What is MMM?
MMM, or Marketing Mix Modelling, is a popular marketing technique used by companies to determine the effectiveness of their marketing activities through the application of an analytical framework that includes four different aspects, namely product, price, promotion, and place (or distribution). It’s normally applied as part of their Sales Forecasting and Planning process but has also been used to gauge the overall effectiveness of their advertising strategy.
To model the sales outcomes over time, MMMs use aggregated historical time series data with variables representing advertising, marketing, and control variables such as weather and seasonality.
The return on advertising spend (ROAS), and optimized budget allocations for advertising are derived from these models, assuming that they provide valid results. The purpose of MMMs is to provide advertisers with answers to questions like these:
1) What was my ROAS on Facebook ads last year?
2) What would my revenue be if I increased or decreased my advertising budget next year?
3) How should I allocate my media budgets to maximize profitability or revenue?
4) How do seasonality, holidays, or day of the week impact sales?
5) How do macroeconomic factors affect sales?
In most cases, MMMs are regression models based on a limited amount of observed data, and such models produce correlational, not causal results.
Why should you implement MMM?
These days, post-IOS 14, marketing mix modeling is especially important. Marketing Mix Modelling is going to become more and more important in the near future, especially since chrome is phasing out third-party cookies.
Even if you have robust multi-touch attribution software like Northbeam or Triple Whale, it’s important to have multiple data points to have a holistic view.
There are companies that help you to set up Marketing Mix Model. However, their offering is catered toward big companies. Their fee can start from about $8k/mo, which is something, not every small business can find a cost-efficient solution.
Best Free Marketing Mix Modelling Guides for Beginners
It’s probably one of the best, and easiest MMM excel articles out there.
This comprehensive guide will show you step-by-step how to create your marketing mix model using excel. You don’t need to have any programming skills or statistics knowledge.
Just follow along and create your model from scratch.
Make sure to read the Q&A section, as there are some good nuggets.
This is a more advanced guide in excel. Or at least I found it so. I recommend checking it out if you already have some knowledge or if you watched the guide above.
Supercharge your Marketing Mix Model with Machine Learning. A Free Guide.
If you want to improve the efficiency of your model and speed up the modeling process, you can implement Meta’s Robyn MMM. I generally found it more accurate than the models in excel I built. This guide will teach you how to build a model using the programming language R and Meta’s Robyn Package.
You don’t need any programming language knowledge, but it’s definitely recommended to have some.
Make sure to read the notes in the R code file, as Meta is constantly updating the Robyn package.
One thing that I found tremendously helpful is reading this guide.
One thing that I did to improve my Robyn model efficiency:
Changing adstock techniques from Geometric to Weibull PDF. This allowed my model to show a more accurate ROI by channel, especially social media platforms.
One other thing I did that almost doubled models accuracy was changing from daily data to weekly data. It allowed the model to make better predictions by eliminating the noise.
Google’s Lightweight Marketing Mix Model (LMM)
Google’s Lightweight Marketing Mix Model (LMM) is one such model that was developed by Google engineers specifically for small businesses.
This model lets you do geo-level modeling right out of the box. This differentiates it from Facebook Robyn, which doesn’t offer this feature currently.
Bayesian approach to Marketing Mix Modelling allows us to integrate previous data into modelling, allowing us to:
- Use Bayesian priors to develop media mix models based on industry experience.
- Provide a report on both parameter and model uncertainty, then apply it to the budget optimization process.
- Use breakout dimensions such as geography to construct hierarchical models with generally tighter credible intervals.
Here’s one guide by Vexpower that touches briefly on this model that I recommend to check out.
It doesn’t look like there are comprehensive Lightweight Marketing Mix Modeling guides for beginners at the moment. However, if you have some Python knowledge, you can just read their GitHub and try to implement your model.
Marketing mix modeling is a powerful tool that can help you to analyze your marketing data.
It allows you to look at the relationships between different variables, which can then be used to draw conclusions about how your business performs. This can be useful for understanding what areas of your business are performing well and which ones need improvement.
One of the most important things to remember when performing marketing mix modeling is that it is not an exact science. You need to keep in mind that the model itself is only as good as the data that goes into it. You should always use high-quality data and make sure that you’re using reliable sources.
You should also be mindful of what type of model you’re using. For example, if you’re looking at a multi-product firm, then you might want a multivariable model instead of a single-product firm’s simple regression model.
Finally, remember that even if your analysis shows that one strategy has higher returns than another, it doesn’t mean that they’re all equal—you still need to consider things like customer loyalty when making decisions about which strategies will give you the best results overall.