An MMM RFP is only as good as the questions it asks. Here is the structure, the questions that separate vendors, and a way to score the answers, with side-by-side comparisons to help you assess the options.
A Marketing Mix Modeling (MMM) request for proposal is only as good as the questions it asks. A weak RFP invites every vendor to answer “yes, we do that,” and you end up choosing on brand familiarity or price. A strong RFP surfaces the differences that actually predict whether the investment pays back: how often the model refreshes, what you own at the end, whether the methodology is explainable, and whether you can ever bring the capability in-house. This guide gives you the structure, the questions and a way to score the answers, then points you to side-by-side comparisons of the main vendors so you can assess the options.
How to structure an MMM RFP
Nine sections cover everything a vendor needs to respond well, and everything you need to compare them fairly.
Background and objectives
- What decisions will the model inform?
- Which markets, brands and channels are in scope?
- What does success look like in 12 months?
Your data and stack
- What data do you hold, and where (Snowflake, BigQuery, other)?
- What is the history and granularity available?
- How should the model integrate with your stack?
Methodology requirements
- What modeling approach is proposed, and why?
- How are adstock, saturation and the base handled?
- How is the model kept current over time?
Operating model and cadence
- How often does the model refresh?
- Who runs it day to day, you or the vendor?
- How quickly are results available after a period closes?
Validation and causality
- How is the model calibrated against experiments?
- How is accuracy measured and reported?
- How are results stress-tested before a decision?
Transition and in-housing
- Can we bring the capability in-house over time?
- What training and enablement is offered?
- What does the path from managed to owned look like?
Tools and access
- What interface do business users get?
- What access do our data scientists get to the model?
- What do we retain if the relationship ends?
Commercial model
- How is pricing structured, and what drives it?
- Is pricing tied to media spend or fixed?
- What is the all-in annual cost including our time?
Proof on your data
- Will you prove the approach on our own data first?
- What does a paid proof phase include?
- What do we keep from the proof regardless of outcome?
Download the editable RFP template
A Word version of these nine sections, with the vendor questions built in and a weighted scoring scorecard ready to fill in.
Download the MMM RFP templateThe questions that separate vendors
Most vendors clear the basics. These are the questions where answers genuinely diverge, mapped to the criteria that predict success. Lift them straight into your RFP.
Ownership and lock-in
- At contract end, do we keep the model, the pipeline and the insights?
- Is the methodology documented and explainable to our team?
- What is your client retention model: partner or permanent dependency?
Refresh cadence
- Is the model rebuilt periodically or refreshed continuously?
- How many days from period close to usable output?
- How is the model maintained as our channel mix shifts?
Transparency and trust
- Can you walk our finance team through how a result is derived?
- Is this a proprietary black box or an explainable model?
- How do you show uncertainty in the estimates?
Coverage and the base
- Does the model account for offline media, promotions, pricing and seasonality?
- How do you measure the long-term brand base, not just media?
- How is the impact of pausing brand investment quantified?
Causal validation
- How do you calibrate the model with geo-lift or conversion-lift tests?
- Can you give an example where a test changed the model’s read?
- How do you avoid presenting correlation as causation?
In-housing and training
- What is the route from managed delivery to in-house ownership?
- What structured training do you provide our team?
- How long until our team can run and validate the model itself?
Data integration
- How do you connect to our data warehouse?
- What is the effort to build and maintain the pipeline?
- How is data privacy handled, and is individual-level data required?
Commercial fairness
- Is pricing fixed or a percentage of our media spend?
- What is the total cost of ownership over three years?
- What happens to cost as our spend or scope grows?
Where to go next
Use these guides to compare vendors and to settle the build, buy and in-housing questions before you issue the RFP.
Frequently asked questions
What should an MMM RFP include?
A strong MMM RFP covers nine areas: background and objectives, your data and stack, methodology requirements, operating model and cadence, validation and causality, transition and in-housing, tools and access, the commercial model, and proof on your own data. Each section should force a clear answer on at least one of the criteria that predict success.
What questions should I ask MMM vendors?
Beyond the basics, ask the questions where answers genuinely diverge: what you own at contract end, whether the model refreshes continuously or periodically, whether the methodology is explainable to finance, how it covers the brand base and offline media, how it is validated against experiments, whether there is a path to in-housing, how it integrates with your data, and whether pricing is fixed or a percentage of media spend.
How do I evaluate and score MMM proposals?
Weight the criteria before reading any response, typically placing methodology, validation and refresh cadence highest, then ownership and in-housing, then commercial model and integration. Score every vendor against the same rubric, and treat reluctance to prove the approach on your own data as a negative signal.
Should an MMM vendor prove the approach on our data first?
Yes. A credible provider will validate the approach on your own data before a full commitment, often through a paid proof phase, and will let you keep the outputs regardless of whether you proceed. This de-risks the decision and tests the relationship.
Put your shortlist to the test
The fastest way to separate vendors is to ask each to prove the approach on your own data. A wastage assessment does exactly that: it uses your own spend and channel data to estimate the recoverable waste in your current setup, with no personal data required and no rip-and-replace. You keep every output regardless of whether you continue.
Start your wastage assessment