Is MMM worth it for a small brand?

Usually not yet. MMM pays off with two or more years of weekly sales, spend that moved, and several channels big enough to show in your totals. If you run one or two channels on steady budgets, a holdout test or a close read of your GA4 paths usually answers more, sooner.

The software will not stop you. Google's Meridian and Meta's Robyn are both free and open source. What a small brand runs short of is data, hours and a big enough budget for a better split to matter.

What one store's data shows

One store's anonymised GA4 export, 1 January 2024 to 21 August 2026. It holds shares of revenue only: no ad spend, no order counts.

Run this one store through the usual MMM checklist and it passes the easy test and fails the one that matters. The export covers about 137 weeks, which clears the two years of weekly data Meta's Robyn guide asks for. But it holds no spend, and Meridian's docs say the dataset must include spend for each paid channel. No spend, no model.

The Channels sheet shows why a brand would want a model anyway. On the Channels sheet, Direct holds 57.7% of revenue in every view, and Paid Social holds 0.0%. A path report can only credit what it managed to label. A mix model never reads labels. It reads weekly spend against weekly sales, so if social ads had run, it could look for their effect in the totals.

The Journeys sheet shows where attribution still earns its keep. On the Journeys sheet, journeys with 1 touch hold 79.5% of revenue and took 0.5 days to buy. Journeys of 2 or more touches hold 20.6% on the Journeys sheet, about a fifth. A weekly model would see most of this revenue land in the same week as the visit that led to it.

What the export cannot show is whether any channel moved sales at all. That takes spend data and a model, or a test. Neither fits inside a file of revenue shares.

Why does "MMM is free now" mislead?

Free software is not a free answer. Meridian's FAQ says the library is free to use and open sourced on GitHub. The bill arrives in hours and in data.

Start with the hours. Robyn's guide calls data collection the most time consuming step and says to allocate at least 4 weeks for it. That is before anyone fits a model or argues about the results.

Then the data. Meridian's docs walk through a cut-down national model: three media channels, two knots, three years of weekly data. That is 156 data points for 10 parameters, roughly 15 per parameter. Even then, the docs say you might be able to glean some directional information. Directional is their word, after three years.

Small budgets add a second snag. Robyn's guide asks whether an activity could have a big enough impact on sales to be measured at all. If not, it says the activity is likely better left out of the model. So the small channel you most want judged can be the one the model has to drop.

And you learn whether the data was enough only after the work. Meridian's docs call their data guidance rough and directional. The most accurate check, they say, is to run the model and see how wide its estimates come out.

What can MMM not tell a small brand?

• Which campaign to cut. Meridian's docs call MMM a macro tool that works at channel level and generally advise against campaign-level models.
• What next month will bring. Meridian's FAQ says it is designed for causal inference, not prediction, and cannot forecast future raw outcome. It plans budgets; it does not predict sales.
• What a channel too small to show did. If the guides suggest dropping it, the model is silent on it by design.

For those questions, a test is the cheaper tool. Google's user-based Conversion Lift needs at least 1,000 observed conversions and a minimum campaign budget of $5,000 USD. Below that size, your own holdout is the realistic option. Switch one channel off in some regions and compare their sales with regions that kept it.

What to do this week

1. Check that your spend ever moved. In Google Ads, go to the Campaigns table, select the segment icon and choose Time, then Week. Pass: weekly cost rises and falls, with some quiet weeks. Fail: it runs flat for months, so a model has nothing to learn from yet.
2. Check for a platform lift study. In Google Ads, go to Lift measurement within the Goals menu and select the plus button. Pass: Conversion Lift appears and your account clears its minimums. Fail: it does not, so plan your own regional holdout on your biggest channel.
3. Price the decision. In Google Ads and Meta Ads Manager, read last quarter's total cost per channel. Pass: a better split of that money would clearly pay for weeks of data work. Fail: it would not, so test one channel and keep collecting weekly data for later.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: FAQs (Google); Amount of data needed (Google); Collect and organize your data (Google); An Analyst's Guide to MMM (Meta); Set up Conversion Lift based on users (Google); Set up Conversion Lift based on geography (Google); Use segments in your tables (Google)

Related answers

• All MMM & measurement questions
• How to check if MMM is worth it for your brand, step by step
• Is MMM worth it for a home goods brand?
• MMM for brands under €10M: what Meridian will not tell you
• Marketing mix modeling