AI-Powered Insights: The Liberation From Spurious Correlations
Many of our customers are much more experienced in this area than I am. I am just summarizing here what I have learned from the pioneers in marketing for large companies.
I have noticed three areas that need to be considered if you want to implement the right methodology as an internal pioneer:
Introducing a new solution in a company is like launching an innovative product on the market. Adoption research has dealt intensively with this problem and identified eight universal reasons for the failure of innovations.
After publishing an early version of our Causal AI method in a peer-reviewed journal in 2008, I started helping companies as well as PhD and Master’s students to use the software. One of the master’s students was Daniel. He was doing his master’s degree part-time while working at a telecommunications company on the implementation of mobile payment solutions. I learned all about the eight adoption barriers from Daniel as he built a Causal AI model.
In fact, the model showed something unexpected for me. It wasn’t the benefits or the simplicity of the product that was the problem with customer acceptance. Mobile payments just weren’t “the process” and didn’t fit with the way people were used to paying for purchases. This change in familiar processes was the biggest obstacle to the product’s growth.
Since then, we at Success Drivers have been using these eight criteria to evaluate new products. The introduction of Causal AI in your company is also a question of adoption and it makes sense to optimize this task along the following eight aspects:
The importance of the eight factors varies depending on the product category. For causal AI, all eight factors are certainly of some relevance. In my experience, however, the first three factors are the most important:
Adoption drivers: Benefits
I often give talks on the topic of causal AI in marketing. Among other things, I often speak at fundraising conferences. In discussions with fundraisers, one argument has proven to be particularly effective. The decision-makers were already aware that we could help them make better decisions or improve customer loyalty.
But what really struck home was this argument: NGOs manage donations and have to use them to pay for their marketing and sales (fundraising). Every Euro that is invested inefficiently does not benefit charitable causes. It is therefore negligent not to use causal AI in fundraising, and looking at ROI rather than pure costs is a moral duty for fundraisers. With this argument, I have gotten to the heart of the existing problem. It is also important to understand what is subjectively perceived as a benefit.
In this book, I have tried to show the advantages of the Causal AI methodology. Which of them solve a problem that actually exists for you? That can change depending on the situation.
Have you just noticed that certain forecasting systems are losing performance? Then highlight the benefits of avoiding model drift.
Is discrimination by AI a hot topic? Then focus on it.
Or is it a problem that the insights of existing systems have led to actions and initiatives that show little success? Then Causal AI could offer the solution here.
Do your areas of application lie in the discovery of strategic insights, in the creation of automated individual marketing decisions such as retention measures? Or is there a need for more effective generation of marketing material through AI, enabled by causal AI?
It is crucial to find a problem area and a field of application that addresses a real problem. A real problem is a topic where a credible solution option always falls on open ears. “Oh, that’s possible?” is the question you hear when you propose such a solution.
Adoption driver: certainty
Isn’t the problem with today’s marketing that customers no longer believe anything? After all, we are bombarded with unbelievable claims every day. Many customers have learned for themselves that you can’t really believe the promises of marketing. Everything is perceived to be “inflated”, decisive factors are exaggerated or omitted.
As a result, marketing as a source of information is becoming increasingly difficult. Content marketing has evolved from this. Peer-to-peer networks are sprouting up. People believe their colleagues more.
If you doubt the benefit, the actual perceived benefit is low. You can boil it down to a formula:
Want to have = Benefit minus Effort times Certainty
This is precisely why a pilot project is usually a good idea. However, it must be designed in such a way that the benefits can actually be experienced. It should be set up with the intention of building the full version if it is successful. This presupposes that the relevant people are involved in the pilot so that the requirements of the organization are already clear in the pilot.
Adoption driver: Uniqueness
Before the streaming speaker brand SONOS launched its Move and Roam mobile models, we at Success Drivers had the pleasure of testing the product “in the market” with a survey. We determined the implicit willingness to pay (Implicit Price Intelligence) and how the product was perceived on the eight Adoption Drivers. Causal AI then showed that willingness to buy and willingness to pay increased significantly when consumers perceived the product as unique. It turned out that most target customers did not understand the specific difference to conventional Bluetooth speakers. This was the biggest obstacle to the product launch.
By placing the Wifi functionality at the center of the communication, Sonos made the launch a complete success.
I think it’s similar with the introduction of causal AI. Many data scientists simply don’t know anything about it. There are many statistical methods that purport to measure the effect of a cause. I still get asked sometimes if correlation wouldn’t be enough.
Multivariate regression is still very often used and sold today as “driver analysis”. If it is important for companies to also capture non-linearities, many market research agencies resort to Shapley value regression, although scientific studies show that this method delivers misleading results.
Recently, market research software providers have been offering an “AI Driver Analysis”. This consists of a random forest method and an impact measurement borrowed from Explainable AI. Anyone who has not dealt with the requirements for causality may well be confused by the plethora of methods. This is because they all claim to measure the influence of causes. That is the reason why I am writing this book. I want to make clear what it takes to understand what drives success (causality).
