The Renaissance of Marketing Mix Modelling

But why is the technique becoming more popular? Is it truly a renaissance, or is it something else? To truly understand this renaissance (literally “rebirth” in French), it’s useful to understand the origins.

What is MMM?

Marketing Mix Modelling (sometimes known as Econometrics) aims to quantify the impact of various marketing activities on sales or other targeted outcomes. Its aims include optimising budget allocation by determining which marketing channels yield the best ROI, measuring the overall effectiveness of marketing strategies, forecasting potential impacts of future campaigns, and accounting for external factors like economic trends or competitor actions. Through these insights, businesses can make informed decisions about their marketing efforts.

This type of modelling goes beyond just evidencing marketing performance; by including non-marketing sales drivers such as the weather and macroeconomic factors, it can reframe all factors which drive demand in a business, thus making it an impactful tool in driving business-wide investment decisions.

Where did MMM come from?

Marketing Mix Modelling (MMM) originated in the mid-20th century but gained significant traction in the 1980s and 1990s. It was a response to the need for a structured and analytical approach to marketing strategy. Its roots can be traced back to the work of Neil Borden, who first introduced the concept of the "marketing mix" in the 1940s and 1950s. Borden's ideas were further developed by E. Jerome McCarthy, who, in 1960, introduced the well-known 4Ps framework, encompassing Product, Price, Place, and Promotion. Over time, marketing practitioners and academics refined these concepts, leading to the birth of Marketing Mix Modelling as a formal methodology in the 1970s through papers like BRANDAID from John Little.

Originally, Marketing Mix Modelling projects were typically expensive, long-term studies requiring a team with a rare skillset, integration of data sources which were often fragmented and the use of scarce computational power – all of which, back in the 20^th century, would be prohibitively expensive for anyone other than the largest of corporations with significant budget to invest in this type of approach.

The MMM Dark Ages

There have been significant developments in technology and consumer purchasing habits - and marketers have always adapted to keep up. Back in 1978, Gary Theurk, a computer salesman, sent the very first “spam” email to early users of the ARPANET advertising his product. By the 2010s, online advertising had exploded, featuring channels like PPC, display ads, social media targeting, affiliate links, email, and more – all of which can be evaluated and monitored by tracking the behaviour of consumers across the internet.

This ability to track individual customers should not be understated. It allowed marketers to look past aggregated and then-expensive measurement techniques like MMM and instead to evaluating single-person journeys using Multi-Touch Attribution techniques. Whilst MTA are complex mathematical models, they are highly automatable and scalable.  It allowed marketers to put rules-based approaches like first/last click or linear sharing (which were originally popular, but not reflective of reality) to one side and factor in how a potential consumer progresses along their purchase journey, and when/how they were influenced.

Although the classic MMM methodology was still used by larger, in-store retailers and FMCG suppliers (particularly those with bricks and mortar stores where purchases cannot be tracked at an individual level), MTA allowed any firm with an online footprint to make faster, tactical decisions to optimise marketing campaigns and provided granular evaluation of advertising performance at the campaign or keyword level. MTA models were typically out-of-the-box, automated, and could churn out reports on a daily, weekly, or monthly basis depending on the use of in-platform or independent providers.

Although MTAs don’t provide complete information (say, about the macroeconomic environment or competitor activity) and are rarely able to include ATL channels, it was felt that digital attribution gave concrete enough reporting metrics to marketing managers desperate to justify their budgets to CFOs… until it didn’t. Changes to cookie regulations, an increase in “walled garden” advertisers, and an increased consumer expectation of privacy provides a major blocker to most MTA solutions.

Faced with being deprived of valuable effectiveness information and the inability to link performance from the wide diversification of marketing channels to real revenue metrics, marketers began to look back to older techniques like MMM to still be able to report holistically on marketing activity and optimise their marketing mix.

The Renaissance

Still, just as the renaissance of the 14-16^th centuries consisted of great leaps forward in art, literature, and science, our modern MMM renaissance would hardly be a renaissance at all if the techniques and mathematics underpinning it remained static.  

These modelling improvements have enabled better model fitting, improved transformations of input variables, better/easier deployment, all facilitated by more data than ever before. What was once the domain of large retailers with significant above-the-line presence is now, through a combination of increased computational power, democratisation of data, and readily available modelling techniques, more accessible than ever .

For example, technology firms Google, Meta and Uber have all released open-source modelling libraries for MMM, with PyMC also contributing for those of us with a Bayesian flair. In addition, many advertising platforms work with companies to ensure that data can be quickly exported for use in MMMs (enabling so-called “real-time MMMs”). At the most introductory level, this would enable a single analyst to run a MMM with nothing more than a few CSVs and their company laptop (though this does carry some danger and we’d advise putting a little more thought into it than this!).

Now, are these innovations the next Printing Press? Clearly not. Is this a renaissance in the sense of long-forgotten practices coming back into the culture? Not really – for many large companies, MMMs haven’t “gone” anywhere and some thought leaders like Les Binet have never stopped championing MMM over techniques like MTA. What we are seeing are smaller companies “waking up” to these old, yet new, approaches and a wider pool of marketers willing to incorporate MMM as a valuable tool in their arsenal. And even better, it’s now no longer as expensive as it has previously been perceived.

The measurement solution

Just as Italian city-states greatly prospered during the renaissance, there were other areas of Europe that were passed by: the Byzantine Empire fell, and Scandinavia/Central/Eastern Europe didn’t particularly benefit at the same time as the west of the continent (though experienced their own renaissance much later).

It’s probably an overstatement to say that companies will either be part of this renaissance, join in later, or perish completely. However, the truth is, marketing mix models will be part of the measurement solution (to a greater or lesser degree) for many companies with a reasonable marketing spend, and it’s simply a matter of “if” not “when”. A:B testing is unsustainable to run all the time (though is incredibly useful for model calibration and answering key questions) and MTA faces serious headwinds (though, for now, is still very useful for primarily digital advertisers looking to optimise tactically).

Want to understand Econometrics in a bit more detail? Check out Econometrics Explained – the handy introductory guide from the IPA to using Marketing Mix Modelling effectively.

Looking to get started? Check out this article from WARC (“Marketing mix modelling: How to get started and ensure success”) or sign up to our upcoming webinar on November 9th, to hear from me and other industry experts.