How Long Does a Marketing Mix Modelling Project Take?

A marketing mix modelling project typically takes 8 to 16 weeks from kick-off to final results. The biggest variable is how long it takes to collect and clean the data. If your data is ready at the start, you can expect results in 6 to 10 weeks.

Timeline surprises are one of the most common complaints marketing teams have about MMM projects. Setting realistic expectations at the start means you can brief your CFO and CEO on when to expect the output, without having to revise the date three times.

Phase 1: scoping and data collection (weeks 1 to 4)

The project starts with a scoping call to agree on the outcome variable (what the model will explain), the channels to include, the time period, and the level of geographic or product granularity needed. This usually takes one to two weeks.

Data collection follows. Your team needs to pull media spend from your agency, revenue data from your finance or analytics team, and pricing data from your commercial team. This is the phase most likely to overrun. Allow three to four weeks even if you think it will be faster.

Phase 2: data cleaning and validation (weeks 3 to 6)

Once the data arrives, the modellers check it for gaps, inconsistencies and outliers. An outlier is a data point that sits far outside the normal range, such as a week where sales spiked because a celebrity mentioned your product on social media. These need to be identified and handled carefully so they do not distort the model.

This phase often reveals missing data or discrepancies between what the agency reported and what finance recorded. Resolving those discrepancies takes time and requires input from your side.

Phase 3: model building and calibration (weeks 5 to 10)

With clean data in place, the modellers build and test multiple model specifications. A specification is a particular set of choices about which variables to include and how to transform them. The team runs many iterations and selects the one that best fits the data while remaining explainable.

If you are using Bayesian MMM (a version of the method that incorporates prior beliefs about channel effectiveness), the calibration step also involves reviewing and agreeing those prior beliefs with you. This adds a week but produces more reliable results.

• Model building typically takes 2 to 4 weeks once clean data is available
• Bayesian models add 1 to 2 weeks for prior setting and calibration
• Each additional market or product segment roughly doubles the modelling time

Phase 4: results, validation and optimisation (weeks 9 to 14)

The modellers present a first set of results and walk you through the key findings. This session usually surfaces questions that require further analysis, such as breaking results down by region or testing a specific budget scenario.

After any revisions, the project moves to budget optimisation. The optimiser uses the model coefficients to find the spend allocation that maximises expected revenue within your budget constraints. This output is what most marketing directors use in their planning process.

What slows projects down most often

The single biggest cause of delays is slow data delivery. Media agencies sometimes take two to three weeks to pull historical spend data, especially for older years. Starting that data request before the kick-off meeting saves significant time. The second most common delay is internal sign-off on the scoping document or the prior beliefs used in Bayesian models.