We are stuck
Marketing Science needs a mindset shift [Chapter 2 of the book DEEP IMPLICIT]
Sergio Zyman, Head of marketing of Coca Cola in the early 80s, was a brave and progressive manager. He not only took “data driven” seriously but also had the bravery to act on the “obvious”. With his project “New Coke” he created one of the most famous marketing lessons in history. The test results had been straight forward. The new Coke recipe was preferred by customers in double blind product testing’s. The numbers spoke a clear message. Numbers don’t lie. It turned out:
They do lie!
“Fact-based” and “Data Driven” is a big illusion.
Our rational mind is designed to solve rather simple problems. Through “test and learn” human learn to use tools. With language we can abstract and derive conclusions based on other facts. Still all this is rather simple. It concerns one or two elements at a time, not more.
How can a species that has such simple rational mind operations build rockets, fly to the moon, build computers and large language models?
The answer: Divide et impera
Latin for “divide and rule” is a roman principle that tells us that dividing complicated problems into simple pieces, simple subproblems, can solve them. Dived 164 by 4? 160 by 4 is 40 plus 1. It’s solvable by separating it into additive parts.
That’s what we do in science. Whatever your computer or even AI is doing it can be broken down to a series auf binary (digital) transistor calculation and produces amazing outcomes.
Roughly the last 200 years of humankind are governed by this scientific principle: Divide your problem at hand into simple subproblems, then solve it.
Science and engineering brought electricity and industrialization, the computer and automation of industries, the internet with the digitalization of the economy, and now AI with the automation of creative and knowledge work.
From a first look, the principle seems to hold unlimited possibilities. Some belief it is the universal principle to progress.
Yet, it is not.
It is for certain types of problems. It’s not for others.
Success of the scientific principle makes us blind to something very unpleasant:
The unknown.
We get stuck in the belief, we can understand everything by breaking it down into pieces. But this is how you can only understand the “known”.
Even worse: This is how you can understand ONLY the complicated …
but not the complex.
In 2011, retail giant J.C. Penney hired Apple’s former retail head, Ron Johnson, to modernize the aging brand. He quickly scrapped discount pricing in favor of “everyday low prices,” redesigned stores, and aimed to attract a younger, trendier audience. But within two years, sales plummeted by billions. Loyal customers felt alienated, new ones never arrived, and Johnson was fired.
What went wrong? It wasn’t one mistake — it was a complex system of interacting factors. The strategy ignored the emotional habits of existing shoppers, who relied on coupons and sales to feel smart. There was a time lag between brand repositioning and consumer perception: while internal teams saw progress, customers stayed confused. And many unknowns — like the emotional role of pricing rituals, loyalty built over decades, and internal resistance — collided unpredictably. The company had applied bold, top-down logic to what was a fragile, adaptive system.
In Marketing most problems are not just complicated, like a rocket ship. They are complex. It’s simply not enough to break it down into pieces.
Why do you think the launch success rate of space rockets is higher than 95% while the launch success rate of a new FMCG / CPG product is invertedly just around 5%?
Complex systems evolve when these three factors come together
· Multiple factors: many root causes influence outcomes at the very same time. It’s hard, costly or even impossible to keep factors constant while altering one of them.
· Time lags: When I bang my head to the wall, I learn very quickly thru biofeedback that this happened and that it can be a bloody exercise. When you stop marketing activities tomorrow, what will happened is …. nothing. You sell the same, you even have more cash in the wallet. Still the “bloody awakening” will come month or years later.
· The Unknown: Unknown factors, lags and the way they interact multiplying the magnitude of the problem. If I only knew when the marketing dollar will come back – next week, month or year? If I only knew which conditions, it takes to get it back – certain creative work, a product that resonates or adequate pricing? Or do I need to nail all of it first to succeed?
Sure enough, smart inventors and scientists tried to handle those kinds of problems during the last hundred years or so.
First, multivariate statistics evolved. Unilever implemented multiple regression for Marketing Mix Modeling in 1919.
Statistics evolved and tried to handle its flaws. So more complicated modeling like Structural Equation Modeling (SEM) and the field of causal modeling evolved.
