Mutinex’s Growth OS and the Evolution of MMMs
Published in Partnership With Mutinex
At Madison and Wall, some of us are, by training, financial analysts first and advertising industry analysts second.
For many years, the advertising industry has tried to position itself as data-driven. Every service provider, technology company and media owner will claim to have the best data, the best methods for organizing and analyzing data, and the best marketing recommendations based on said data. Marketing mix models (“MMMs”) were one particularly important product that relied on data and was critical in supporting data-driven marketing. As financial professionals, and notwithstanding the critical role of intuition in a world that can’t always be compared to the world that came before, we have always believed more marketers should rely more on these tools. However, there have always been practical reasons why usage has been limited or ineffective.
In the middle of the century’s first decade, we saw relatively simple (if laborious) Excel-based models that accomplished this task. From at least the mid-2010s there were software products automating the ingestion of data and production of insights used to inform the choices marketers made. At the time, we couldn’t help but wonder if there were some “smoke and mirrors” in the products of that era. However, following our recent interview of Mutinex’ Henry Innis, we asked to see a demonstration of his company’s GrowthOS. It’s evident that great strides have been made in this field over the past decade.
Here were several important features of GrowthOS which we thought were noteworthy:
It ingests unstructured data
It relies primarily on generalized marketing dynamics (and a much bigger body of research and data) rather than individual marketers’ spending and sales patterns to anticipate marketing outcomes
It applies artificial intelligence to produce written analyses, PowerPoint texts or other outputs that its users require
More generally, our first impression of the product is that it is very user friendly, relying as it does on natural language. Do you want trend graphs of performance by channel over time? Or channel specific recommendations for the coming quarter? Or maybe you want to “double click” on the relative performance of TV advertising. Or perhaps you want a geographical analysis of historical marketing spend/success. You can ask and you will receive!
Taking a step back to look at this space more broadly, as financial professionals and mindful of tools used to drive optimization in other industries, we can see how the following attributes should be important in the evolution of MMMs, regardless of provider:
New data and market conditions should cause underlying marketing mix models to evolve. At the same time, they should be able to explain why it makes one recommendation now vs a different recommendation in the past
MMMs should be appropriate for the user. If there is a meaningful difference between underlying models and different segments of marketers, there should be different MMMs
Users should not be burdened by having to use an MMM with a fixed frequency of data updates. Instead, each user should be able to determine (a) when enough new data is available to justify an update to the analysis and (b) at what phase in the planning process the insights are most useful
Whatever volume of data a marketer has should be sufficient to power the model
A user should be able to implement real world constraints into the model, so the “optimized results” are truly actionable
More to point, the tool should heavily inform the choices that marketers actually make
The entire process (data input, model optimization, report creation) needs to be fast enough so that the insights don’t become stale and there is a true payoff in terms of value vs time spent in the tool.
As discussed earlier, there has been no shortage of data, software and service companies who hype their ability to transform data into actionable insights. We don’t doubt there are still improvements to be realized across this space.
Still, while marketers are not always equipped with well-resourced teams of data scientists, they have usually had plenty of data to work with, and they need the ability to transform their data into something useful. Insights need to evolve just as the industry does. To the extent that Mutinex’s product maps to these needs and goals for a given marketer, superior outcomes should follow.