-1.png?width=1132&height=292&name=Asset%201@2x%20(12)-1.png "Odins.ai Logo")
By the Odins team · Updated October 2026 · About 6 minutes
"Are we big enough for marketing mix modelling?" is the first question most buyers ask, and the one vendors answer least clearly. The answer has little to do with company size. It depends on how much you spend on marketing, how much that spend varies, and how clean your sales signal is. This guide gives the thresholds we use, why they exist, and what to do if you sit below them.
The short answer
Marketing mix modelling starts to pay off at a marketing budget of about EUR 1 million a year, and there is no upper limit: the same model covers budgets in the hundreds of millions. What the model needs is variation in spend across channels and over time, and a sales signal it can learn from. Revenue, headcount and industry do not decide it. Below the threshold, the better first investment is usually a clean cross-channel dataset.
Why there is a floor at all
A marketing mix model estimates how sales respond to each channel by looking at what happened when spend went up and down. If spend never moved much, or the total is so small that its effect drowns in normal sales noise, the model cannot separate one channel from another. It will still produce numbers. They will have wide uncertainty ranges, and a budget decision built on them is a guess with a chart attached.
So the floor is about statistical signal. Around EUR 1 million a year is where we see enough spend, spread across enough channels, for the model to place each channel's response curve with ranges tight enough to act on. It is a rule of thumb from our own onboarding, not a law. A company spending somewhat less with strong seasonality and two clean years of history can be modelled; a company spending more with flat spend and a noisy sales feed can be harder.
Three things that matter more than the headline number
| What the model needs | Why it matters | Rule of thumb |
|---|---|---|
| Variation in spend | The model learns from change. Channels that were switched on and off, scaled up for a campaign or paused over summer give it something to read. | Spend that moved by a third or more in at least some channels over the period. |
| History | Seasonality is the biggest non-marketing driver of sales. The model needs to see the same season twice to tell a seasonal lift from a marketing lift. | Two years of weekly data is the comfortable floor; six months is the technical minimum, with wider ranges as the price. |
| A sales signal | Revenue, orders, leads, sign-ups or approved applications: one outcome the model is built around, measured consistently, ideally daily. | One clean outcome series beats five partial ones. |
One model per brand and market
The threshold applies per model. A group with several brands or markets gets one model per brand and market, so the question is whether each of them clears the bar on its own. Where one brand has thin data, the model can borrow signal from a related brand in the same group, which helps a smaller brand get a usable model sooner.
What to do if you are below the threshold
Three sensible options, in order of how much they cost.
- Start with the data. Collect spend from every channel, online and offline, and your sales data into one structured dataset. It is useful on its own for reporting, and it is what the model will need the day you cross the threshold. Our guide to data for marketing mix modelling covers what to collect.
- Create the variation yourself. If spend has been flat, plan a few deliberate changes: pulse a channel up for four weeks, pause another for four. Done with a little structure, those changes are what the model will read next year.
- Model the biggest brand first. In a portfolio, the largest brand usually clears the bar before the others. Start there and let the smaller brands join as their data grows.
What to do if you are well above it
Nothing changes in the method. Larger budgets give the model more to read, more channels can be modelled separately, and the budget-level question becomes the valuable one: at what point does the next unit of spend stop paying for itself? That is answered by marginal return at every spend level, which we go through in finding the right total marketing budget. Companies with budgets in the hundreds of millions run on the same model as companies at the threshold; the difference is how many channels, markets and products it is split into.
Where Odins fits
Odins is a managed marketing investment platform for any company with a marketing budget from EUR 1 million to several hundred million a year and a sales signal the model can learn from. We connect the data, digital and offline, build and maintain Bayesian models and deliver monthly recommendations our team reviews before you see them. If you are unsure which side of the threshold you are on, the first conversation settles it; we say plainly when the data is not there yet.
Frequently asked questions
Is there a minimum revenue for marketing mix modelling?
No. The model learns from marketing spend and its effect on sales, so the marketing budget and the sales signal decide the fit, not revenue, headcount or industry. A company with modest revenue and a marketing budget around EUR 1 million a year is a better fit than a much larger company that barely advertises.
Can we run MMM with six months of data?
Technically, yes. The model can be built, but it has seen each season at most once, so the ranges around seasonal effects will be wide. Two years of weekly data is the comfortable floor. With six months, strong starting assumptions about each channel carry more of the load, and the first months of forecasts should be read with that in mind.
We spend a lot, but almost all of it on paid search and Meta. Is that enough?
Usually, as long as the spend has varied. Two channels at steady levels are hard to separate; two channels that have moved independently are not. The model also needs the non-marketing drivers, such as price promotions and seasonality, so it does not credit the channels for sales they did not cause.
What happens if we are a bit below the threshold?
We say so, and we recommend starting with the data platform rather than a model. Structured spend and sales data from every channel is valuable on its own, and it means a model can be built the day the budget and the variation are there.
For the full background on how the method works, read our complete guide to marketing mix modelling, or book a demo and bring your budget and your sales data; the fit question takes ten minutes. How long the first model takes once you start, data collection included, is in how long marketing mix modelling takes to set up.