The nineteen seventies brought more awareness about the dynamics of interconnected, time lagged systems that can produce chaos while being 100% deterministic (chaos theory and cybernetics evolved). Systemic science tries to handle it with tools – from the Vester matrix to System dynamics software.
The usefulness of all those methods was limited for a simple reason: It tried to model the world based on “the known”. Traditional causal modeling requires you to know the parts. A traditional MMM requires you to “know” channels are independent and linear (or saturated nonlinear or xyz). It finds parameters of a reality you need to know upfront.
But complex systems are not that way. They have considerable parts of “unknowns”.
And that is where now AI (or Machine Learning) comes in by simulating the capability of LEARNING from data. “Learning” is filling the unknown with insights based on experiences or data.
In essence, it is not just determining the parameters of a known formula, but finding the new previously unknown formula itself.
Complex systems require explorative methods of machine learning. (AI is a synonym term and is used to describe systems that show intelligent behavior.) The AI we all use (Large Language Models) was built by modeling associations in text data. Its value depends on the data used. The AI we need in Marketing to understand the real-world relationships on your companies and customers reality is called “Causal AI”.
Now. 25 years after I published a method in my dissertation, marketing science slowly is awakening. Suddenly explorative papers become possible under the umbrella of “empirics first” research. Even in an age of AI it may take another decade to have Causal AI dominating the marketing science landscape.
But even these methods have limitations.
They find the unknown by building on some weaker knowns. The scientific principle stays in play. Breaking down the known and infer pieces from there – of cause in a much more sophisticated manner.
I learned this in a recent project in which we wanted to learn how Bundesliga clubs can win fans. We interviewed experts and managing directors of these clubs. In parallel we run a qualitative study where we applied a different method that taps on the intuitive wisdom of people. A method I will share in this book.
As a result, we added one single item to the Bundesliga study. Running Causal AI then revealed that this item that experts could not voice, turned out to be -by far- the most important reason to win fans.
What if serious unknowns exist? What if you don’t have data for many relevant factors? What if you don’t even know what’s potentially relevant? Ask your data team. What do they measure? A lot. But do they measure the relevant? No, just the available.
They rarely measure how customers feel and think at any given moment in the customer’s journey. This most relevant information is missing because it is hard and expensive and sometimes even illegal to measure.
“You don’t know what you don’t know”
And there is where the classical scientific method breaks down. Unknown ways how causes interact, unknown how long it takes to effect, unknown what’s unknown.
Marketing science today simply sticks the head in the sand. It willingly ignores everything it does not know and pretends bogus theories are useful just because they are often enough sited.
Companies today are not well advised to sit back and consult marketing textbooks and research papers. A lot of it are half-truth at best.
Marketing is complex. This means brands need to do there own research because success laws can be very context dependent.
To cope with this reality, we need a deeper scientific understanding and a deeper more adequate practice.
I call this “Deep Science”.
Deep science uses approaches to cope with problems that are complicated (Multi-factorial and time-lagged) and have unknown parts. Deep science is enabled …
· On the qualitative side by: Objectified ways to leverage human intuition to get truthful hints where data are scarce or incomplete
· On the quantitative side by: Using Causal AI to explore truth in data.
For decades, we tried to outsmart complexity with confidence, numbers, and elegant theories. But beneath all of it lay the same quiet truth: we were afraid of what we couldn’t measure.
Deep Science gives us permission to stop hiding from the unknown—and start listening to it. And that’s where the real breakthroughs begin.
The world keeps telling us to optimize the known. But the companies that change everything are the ones that dare to explore the unknown.
Deep Science is not another tool. It’s a new way of seeing. It does not try to “squeeze” the real world in a fantasy model, it is building that model. Constantly and on the fly.
Once you accept that the real world will never obey to your models,
you start being humble and stop overconfidence
You stop validating and start creating.
You stop brain fog and start tuning in all our senses and data.
The benefits loom thru dozens of case studies already:
10x the impact of marketing.
p.s. How to apply deep science in marketing using the Deep Implicit Framework will be the topic of my upcoming book “Deep Implicit” and of the substack articles to come.