A random forest approach (or another common AI approach) does not meet the requirements that we developed in the third chapter. The more strongly the drivers correlate with each other, the more distorted and therefore unusable the results become.
When David from T-Mobile USA commissioned us back then, it was primarily his initiative and decision. Of course, he asked his boss for permission, but she trusted him and didn’t start her own evaluation. He listened to everything in two online meetings with me and discussed his challenges with me. After that, it was clear to him that he wanted to pilot the method. The results were so convincing that he had a retainer fitted.
After all these years, I’ve always wondered what kind of people want to try out Causal AI without months of internal discussions. I think the term “early adopter” sums it up best. For most people – including managers – it’s not enough to find something plausible or have proof of its effectiveness. They need social support (in B2B: the organization). References alone are not enough either. “They all have logo wallpapers,” is the response.
According to marketing literature, around 15% of people are early adopter types. These are people for whom their own professional judgment is enough to decide whether something makes sense and should be tried out.
When I presented our work with SONOS at the 2018 ESOMAR Congress, the head of market research at Microsoft – Reed Cundiff – was on the panel that selected the presentations. He immediately realized that our technology could help Microsoft.
“If the head of one of the largest market research departments in the world gives his department heads a tip, it will be done”, you might think. But in my experience, that’s not how it works. For good reason, line managers hardly interfere in the work of their teams, who then find ways to “get rid of” projects that, in their view, bring little added value.
Once again, we were lucky that another early adopter – Dr. Rajul Jain – had taken over the testing of our method. She was so convinced of the methodology that she did the internal persuasion work with great commitment.
In my view, it is therefore crucial to identify early adopters in the company. You need these people to gain a foothold in a larger company.
If you are an early adopter yourself, you would do well to find allied early adopters. This can slowly create a critical mass that encourages the sluggish masses to follow.
If your company is so big that you can’t get a good idea of all your colleagues, ChatGPT can actually help (no joke!) We tried it ourselves once. We used the LinkedIn Navigator to find 1000 people who belong to the target group. Then we searched for each person’s self-description in LinkedIn and asked ChatGPT to what extent this person is an early adapter of our service. The 50 people with the highest scores received a book from me in the mail with a cover letter.
I know from experience that such mailings are very rarely followed by replies or even project inquiries. This time it was different. We immediately received a project briefing from a large company. Our offer was convincing and we carried out our first joint project.
Persuasion is a tricky business, because it involves “imposing” an opinion on someone – even if you mean well. To exaggerate, it is essentially about the attitude of the knowledgeable person who wants to convert the ignorant. But nobody wants to be converted against their will. People sense this, no matter how sophisticated the tactics.
That’s why, in my view, your own attitude is crucial. I like to ask myself the question: “Can I bear to let the other person have their opinion, even if I don’t like it? Respect for others means giving them the freedom to choose. And even more. It is regularly helpful to remind ourselves to be humble. Because the more convinced we are ourselves, the more blind we can become to completely different points of view. Views that may later prove to be wiser than our own.
With humility and respect, we can go into the process with the intention of inspiring. This will be more fruitful in the long term than narrow-minded persuasion.
Interests
When I drive through my adopted home of Cologne, I see a Ford B-Max on the road every day. How can that be? Production of this model was discontinued in 2017 due to a lack of demand. My wife has been driving the car for 10 years and I find it super practical. The lack of a B-pillar and the sliding doors are just great. As I am a self-confessed (lonely) fan of this car, I notice this model every day.
This is known as the “cocktail party effect”. Our subconscious filters the available information and makes us aware of those that it considers relevant.
This process is constantly active, even when we are listening to someone giving a lecture. Studies show that only around 5% of what is said is remembered. These are mainly the arguments and information that our subconscious plays back to us.
What information is this? It is information that appears to be “relevant” to us in some way. In other words: Perception is interest-driven. If I am a CFO who wants to have a profit at the end of the year, then I am interested in everything measurable that contributes to this profit.
If I am a creative brand manager who “knows” that creative performance is difficult to measure, then I am interested in arguments that support my point of view.
So if I want to inspire someone to think outside the box, I should show how my thesis serves their interests. The reference to their interests makes the argument relevant and is then examined.
An early adopter is interested in an objectively better solution. A CFO is interested in a solution that delivers demonstrable and risk-free bottom-line results. A team leader who is to be promoted to another team in two years is interested in results that show visible success after one year.
It is actually trivial, but it is work to put yourself in people’s shoes and find out where their interests lie. It may be trivial, but it contradicts our world view that we should all be professional and objective and therefore all serve the same corporate goal. Putting this into perspective is an important step in being able to inspire colleagues.
Case studies
I have spent 20 years – including my doctorate – in the education system. The didactic method of science is deductive. This means that a theory or thesis is first discussed abstractly and then concrete examples are derived from it. The way people learn is exactly the opposite. Learning is inductive. It starts with examples. Only when we have heard at least one are we ready to absorb the general learning from it.
