MASS Analytics

So more and more brands are in-housing their marketing mix modeling operations into a central team, or a Center of Excellence, managing the MMM projects for all the markets they operate in.  And it makes sense. From the global HQ perspective, this setup provides a unified dashboard to compare and optimize media budgets across countries. Within each country, further optimization can be done to determine the best channels and media investments.  So reporting, KPIs, measurement, and optimization are aligned because it is done by a single team. However, where this approach could go wrong is when the local marketing teams are not involved in the decision-making process.  They might see the central team as imposing strategies without considering local conditions & insights, then impose data requirements and deadlines on them.  This just leads to friction, delays and resistance in data sharing and implementation of recommendations. The focus needs to be on the ADOPTION of MMM results. That is the end in mind.  This is why I recommend an iterative approach to MMM implementation.  In the iterative approach, we make sure the local teams’ perspectives are taken into account as part of the MMM model iterations.  We listen to what they say and take into consideration their hypotheses when it comes to finalizing the results. They are part of the process, and the results are shaped with their perspectives in mind.  So when it comes to adopting these results, you get something that is already adopted by the local teams because they have been part of the journey.  Adoption is key. You need to win that battle first. Otherwise, you'll be spending millions of dollars on MMM and nobody in the local marketing teams will be applying them because they don't feel that it represents their reality.  

Before you even claim that a marketing channel reached saturation, you’ve got to look at all the elements that make up its execution. In media planning, the focus is on what channels to invest in and how much.  But that's not enough, because the way you execute spend within a channel matters. A marketing mix modeling analyst should not just rush to cut budgets if a channel starts to underperform. They need to identify WHY that channel underperforms and suggest improvements. You might be spending $2 million on programmatic ads, and it seems saturated. And you know what? With the way it has been executed, it probably did saturate. However, changing the execution could probably allow you to spend up to $5 million or more because factors like targeting, format, messaging, and laydown all matter. So if I give you an additional $1 million to invest on advertising, as an MMM analyst you need to tell me:  - How many campaigns should I be running? What type?  - When should I be running them?  - What type of message should I have on them?  - How can I execute this media week-on-week in a way that doesn’t saturate one week over the other? (To take advantage of the carryover effect in the campaign build up) All those components will make it very difficult to saturate a channel. And the chances that all these combinations have been fully optimized by the brand or agency are slim, meaning there is probably room for better execution. This is not to say that saturation never happens. It obviously does.   What I’m saying is, let’s not jump to conclusions. But if you don’t care how that media spend would be executed, because you think that’s too complicated, or that’s not part of your job…  Then you are only providing half of the answer to your client. 

Recently debriefed MMM results for a $1+ billion revenue retailer, this is what happened: So this retailer had never invested in television advertising in their key region. Now their marketing team strongly believed that TV ads would significantly boost their performance. They knew they worked, and they saw their competitors using them effectively in recent years. Only problem? They struggled to convince the CFO to allocate the budget due to the high costs associated with TV advertising in this region. So when they approached us for a marketing mix modeling project, one of the key questions they wanted answered was whether or not to invest in TV.  Fortunately, they came prepared. They had data for their competitors’ TV campaigns that we can model. After we thoroughly analyzed this data over key business metrics, what we learned was huge: Competitors’ TV activity caused revenue loss estimated at over $50 MILLION. Not only that, but the analysis also showed that this led to a 2% decrease in their brand consideration. So when the CFO saw just how much TV advertising by competitors was siphoning off their sales, he understood that something must be done to counter these declines. Based on this insight, the company agreed to a pilot investment in TV ads during a high reach season to assess its commercial effectiveness. Marketers, if you’re struggling to convince your leadership, remember this:  Your stakeholders are on your side, they just need the evidence to support you. With the right data and robust modeling, you can make the case to drive your marketing forward. 

The best MMM analysts I have known are the ones that are convinced that the data preparation phase is not a burden. They realize it’s a key success factor in the whole process, and they give it the time and attention it deserves. The pre-modelling phase needs to be done properly, without jumping into conclusions. And to do that, you need to have attention to detail. When you look at something, you should seek to discover what it's telling you, not just look at it with zero critical thinking. I fondly remember my ex-boss Marion, back when I was starting in MMM. She used to graph (as in physically print copies of graphs) every single variation of every single dependent variable with every single independent variable, post and pre transformation. And she would spend days looking and contemplating those graphs in her “yellow folder” and really understanding what the data is trying to tell her.  And then based on that, she would go to the analyst and request further data transformation. And she was in most cases spot on when it comes to creating the right variables that will make sense to the model.  Obviously, a lot has changed since then. But we need to bring back that spirit of in-depth enquiry while leveraging the new capabilities available to us today. Unfortunately, we are now losing this more and more as people think that we have this AI-automated MMM all figured out.  They think that everything would be done by the machine, but we know that is not true and that there is a big part that still needs to be done by the analyst. People are in such a rush to get into modeling. They want to reduce the time for everything that comes before that and jump straight into the econometric equation. When we know through experience that this is not the optimal way to handle your MMM!  There is a great deal of attention and due diligence that you need to make before modeling begins.  If you have an analyst who is not interested in the pre-modeling part (data preparation, data exploration, identifying trends…) then obviously what you will end up with in the final result will be poor quality transformation, poor quality data, and obviously a poor model.  Because what could happen if you disregard that first part of pre-modeling is that you will just believe whatever result the model says to you without really exerting any type of due diligence to make sure that the variables as they come into the equation, the coefficient as they are estimated, make sense business wise. 

One of the traits that drew me to the team at MASS Analytics was their enthusiasm for analytics excellence and business growth. When making my decision that was not the only reason to accept the offer to be part of the team. In this blog, I talk about the 4 reasons why I joined MASS Analytics. 👉🏻 https://bit.ly/45lS2Ys