MMM-PyMC3-Marketing

What is a Marketing Mixed Model?

Think of MMM, or marketing mix modelling, like a recipe book for a business’s marketing efforts. It helps figure out which ingredients (like online ads, social media, or big sales events) give the best results in terms of sales. By looking at past data, MMM works out how much each part of the marketing mix contributes to sales and helps decide where to spend the next marketing dollar to get the most bang for the buck. It’s like having a financial GPS for your marketing road trip – it points to where you might want to go next based on the paths you’ve taken before.

What can the MMM model help the marketing team understand?

MMM helps the marketing team answer some crucial questions:

1. Efficiency: Which parts of our marketing spend are giving us the best return? Is it social media ads, TV commercials, or something else?

2. Allocation: How should we split our budget across different channels? Where should we invest more and where can we cut back?

3. Timing: When is the best time to run our campaigns? Are there certain times of the year when our marketing is more effective?

4. Impact: How do our marketing efforts affect sales? Can we quantify how much each campaign actually contributes to our bottom line?

5. Forecasting: Based on past performance, what can we expect in the future? How can we predict and plan for the impact of our marketing spend?

6. Optimization: How can we tweak our marketing strategies to improve performance and get more value from our investment?

Basically, it's about understanding what works, what doesn't, and how to get the best results with the resources available.

Why do we need an MMM model, can't we simply track the way customers reach us using cookies?

The marketing mix model (MMM) came into the spotlight, especially as privacy concerns grew and the digital world began to clamp down on tracking user behaviour. With regulations like GDPR in Europe and CCPA in California, plus browsers phasing out third-party cookies, the good old days of following customers’ every click to see how they respond to ads are slipping away.

This shift means marketers can't rely as heavily on detailed tracking data to see how effective their ads are. So, they needed another way to understand what's working. Enter MMM. It steps back to look at the bigger picture, using broader data like sales figures, marketing spend across channels, and even external factors like the economy or seasonality to help untangle how different parts of marketing efforts contribute to sales.

MMM doesn't need to track individual customers to make these insights. Instead, it works with aggregated data over time, sidestepping the privacy concerns that come with detailed user tracking. It's a more privacy-friendly way for businesses to make smart decisions about where to spend their marketing dollars.

References:

1. Mmm example notebook#. pymc. (n.d.). https://www.pymc-marketing.io/en/stable/notebooks/mmm/mmm_example.html
2. Sanchez, M. (2024, March 13). Why is validation challenging in MMM? what are the different ways to validate?. Recast. https://getrecast.com/why-is-validation-challenging-in-mmm/
3. Jin, Y., Wang, Y., Sun, Y., Chan, D., & Koehler, J. (2017). Bayesian methods for media mix modeling with carryover and shape effects. Google Research.
4. Anderson, Anthony (2024). Multi-Region Marketing Mix Modelling (MMM) Dataset for Several eCommerce Brands. figshare. Dataset. https://doi.org/10.6084/m9.figshare.25314841.v2