A real MMM walkthrough from an expert

Published Mar 3 2023 4 minute read Last updated May 29 2024
MMM expert case study
Contributors Dropdown icon
  • Sean Dougherty
    Written by Sean Dougherty

    A copywriter at Funnel, Sean has more than 15 years of experience working in branding and advertising (both agency and client side). He's also a professional voice actor.

Marketing mix modeling can be quite an intricate and complex subject matter. There are loads of variables to take into account, lots of data to keep track of, and somehow it always feels surreal.

It’s enough to sometimes make your head spin — which is why we sat down with Charlotte Lundberg of the marketing intelligence agency Nepa. She walked us through a great deep dive of MMM with a real-life example that they implemented for one of their clients. 

There’s loads of good stuff below, so strap yourselves in. 


An overview of the case

Nepa’s client (we’ll keep things anonymous, of course), is a market leader in the retail space. As with many retailers, the client had traditionally focused on activating short-term sales with their ad spend. That is, they were looking at lower-funnel leads who would convert in a couple of weeks or a month. 

However, the client approached Nepa looking to incorporate longer-term tactics that would build brand awareness. The client wanted to get a sense of how much to spend on this long-term approach and how that might affect their sales-driven marketing efforts. 

That meant Nepa would need to answer a few core questions:

  1. Should they increase their total media budget or not?
  2. Is the split between different media channels well balanced, or should they re-prioritize?

What is the process for an MMM project?

Well, it starts with the main business questions above. Both Nepa and the client agreed on the areas of exploration, setting the scope of the project. 

They then needed to identify the data that would be needed for the modeling. Where would it come from? Who has access to it?

Once the data is collected, it is visualized to make sure everything looks correct, and also to make sure it will be valuable for the modeling. 

Step four in the process is the actual modeling. Once the modeling is done, both parties review the results and make a plan for next steps. 

See? Marketing mix modeling isn’t so bad once you break it down. So let’s take things a step further and examine each of those steps. 

What data is needed for MMM?

The answer will change depending on the objectives and the client. In Nepa’s case, they determined that they would need the client’s weekly sales — both online and offline sales. They also determined that they would need to factor in the weekly media spend, weekly brand tracking data, and other industry-specific variables. 

As with any other data modeling, the results become more accurate as you add more data. But you need to walk a fine line. 

Nepa decided to use three years worth of data. Any more than that, and they might have fed outdated information into the model. Think about it: if a business has existed for 50 years, you wouldn’t model 2023 marketing spend on sales data from the early 70s. 

Other factors can make your data too far from the current business reality to be useful. If you’re a younger company, your business dynamics and customer interactions are likely in a state of constant change - making historical data analysis only so useful. Also, if you’ve recently acquired another company or drastically changed your offering, you may need to base any modeling on data after those major changes. 

All of these factors are why Nepa tends to recommend using three years of data for the best models. 

Putting it all together in a tidy visualization

Once Nepa assembled all of the relevant data and applied some proprietary filters to reduce various noise, they visualized the data for review with the client. 

While it might seem like a simple step, it’s also incredibly important. Things happen, and data can become corrupted — whether by human error or other factors. It’s important to review the visualized data with the client’s to make sure everything is accurate. 

For instance, the client might remember an odd spike in sales a year ago that may not be reflected in the data. By seeing the visualization, and the lack of that one random spike, plans can be made to review and re-import any data that could be missing. 

Plus, what may seem like an odd spike in sales may actually be an emerging trend that you can take advantage of later. You never know. So make sure everything looks spot on before running the model. 

As the old builder’s saying goes: measure twice, cut once. 

Let the modeling begin!

To implement the modeling, Nepa ran two models for their client. The first model focused on the direct effects of their marketing (the short-term) on sales. The second model looked at the effects of that marketing on brand.

To be clear, this is not attribution modeling. Instead, MMM can consider long-term sales activities and it isn’t as sensitive to any regulatory changes (i.e. GDPR). 

By viewing the effect on sales and brand awareness, Nepa is able to calculate ROI per media. 

For this specific client, Nepa was able to see that by shifting some media spend, they could actually increase total sales by 15 percent without increasing the media budget. Wow! 

They could also use the model to analyze marketing performance against the law of diminishing returns. That is the rule of economics whereby you begin to see less incremental returns as you spend more. 

In this case, Nepa saw the client could still increase their budget by 15 percent while optimizing their media mix to achieve a 25 percent boost in sales numbers. Any more than that, and the increases in sales performance would start to shrink. This gave Nepa the ability to present a conservative and aggressive plan:

Conservative: Maintain budget and just optimize the mix = 15 percent sales increase

Aggressive: Increase budget 15 percent and optimize mix = 25 percent sales increase

The results of the modeling

While the client opted for the more conservative approach, they still implemented the suggestions and actually saw better than expected results. Great news, indeed. 

It just shows how powerful MMM can be for a business. That particular client didn’t need to spend any additional money, just spend their existing budget more wisely. 

If you’d like to hear Charlotta walk through the case, study be sure to check out our latest Funnel Tip above. And don’t forget to subscribe to our YouTube channel where you can get all the useful hacks and tricks you need to become a better marketer. 

Want to work smarter with your marketing data?
Discover Funnel