Raising Media-Driven Revenue With Market Mix Modeling

The reason why we measure more and more is straightforward: so that we can forecast the performance of different strategic scenarios, and thereby help the brands we partner with optimize their media efforts. And just like any other discipline within advertising, the field of media continues to evolve, so let’s put a spotlight on what matters right now and will support your media measurement. 

Welcoming incrementality in the media world. 

First, let’s take a step back and look at what incrementality entails. Simply put, it refers to the lift in conversions or sales that can be attributed to a specific advertising campaign above those that would have occurred regardless—also known as the base. Incrementality has recently been adopted by us media folks, and the term has risen in importance because it’s a media measurement solution that isolates the incremental uplift. This matters because otherwise you can’t tell which media is driving growth and which is just harvesting conversions that you would have gotten anyway. As such, incrementality delivers a far more accurate view of how your media channels are driving conversions.

For example, traditional multi-touch attribution (MTA) often fails to separate the base from the uplift of the advertising campaign. This can lead to overstated results. Instead, in order to accurately measure incrementality, it's important to use MTA in conjunction with incremental techniques like market mix modeling (MMM). This way, you can better understand the true impact of advertising campaigns, move from ROAS to ROI, and as such have a more sensible conversation with your finance teams on the effectiveness of media.

How market mix modeling has got media measurement’s back. 

Market mix modeling—sometimes referred to as media mix modeling, but I prefer the former—is certainly not new to the scene, and this technique has been around in its commercial application to understand media uplifts for several decades now. However, the discipline has significantly improved, especially in the last few years.  

Contemporary MMM has come a long way. In the old days, annual updates would take months to bear results, while today you can get a pilot up and running within six weeks and use automation and machine learning to obtain monthly updates in just a matter of days. Besides, visualizations have also become much better, as today’s reporting dashboards offer analysts a plethora of ways to approach the data sets.

It's important to note that market mix models consider the whole market—including drivers like promotions to pricing, the recent pandemic, seasonality and more—and thus offer a holistic view. If you fail to take these other factors into account, you can’t get an accurate read on media and risk overstating its impact. As such, we’re seeing more and more brands partner with specialist MMM experts to help build the market mix models, or work with them to in-house this capability.

I have to point out that some players out there might say they execute “media mix modeling,” but are actually just building a simple regression with media variables or using multi-touch pathway techniques (which is not an incremental analysis). What’s so concerning about this is that they offer so-called MMM solutions at very cheap rates, which may sound appealing, but the damage of using these cannot be underestimated. Basing your decisions on a cheap but bad model could go wrong and cost you over 40% of your media-driven revenue—compared to an increase of roughly 30% if the technique is applied properly. You can make the call on what’s best for your brand.  

Leveraging AI to accelerate our analysis. 

Another very timely reason why I’m so excited about applying market mix modeling is the recent rise of artificial intelligence and the automation solutions that have stemmed from it—AI has been advancing fast in various areas, and it did not forget about MMM. 

At Media.Monks, we’re bullish about AI. That said, we also know that it’s important to be cautious and do our due diligence, especially as we see many AI providers claiming to build market mix models without having the right experience and tools to do so. When it comes to MMM, we believe that AI and automation solutions can be incredibly useful in speeding up the process, but of course there are also some instances that require manual labor. Let’s take a look.  

Raw data and processing. This can be automated using APIs or templates to stream data in, and then pre-ordained processes automate cleaning, saving lots of time. Beware of providers who take several months to initially onboard data pipes, as you really should be up and running in a matter of weeks.

Initial models. We use evolutionary algorithms to automate the initial model build, running thousands of models instantly in the cloud and scoring them, which enables us to arrive at a base model much faster and save weeks across MMM projects with multiple KPIs.

Final models. Note that this (still) requires manual intervention with a very experienced modeling team. We need to sense-check the models, triple-check the data, and use our extensive experience to spot any anomalies and alternative analysis to interrogate any controversial findings.

Sales effects and ROI calculations. These can be automated without the use of AI—this is just a process that can easily be repeated using code.

Automated reporting. Once all the numbers are calculated, it’s easy to automatically populate dashboards and media optimization tools. One thing that can’t be automated, however, is the answering of bespoke client questions around most effective second length, audience, and more. 

Engagement. Reporting ROIs and optimizations is one thing, but gaining an understanding of and trust in the models is another. Therefore, in the early stages of MMM engagements, it's imperative to have people who can explain the models and results to the wider team—not just marketing, but also finance, sales, the board, to name a few. My advice would be to circle back to this in later stages, once people understand and trust the model, and then you can move to more automated reports.

In short, automation can replace a lot of the heavy lifting of data and results processing and visualization, while AI can be used in the initial modeling stage. But what can’t be replaced is the sense-checking, interpretation, and experience of a good modeler to ensure the results are robust, realistic, understood and therefore usable.

Decreasing time, while increasing results. 

In the context of economically uncertain times, a time-saving—and thus cost-saving—solution like market mix modeling, especially when it’s powered by AI and automation, comes in very handy. Based on these models, media measurement typically enables brands to forecast different sales scenarios. In turn, having a robust forecast of performance is critical in justifying different strategic scenarios to the board, owners and investors of a company.

Incrementality is critical in the quest for accurate ROI, and MMM is a main way to get there. Though this technique has been around for decades, its pace of change and adoption rate is accelerating, which I’m sure will be further driven forward by AI. That said, in order for you to reap the many rewards of this tried and tested technique, it’s critical to work with a media partner who includes the whole mix of sales drivers and can take your models from sheer numbers to clear business actions.