This is a YouTube video titled "What is Marketing Mix Modeling and why is MMM validation important? | Modern Analytics Ep. 1" created in January 2024. The content is a presentation by Michael Kaminski, an econometrician and marketing science researcher, explaining the concept of Marketing Mix Modeling (MMM), its history, challenges with digital tracking, and the critical importance of validating MMM models to ensure accurate marketing budget decisions. The video is part of a series aimed at helping marketers understand and apply modern marketing measurement frameworks.
The incrementality revolution is here. Just measuring touch based attribution isn't good enough anymore. Modern marketers need modern methods to really know how much their marketing channels are actually driving their business instead of just relying on vanity metrics. Welcome to Modern analytics for marketers, our YouTube series that helps marketers learn and put into practice modern marketing measurement frameworks. Today we're going to cover a brief background on MMM and then we'll talk about why validation is so important. In future episodes, we'll cover all of the different ways to validate an MMM and how you can put these ideas into practice at your organization. I'm Michael Kaminski, an econometrician entrepreneur and marketing science researcher. Welcome to Modern analytics for Marketers presented by Recast let's start with a story. Let's imagine a brand called Gourmet Galaxy that sells pots and pans online as well as through retailers like Target, Walmart and Amazon. Gourmet Galaxy spends about $20 million a year on a mix of marketing channels. These include digital marketing channels like Meta and Google as well as offline channels like direct mail, radio and tv. How do the marketers at Gourmet Galaxy know how to invest their marketing budget? What's the true incremental ROI of those Facebook ad dollars? How much is TV advertising driving sales in store at Walmart and Target? If they just look at their digital tracking or multi touch attribution reports, digital channels that drive sales on direct to consumer website get all of the credit. But something's not right here. Gourmet Galaxy knows that their digital tracking reports are over crediting channels like branded search while giving almost no credit to tv. Even when their TV spin seems directly correlated to sales at Walmart and Target. What's going on? How can they make better decisions as a business? It turns out that marketing mix modeling can help to solve this problem, but only if the model has actually been validated. Before we get to the answer on how this works, let's cover some history of mmm. MMM has been around for a long time. For over 50 years, sophisticated marketing organizations from McDonald's to Uber to Airbnb to Coca Cola have been using statistical and econometric methods to try to understand how their marketing channels are working and how they can invest smarter to get more efficiency. Back before the Internet, there was no way to track a consumer from advertising touchpoint to a purchase. If you think back to the Madman era, people saw ads on TV or in print or heard something on the radio and they went into a store to make a purchase in cash. There was simply no way to know if an ad impacted the purchase for any individual person. So how did CMOs measure advertising effectiveness back then? Well, they used the power of math and statistics to measure advertising effectiveness. Brands hired statisticians and econometricians to compile their data on marketing activity and sales and then used statistical models to find find the patterns in the data. This allows them to answer questions like when we control for everything else, how many sales does an additional $5,000 worth of TV spend actually drive? Many consultancies were created to help these brands do this complex analysis. The problem was the process was slow and expensive since it required data experts to gather and clean all of the relevant data sets and expensive statisticians to do the actual analysis. Because of this, MMM models were really only built by the biggest brands in the world that had budgets to pay for these expensive and time consuming analyses. With the rise of E commerce over the last 15 years, MMM fell out of favor as brands started relying on digital tracking methodologies to directly match customers to the ads that they saw prior to purchase. Huge brands were built solely by advertising on channels like Facebook and Google Search while selling online. These were exciting times, but they masked deeper problems. It turns out that while digital tracking methods can tell you who saw or clicked on an ad prior to Purch purchase, they can't tell you if that ad actually caused the purchase. This was a small problem for some brands, but a really big problem for others. Beyond the inherent limitations of these methods, the glory days of digital tracking are over. Between iOS 14.5, app tracking, transparency policies, GDPR, Chrome, killing the Cookie, and a slew of other privacy promoting regulations and policies. It has gotten a lot more difficult to accurately track individuals across the Internet in order to see the full story of of what advertisements they saw or engaged with prior to purchase. Because of this, MMM has surged back into popularity. Brands large and small are evaluating how to use these types of models to measure the true incrementality of their marketing spend. There have been a bunch of cool new open source libraries released to the public, including Robin from Meta Lightweight, MMM from Google, and lots of marketing analysts are testing out these new methods for the first time. It's really exciting, but also really scary. MMMs are incredibly complex models and while there are millions of ways for them to go wrong, there's only one way for them to go right. Whether you're just getting started with MMMs or you've been working with them for a long time, it's critical to think deeply about how you and your organization will know that you've gotten to the right model. Or if you've just accidentally landed on a model that will end up leading you astray and potentially cost your company millions of dollars in wasted ad spend. With so much on the line, how can we make sure we've landed on the right model? With validation. Validation is the process by which we will prove to ourselves and others to the best of our ability, that the modeling results we've obtained are true reads on incrementality and can be used for decision making and budget optimization. Remember, that's what this whole exercise is about. If we can't trust the output of our MMM for budget optimization, the whole effort has been a waste of time. So in the remainder of this series, we're going to be talking about how to validate your model, how to build trust in the results, and how to help your organization drive millions of dollars of improved marketing performance. In the next video in this series, we'll talk about why validation of MMMs is so challenging, what people do that's wrong, and all of the different correct ways to actually validate your MMM model.
