A good MMM brief defines the business decisions the model needs to inform, not just the data you have available. Methodology follows from the decision. Start with what you need to know and let the partner propose how to find it out.
Many MMM projects go wrong not because of modelling errors but because of a poorly written brief. The client asks for an MMM without specifying what decisions it needs to inform. The partner builds a model and delivers outputs. The client receives a detailed decomposition of last year's sales and has no idea what to do with it. The brief is where you prevent that outcome.
Section 1: business context
Explain your business clearly: what you sell, who you sell it to, how customers find and buy from you, and what your revenue model looks like. Include anything that makes your business non-standard: long purchase cycles, subscription revenues, multi-market operations, heavy seasonality, or a recent business change (rebranding, new product launch, new market entry) that might affect the historical data.
Section 2: the decisions you need to make
List the specific budget or strategy decisions that the model output will feed into. For example: we need to decide how to allocate a 15-million-pound budget across eight channels for the next financial year. Or: we need to decide whether to invest an additional 2 million in TV or use it to scale paid social. Being specific here shapes the model design. A model built for budget allocation needs different features from one built for campaign evaluation.
- State the decisions explicitly: what will we decide differently based on the model output?
- State the timelines: when do these decisions need to be made, and therefore when do you need model outputs?
- State the budget authority: who makes the final budget decisions and what evidence do they find credible?
- State any constraints: are there channel commitments or floors that the model output cannot override?
Section 3: your data
Describe your data availability honestly, not optimistically. List each data source, the time range available, the granularity (daily, weekly, monthly), and any known quality issues. If your spend data has gaps, say so. If your sales data changed methodology two years ago, flag it. Partners who discover data problems mid-project face delays and cost overruns that a clear brief would have avoided.
A brief that overstates data quality wastes everyone's time. A partner who discovers your data is messier than described will either spend extra unplanned time cleaning it (at extra cost) or build the model on imperfect foundations. Be honest about what you have.
Section 4: success criteria
Define what success looks like in advance. Is it model accuracy below a certain error threshold? Is it a set of channel response curves that inform a specific budget decision? Is it a recommendation that your CMO and CFO both find credible and actionable? Writing success criteria before the project starts prevents a scenario where the deliverable is technically complete but commercially useless.
Section 5: stakeholders and process
Name the key stakeholders who will receive and act on the output. Tell the partner what level of statistical sophistication to assume in their outputs and presentations. Agree the review checkpoints during the project: a check-in after data ingestion, a review of interim model outputs, and a final presentation. Projects without agreed checkpoints drift. Agreed checkpoints keep the project on track and allow you to course-correct early.
How long should an MMM brief be?
Four to eight pages of substance, not counting appendices or data documentation. Long enough to give the partner genuine context and specific requirements. Short enough that they will actually read it carefully. A brief that is too short leaves too much ambiguity. A brief that runs to 30 pages of background material buries the important requirements in detail that the partner may not retain.
Should we share our brief with multiple partners before choosing one?
Yes. A competitive brief process, even with two or three partners, gives you a basis for comparison and often surfaces questions and considerations you had not thought of. Ask each partner to respond to the same brief so you can compare their proposed approaches on equal terms. A partner who asks good clarifying questions in response to your brief is likely to be a better collaborator than one who accepts everything without question.
We have never run an MMM before. How specific do we need to be about the methodology?
You do not need to specify the methodology in detail. That is the partner's job. What you do need to specify is the business context, the decisions the model needs to inform, the data available, and your success criteria. The partner proposes the methodology to meet those requirements. If you find yourself writing a brief that specifies Bayesian versus frequentist approaches without knowing the difference, you are writing the wrong brief.