Zohar Bronfman is the CEO and cofounder of Pecan.ai, a predictive analytics platform built to solve business problems.
getty
Stressed about your business’s future? Maybe it’s time to look to the past for solutions.
History buffs know the past offers valuable insights for navigating the challenges of the present.
In today’s marketing world, we’re seeing the renaissance of a practical solution that was a notable innovation in its time.
That solution is marketing mix modeling or MMM. MMM is a powerful tool for understanding marketing effectiveness and planning marketing strategies.
And while it’s recently become a hot topic due to broader changes in the marketing and data landscape, MMM is surprisingly old. Initially developed from models used to predict how political campaigns would affect voting patterns, MMM’s earliest forms date back to the late 1960s. That version of MMM ran on computers programmed with punch cards.
You know that other thing about history—how it repeats itself? That’s true here, too. Today, MMM is seeing a renaissance in the context of AI and machine learning. Looking back at this history reveals a lot about successful marketing today.
Marketing With “Electronic Brains”
Jill Lepore, Harvard University historian, wrote in If/Then: How the Simulmatics Corporation Invented the Future about the company Simulmatics. Simulmatics was founded in 1959 and initially sought to influence U.S. politics through the power of data.
The company developed what the New York Times called “an electronic brain designed to estimate voter reaction to campaign issues [that] made strategy recommendations.” The idea was to have computers model when and where candidates should advertise to win elections.
However, along the way, Simulmatics realized this challenge applied beyond political campaigns. Ad agencies and marketers also struggled to decide exactly where to allocate their resources to achieve business goals.
But as Lepore describes, the Simulmatics experts realized that “... it’s not possible to build a model without data, and these media companies had surprisingly little data about who read the books and magazines they published or listened to or watched the records and movies they produced (television … was an exception).”
A lack of relevant data about marketing channels? Difficulties in understanding and planning for consumer behavior?
Are any marketers feeling this all-too-familiar vibe?
The early efforts at developing and refining MMM came from a similar scenario we face today. While we’d now replace some of those channels with mobile devices, influencer marketing and out-of-home advertising, the problem remains the same. There’s just not always enough data to guide the most efficient use of marketing resources.
The Data And Computing Gap
Ultimately, Simulmatics’ modeling—sold as “Media-Mix”—simulated an ad campaign to predict its effectiveness. Ad agencies that collected consumer data could use Simulmatics’ methods to predict campaign performance more accurately.
In the 1960s, collecting data for model building was much more difficult, however. The Simulmatics team even had their children generate data from television schedules.
The slow, laborious process of data collection was just one issue these MMM pioneers faced. The other challenge was computing power. With computers that still relied on magnetic tape, punch cards and vacuum tubes, there was nothing fast about building and maintaining an MMM model.
Even the largest early Simulmatics clients, like Nestle, Colgate-Palmolive and General Foods, would likely have updated their models infrequently. So even while consumer behavior undoubtedly shifted during the months between new models, they’d still have to base plans on outdated information.
The Rise Of ML-Powered, Automated MMM
Fortunately, marketers no longer need magnetic tape for computing, nor do they have to recruit their families to help with data collection.
Today, automated data flows make it far easier to gather data from many marketing channels and information sources. Those include CRM systems, customer data platforms, ad platforms and web analytics. Data can also be integrated from relevant external sources, like industry or weather data.
An MMM model can then estimate the likely impact on revenue or another business outcome of each of those channels, as well as those for which data isn’t readily available.
Additionally, powerful cloud computing has made the modeling process far more flexible, fast and easily updated. Changes can be implemented in minutes instead of weeks. Instead of waiting months for completed models, marketers can refresh models weekly to make decisions based on the most recent data.
Machine learning and greater computing have also made these models more accurate. By identifying gaps between spending and targeted outcomes, these trustworthy models can guide the optimization of marketing budgets across all channels—even those where data may be limited or sparse.
Another new aspect of today’s faster, more robust MMM is the ability to rapidly simulate different budget scenarios and their potential impact on business outcomes. Simulation tools offer actionable recommendations for how much spending to allocate when given a specific set of goals and constraints.
Finally, just as in the Simulmatics days, some consumers are concerned about the privacy and security of their data and how it’s used. With MMM, only channel-level data is needed, not individual-level data. So MMM is a marketing measurement tool that respects privacy policies and considerations, while still helping marketers make reliable data-driven decisions with up-to-date information and simulations.
With the addition of machine learning capabilities, MMM isn’t just making an ordinary comeback—it’s at the forefront of a marketing revolution.
Moving Into The Future With MMM
Another benefit of today’s MMM is that it no longer requires advanced statistical skills and specialized expertise. Marketing organizations can adopt MMM in various ways, including low-code predictive analytics platforms, open-source packages for use by experienced data scientists, or external consultancies.
The willingness to adopt a truly data-driven approach to marketing strategy and planning is just as important as technical readiness. At Pecan, our research has revealed that over half of marketers still feel many decisions are based on “guesswork,” despite the available data and technologies.
But if your organization is ready to move past gut intuition and toward a more confident, future-focused marketing strategy, the new-and-improved marketing mix modeling may be just the ticket for your team—no punch cards required!
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
