Today’s best marketers are relying on Marketing Mix Modeling (MMM) to provide an accurate way, backed by science, to allocate budgets across all their channels. In this guest post, Michael, co-founder of Recast dives into the topic and why it’s critical to have good, fast data for your MMM.

We’re excited to share a guest post from Michael Kaminsky, co-founder and co-CEO of Recast, a startup that is re-inventing Marketing Mix Modeling for modern marketers. We hope you enjoy this.

Marketing Mix Modeling (MMM) has a long history. For more than fifty years econometricians and statisticians have been helping marketers measure the effectiveness of their marketing activity via complex algorithms and analyses.

The traditional MMM approach

The traditional approach to MMM was to treat every analysis like a separate project that involved:

1. Gathering and cleaning all of the relevant data

2. A human-intensive modeling process where statisticians used their judgment and experience to build a statistical model

3. A presentation of the results to executives

This whole process was very time-intensive and expensive, so most brands would only do this once or twice a year.

Common challenges

This process has a lot of shortcomings which have been discussed a lot in the last few years as MMM has surged back into popularity due to the reduced ability to track customers across the internet associated with changes like iOS14.5’s app tracking transparency (ATT) initiative as well as privacy-focused initiatives like GDPR. The main shortcomings are:

1. The marketing world changes quickly and marketers need to be able to respond faster than once or twice a year

2. The results of the MMM report are difficult or impossible to verify so the brands have to just trust that the statistician did everything right

3. The reports that the executives are looking at in the MMM don’t line up with the operational data marketers are using for actual day-to-day channel management

   
   

These are all critical problems that made MMM much less useful than it could be. A lot of these problems stem from the workload associated with gathering and cleaning all of the relevant data which historically existed in many different databases, some owned by the brand and some owned by a variety of different agencies (if they existed at all).

The modern MMM approach

Today, there is a huge opportunity to improve the approach to MMM simply by making the data required for MMM easier to access and work with. Once brands get all of their important historical marketing data into a clean data warehouse table, there are a number of immediate benefits:

1. The business can accurately report out on what’s actually happening (total spend by channel, blended ROAS or CPA) in a clean and consistent way;

2. MMM projects go much faster since the first step of aggregating and cleaning data is already done;

3. MMMs can be trained and deployed much more rapidly since the MMM model can run off of data pipelines connected to the data warehouse instead of from excel files cobbled together by hand;

4. All analyses are working off of the same single source of truth so there aren’t inconsistencies between the MMM and other reporting the marketing team is doing.

   
   

Brought together, these improvements are incredibly valuable. They both reduce the overall costs of running MMMs and actually allow the MMM to be more actionable and verifiable because the feedback loop can be much shorter.

In general we are seeing more and more businesses making it a priority to invest in the tooling to get all of their marketing data into a clean and regularly-updated marketing data warehouse whether they’re going to pursue an MMM project or not. Being able to continuously monitor trends in marketing activity and business performance is incredibly valuable on its own, and then once you’re ready to embark on an MMM implementation you’ll already have all of the relevant data ready to go.

That said, preparing data and building the model is just the first step. Turning the MMM model into better marketing decisions it's often the biggest challenge - read more here.

About the author

Michael Kaminsky is a co-founder and co-CEO of Recast, a startup that is re-inventing Marketing Mix Modeling for modern marketers. Michael was trained as a statistician and econometrician focused on helping people make better decisions. He's passionate about taking cutting-edge statistical techniques and using those to build tools that help marketers drive better performance for their businesses. Interested in reading more from Michael?

Check out his recent blog post on Lenny’s Newsletter about how top brands (including Clarisights customers Uber and HelloFresh) measure marketing’s impact.