AI-Powered Insights: The Liberation From Spurious Correlations

April, 21, 2024

Chapter 6


Why, how and when to start

Dr. Frank Buckler

Founder, Success Drivers

As a research assistant at the University of Hanover, I have read and evaluated countless diploma, master's and bachelor's theses. I always read the last chapter first. That's why I'm summarizing the most im-portant findings here first.

The most important lessons learned can be summa-rized as follows:

Marketing expertise is the decisive design element of data analysis for marketing. The equation of correlation and causality is the cardinal error that still unites management and data science today. The dominant hypothesis-driven approach to data analysis has its limits. At best, it produces racing cars that are designed to drive over open, hilly terrain and too often fail.

What is AI? Statistical modeling finds the parameters of a fixed, predetermined formula. Artificial intelligence finds a new, optimized formula, not just its parameters. However, AI suffers from model drift, discrimination and the risk of performing poorly in live operation. Explainable AI does not solve the problem and offers a dangerous false transparency. The solution is called Causal AI. This is for AI like the filter system of a distillery. Ultimately, it is an absolute must.

And this is how you proceed with Causal AI: Start with a “blank sheet of paper” and write down what could influence your target variables. Then obtain the data for this. In addition to internal and external data sources, also consider market research. It offers unique opportunities to understand the inner life of customers.

Model the data with an AI algorithm that ensures in the learning process that predictions are only based on causal causes. Model a causal network, not a pure input-output relationship. Open the AI black box with suitable simulation algorithms. Algorithmically check the causal directions and search for confounders. Optimize your model and recalculate until it is meaningful and useful enough. Finally, standardize everything in one process – from the data, the preparation, the method to the preparation of the results.

Why Causal AI?

AI is present in all marketing processes, be it in insights discovery, operational and strategic decision-making or the generation of marketing content. All of these algorithms suffer from model drift, discrimination and limited reliability of predictions. The risk varies depending on the use case and context. In most cases, Causal AI is a step forward compared to the status without AI. It opens up a new level of development for more performance and less risk.

How to get started?

Find internal allies who are early adopters. They also think about where there is a real problem in the company that Causal AI can solve. Too often we tend to walk around with a belly laugh. “Causal AI can do everything.” That confuses and confused people don’t make decisions.

A pilot project often makes sense because it carries a limited risk. However, in order to exploit the upside potential, the pilot should be set up in such a way that it serves as a real test case for the rollout and provides clear evidence. If necessary, set a good example and show how you think others should act. Be a role model.

But would you rather wait?

Are you convinced that the cost of the Causal AI Pilot is already less than the benefit? What is the argument for waiting?

Ask yourself how much risk (model drift, discrimination) you can avoid by not waiting. Also ask yourself how big the long-term disadvantage is if you are too late on the learning curve.

If you are an early adopter, these questions are not relevant for you. AI is the value driver of the economy of our time. Causal AI takes AI to new dimensions. The future of AI is “causal”. 

Causal AI is also and especially an opportunity for marketers not to lose their position in the decision-making process to data science and be unjustly substituted by AI.

Data science needs the expertise of marketers.

Causal AI needs you – as a marketeer and as a human!

Success Drivers GmbH
Johann-Heinrich-Platz 4
50935 Cologne

Success Drivers GmbH holds the copyright for any material, included translations, published at