal reads now make your MMM better later. Regular reads build the habit of asking causal questions, give you channel-level incrementality priors to feed the model once the data exists, and keep you from making two years of budget decisions on last click while you wait to be "MMM-ready." When the CFO asks why marketing deserves its budget next year, you want a track record of causal answers, not a promise to hire a statistician. Our CFO budget-defense kit is built for exactly that meeting.
Causality Engine is the bottom row of the table. GA4 export in, causal read out, in 5 to 10 minutes, with no pixel and no annual lock-in: €99 per read, or €299 per month on Pro. If you want to sanity-check that against an MMM proposal, see the full pricing. For context on what normal looks like, our 2026 DTC attribution benchmarks aggregate 1,300+ reads from brands at your stage.
Key takeaways
- Meridian went GA in January 2025 and shipped Scenario Planner (February 19, 2026), GeoX and Meridian Studio (May 2026), and an Analytics 360 integration at Google Marketing Live 2026. The software is free; the data and analyst time are not.
- Marketing mix modeling typically needs 2 to 3 years of clean weekly data and enterprise-scale spend to stabilize. Below that, coefficients wobble and confidence intervals swallow the answer.
- Roughly 104 weekly observations cannot reliably split credit across channels that move together. That is arithmetic, not a Meridian bug.
- A causal read on a GA4 export answers the sub-€10M version of the MMM question: which channels are incremental, at €99 per read instead of an analyst quarter.
- Causal reads now, MMM later: regular reads give you decisions today and incrementality priors for the model you will run when you have the history.
Further reading
Get attribution insights in your inbox
One email per week. No spam. Unsubscribe anytime.
Key Terms in This Article
Attribution Model
An Attribution Model defines how credit for conversions is assigned to marketing touchpoints. It dictates how marketing channels receive credit for sales.
Attribution Window
Attribution Window is the defined period after a user interacts with a marketing touchpoint, during which a conversion can be credited to that ad. It sets the timeframe for assigning conversion credit.
Causal Inference
Causal Inference determines the independent, actual effect of a phenomenon within a system, identifying true cause-and-effect relationships.
Confidence Interval
Confidence Interval is a statistical range of values that likely contains the true value of a metric. In marketing analytics, it quantifies uncertainty around estimates, indicating the precision of an outcome or causal effect.
Counterfactual
Counterfactual is a hypothetical outcome that would have occurred if a subject had received a different treatment.
Incrementality
Incrementality measures the true causal impact of a marketing campaign. It quantifies the additional conversions or revenue directly from that activity.
Marketing Mix
The marketing mix is the set of actions a company uses to promote its brand or product. It traditionally includes product, price, place, and promotion.
Marketing Mix Modeling
Marketing Mix Modeling (MMM) is a statistical analysis that estimates the impact of marketing and advertising campaigns on sales. It quantifies each channel's contribution to sales.
Related Articles
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
Ready to see your real numbers?
Own the budget? Upload your GA4 export and see which channels drive incremental sales, with confidence intervals, in minutes. Have to defend it? Start with the live demo and take the read to your CFO.
Full refund if you don't see value.
Stay ahead of the attribution curve
Weekly insights on marketing attribution, incrementality testing, and data-driven growth. Written for the person who owns the budget and the person who has to defend it.
No spam. Unsubscribe anytime. We respect your data.
Frequently Asked Questions
Is Google Meridian really free?
Yes. Meridian is open-source software that went generally available in January 2025, and the license costs nothing. The real costs are the data and the people: figure on 2 to 3 years of consistent weekly spend and revenue history, plus analyst time to prepare inputs, validate the model, and refresh it. For brands without an analytics engineering function, those hidden costs usually decide whether marketing mix modeling is practical at all.
How much data does marketing mix modeling need?
Practitioners generally want 2 to 3 years of clean, consistent weekly data, roughly 100 to 150 weekly observations. The model must split credit across channels that often moved up and down together, so more history and more spend variation produce more stable estimates. With less history, coefficients swing between refreshes and confidence intervals widen until they include opposite conclusions about the same channel.
What is the minimum budget for MMM to work well?
There is no official threshold, but practitioner experience puts reliable MMM at enterprise scale: think €10M+ in annual revenue with meaningful spend across several channels and regions. That scale creates the variation the regression needs to separate channels from each other. Below it, the model still runs and still produces numbers, but the outputs are rarely stable enough to move budget on.
What is the difference between MMM and a causal read?
MMM is a top-down regression on weekly aggregated spend and revenue, built for allocating large budgets across many channels over years. A causal read works bottom-up on daily, transaction-level GA4 export data, estimating what would have happened without each channel. MMM answers "how should I allocate next year's budget across everything." A causal read answers "was last month's spend incremental, and where."
When should a small brand invest in MMM?
When three things are true: your weekly tracking history runs 2 to 3 years and you trust it, your spend is large and varied enough to separate channels statistically, and the decision on the table is budget allocation at a scale where a 10% improvement covers the analytics cost. Until then, causal reads on your GA4 export answer the same budget questions with far less data and far less waiting.
What changed in Meridian in 2026?
Google shipped Scenario Planner on February 19, 2026 for simulating budget allocations, then added GeoX experiment design and Meridian Studio, a managed interface, in May 2026. At Google Marketing Live 2026 it announced integration with Analytics 360. The updates make Meridian more capable for enterprise teams, but they do not change its underlying data requirements.