MMM Software or a Consultant for a Multi-Brand Company? What a Platform Can't Plan

Updated: 5 days ago
Short answer: MMM software models the data it's fed. It can give you quick answers and help you shift budget between channels, but it doesn't ask whether a competitor's price cut, a heat wave, a shelf change or a PR hit explains the swing in sales, and it can't build a plan for a portfolio of brands. For a company with several brands or markets, those gaps multiply, because every brand and market has its own version of those questions. The model has to be built by someone who understands strategy, because the strategy has to be inside the model.
When I built the first marketing mix model for one of my clients, a large online marketplace, the big measurement firms pitched wiring every data source together and letting the models run. The client later paid an outside firm close to a million dollars for a model that matched mine wherever we used the same variables, and missed weather, COVID, competitors, PR and TV everywhere else. That was one business. A company with fifteen brands across several markets has fifteen chances to make the same mistake.

1. What's the difference between MMM software and a consultant-built model?
Both use the same kind of statistics. The difference is everything around the math. A real marketing mix model starts with questions about the business: what changed, what competitors did, what the weather did, whether distribution or shelf position moved, whether the press picked something up. Each answer becomes a variable, a test or a reason to treat a stretch of data differently. Software skips that step. It takes the spend and sales data it's given, applies default settings and produces a result. If sales jumped in March because a competitor ran out of stock, the software credits whichever channel spent the most in March. I've written more about why that thinking has to come before the model in can't I just have AI build my marketing mix model.
2. Why do several brands or markets make software riskier?
Brands in a portfolio share some media, like a corporate TV flight or a retailer promotion, and run their own for everything else. One brand's promotion can pull sales from a sister brand. A premium brand and a value brand react to price, weather and competitors in different ways. Each market has its own retailers, competitors, seasons and data quality.
All of that has to be designed into the model. Sometimes each brand needs its own model. Sometimes brands can share information while keeping their own results. Shared media has to be split across brands in a way you can defend to each brand's leader. Software that pools brands that behave differently, or splits them so finely that each has too little data, gives you results that look precise and can still move money in the wrong direction.
3. Who should build a multi-brand model?
One of the biggest mistakes I see is handing the model to someone who only builds models: a statistician, a data analyst, or a firm that does nothing but MMM. The statistics matter, and they're the smaller part of the job. The person building the model has to know which questions to ask before it runs, and those questions come from strategy. Did a competitor reprice? Did one brand change its positioning? Was a launch delayed in one market? Do the brands in the portfolio support each other or compete for the same customer?
You need someone who has set marketing strategy, built spend forecasts and defended budgets at the CMO level, or who has worked closely with people who have. They build the strategy into the model. Without it, you get garbage in, garbage out, however sound the math. That's how I build models: I've run marketing as a fractional CMO and I have the PhD-level statistical training, so the strategy and the model come from the same person.
4. What can MMM software actually do for you?
Software is good at speed. Once it's connected, it can give you a quick read on how your channels compared and suggest small shifts in budget between them. For routine adjustments, that has value. But it works from the data it's fed, and much of that data comes from inside the walled gardens of Google, Meta, Amazon and the other platforms, each reporting on its own performance. So software can help you realign budgets a little within the channels it can see. It can't tell you what the rest of the business and the market were doing, and it can't develop a plan.
5. Why can't a platform build the plan?
A plan for a portfolio decides what each brand is for, which brands get more investment and which get less, how much to spend next year and where, what happens if a competitor moves, and how a launch or a price change will ripple through the rest of the portfolio. That takes forecasting spend scenarios, judgment about positioning and competitors, and someone who can tie the model's results back to the business strategy. A model built with that strategy inside it can support those decisions. A platform reading last quarter's data can only tell you what last quarter looked like.
6. What does each option cost for a multi-brand company?
For a portfolio, the cost depends mostly on how many brands and markets need their own model and how often you refresh. Software is a subscription: Improvado's 2026 roundup puts self-service platforms at about $24,000 to $60,000 a year and managed engagements at $50,000 to $200,000 or more. A consultant-built model is a project. Mine typically run $8,000 to $40,000, and the most complex builds, with lots of messy data and both regional and national models, can reach about $75,000. I break this down in how much marketing mix modeling costs.
7. When should a multi-brand company bring in a consultant?
For the first model, almost always. The first build decides the structure, the variables and the data pipeline, and every refresh after that inherits those choices. It matters even more when you have shared media, retail and ecommerce sales mixed together, offline channels like TV or direct mail, data that needs real cleaning, or results that will go to a board. I've been brought in by a healthcare client to rebuild a model after the first one, built elsewhere, couldn't be trusted. Rebuilding always costs more than building carefully the first time. Once the structure is right, routine refreshes can run in a tool like Google's Meridian, as long as someone keeps asking what changed in the business since the last run. I compare the free tools in Meridian vs. Robyn vs. hiring an MMM expert.
Building the case internally? See how to build a business case for marketing mix modeling, including how to find each channel's point of diminishing returns.
8. What should a multi-brand company ask before signing with anyone?
Ask how they'll structure the model for each brand and market. Ask how shared media and corporate campaigns will be split, and how they'll measure one brand pulling sales from another. Ask which outside factors they'll include for each market, such as competitors, weather, pricing, distribution and PR, and who decides. Ask who on their side has set strategy or built spend forecasts, not just models. And ask whether the results will turn into a plan for the portfolio, or just a dashboard. For the general hiring questions, see how to choose a marketing mix modeling consultant.
If you're weighing a platform against a custom build for a company with several brands or markets, I'm happy to look at your data and tell you what it would take. See how I build marketing mix models, or book a free 30-minute call.






