Kauzify
Marketing Mix Modelling

Know what each marketing channel really brings in, and where to put your next dollar

Marketing Mix Modelling (MMM) measures the impact of every marketing activity on your business KPI, and shows how to set and split your budget across channels and platforms.

Meridian
Built with Meridian, the open-source MMM developed by Google.

What MMM lets you do

Measure each channel's impact

See how much of your KPI each channel and platform actually drives, separated from your baseline, seasonality and other factors.

Optimize your budget

Find the return on investment of each channel and how to reallocate spend across channels and platforms to get more from the same budget.

Our process: the CRISP-DM data science cycle

Every MMM project follows the same proven steps.

  1. 1

    Business & problem understanding

    We learn your business, your marketing and the questions the model must answer.

  2. 2

    Define success criteria

    We agree on the KPI, the decisions the model supports and how we judge it accurate.

  3. 3

    Collect data

    We gather the KPI, marketing activity by channel and platform, and control variables.

  4. 4

    Validate & prepare data

    We check, clean and align every series so the model learns from reliable data.

  5. 5

    Modelling

    We build and tune the model of how each channel drives your KPI.

  6. 6

    Evaluation & results

    We check the fit, then deliver channel impacts, returns and budget recommendations.

A living model that keeps getting more accurate

We don't deliver a model once and walk away. We keep the model, refresh it with new data, and calibrate it with incrementality tests until it reflects your marketing environment closely.

  1. Month 0

    First model version

    We release the first version of your MMM with its results.

  2. Months 0–6

    Incrementality tests

    We plan and run incrementality tests to validate the model's channel estimates.

  3. Month 6

    Refresh & calibrate

    We refresh the model with the new data and calibrate it with the test results.

  4. Every cycle

    Repeat

    New tests, new data, a new refresh: each cycle brings the model closer to your reality.

Example results

Example output from a model run on sample data. Channel names are anonymized.

Model fit: expected vs. actual revenue, with the baseline without marketing.
Model fit: expected vs. actual revenue, with the baseline without marketing.
Contribution over time of the baseline and each marketing channel.
Contribution over time of the baseline and each marketing channel.
Share of revenue from each channel, with spend and return on investment.
Share of revenue from each channel, with spend and return on investment.
Return on investment by channel, with its credible interval.
Return on investment by channel, with its credible interval.
Response curves: how incremental revenue grows with spend in each channel.
Response curves: how incremental revenue grows with spend in each channel.
Recommended budget allocation: the spend change for each channel.
Recommended budget allocation: the spend change for each channel.
Incremental revenue gained by the optimized budget, channel by channel.
Incremental revenue gained by the optimized budget, channel by channel.
Each channel's current and optimized spend on its response curve.
Each channel's current and optimized spend on its response curve.

What we need from you

At least 2 years of data for everything:

Your KPI
The business metric to explain, such as revenue, sales or sign-ups.
Marketing activity
Spend and activity for each channel and platform.
Control variables
Factors that affect the KPI but are not marketing, such as competitors, discounts and price.

Where we build it

It's flexible: we can develop the model in our environment, or directly in yours.

What you receive

  • An HTML report with charts and results
  • CSV files with the results
  • A PowerPoint presentation that we walk you through

Ready to see what your marketing really drives?

Discuss your MMM projectPricing