As far as it made sense to me (and I was attentive enough), I have done exactly that in this book. First the example, then the lesson. I am also deeply convinced from my own professional life that the presentation of theses is not understood without prior examples. Because “understanding” means relating abstract knowledge to other concrete knowledge. This is how it becomes conceivable.
I myself am a victim of my educational path and try to get a little better every day. Do the same! Your audience will thank you for it.
Storytelling
The story is the refinement of a case study. Mankind has always been preoccupied with the question of how events can be presented in such a way that the audience listens attentively to a story. The basic structure of an optimal story has not changed since the ancient Greeks.
The art of storytelling was cultivated by the ancient Greeks and continues to be practiced professionally in the film industry. In recent years, management writers have taken up this art and transferred it to the world of management.
There are great books on this, which I cannot summarize perfectly here. Essentially, however, the point is that case studies need some central stylistic devices and a dramaturgy.
As a stylistic device, we need a hero who is as “likable” as possible, a person with whom the audience can identify. You also need a villain who has to be defeated.
The dramaturgy classically begins in a state of equilibrium, which is immediately followed by a catastrophic event. This is the entrepreneurial challenge. The resulting suffering is then attempted to be overcome with the help of a solution. But the first attempt fails. The tension mounts. This “attempt to solve” and failure can be repeated several times in a dramaturgy. Finally, there is a finale in which the hero confronts the villain (who defends the challenge) with a new solution and finally wins (eliminates the challenge).
Anyone who succeeds in packing case studies into such stories captivates the audience, entertains them and wins their hearts.
Evidence and safety
At the beginning of my career, I was Marketing & Sales Director at the global market leader for industrial packaging. In this role, I was also responsible for innovation. We developed new industrial drums and transport solutions. I even have registered a patent for a transport solution. It was not easy to find early adopters on the customer side. That’s why I concentrated on clearly communicating and demonstrating the added value of the solutions.
In this process, it was only over the years that I realized that something else was much more important in the industry than offering better products and services. Customers often put dangerous or expensive products in the packaging. Every leak and every accident was a nightmare for everyone involved. Moreover, packaging was only a minor cost factor. If the production process was disrupted by packaging or logistics, the costs were many times higher. Safety and reliability were the be-all and end-all. That was the “language” of the industry. Continuity, brand, reliability and quality were the attributes of the winners.
The limbic center of the brain plays an important role in what we call the subconscious. It turns out that people react with different affinities to certain limbic motivators. Brain research has identified three dimensions that drive all people: stimulation, dominance and balance.
Early adopters are particularly attracted to the new, to innovation. This releases dopamine. The topic is exciting and thrilling. Stimulation is the preferred motivator. This is why early adopters are the first target group for innovations such as Causal AI.
Leaders are attracted to the best and to performance. They release testosterone and seek success and prestige. Dominance is the preferred motivator. People with strong leadership skills are usually managers.
Protectors and caregivers are attracted to security, harmony and balance. Oxytocin is released and a feeling of home, security and love is conveyed. Balance is the preferred motivator.
Everyone has their inclinations and these also lead to a career choice. In order to successfully promote innovation in a company, all three aspects must be addressed. The aspect of the new is inevitably linked to innovation.
To inspire the traditional manager, more needs to be added: evidence, i.e. proof of performance. What are the factual advantages of the solution? How can they be translated into performance indicators that are important for the decision-maker? These are the questions that need to be answered in order to inspire performance and results-oriented decision-makers.
We need results from comparative studies that clearly demonstrate the superiority of causal AI. Depending on the use case – discovering vs. deciding vs. generating – the comparative studies must be structured differently.
The advantages of causal AI “model drift” and “discrimination” both address the issue of risk avoidance and safety maintenance. Depending on the contact person, they are therefore only motivating for a specific target group.
Metaphor
If you have read this book carefully, you will have noticed many metaphors. Wherever possible, I have tried to organize the chapters according to the pattern “Story > Learning > Metaphor”.
I was inspired by Oliver Raskin. He is currently Head of Insights at MIRO. Oliver told me about his “superpower” at a dinner together when we were talking about how Insights can gain a foothold in the company.
“Whenever possible, I summarize everything in a metaphor,” he said and told me how this often works wonders.
We are often dealing with very abstract things. “Causal AI” – what could be more abstract? It makes sense to use a simile to give the whole thing an image. Because what does understanding mean? Understanding is the process of relating something new to known units of knowledge. The more concretely these units of knowledge are anchored, the more comprehensible the statement becomes. Images are very concrete units of knowledge.
It’s like trying to build a house out of bricks. If the stones are soft, made of cotton candy or soft lumps of slime, the house won’t hold. Solid stones made of concrete are like concrete pictures. They create a solid house, an understandable statement.
A picture is worth a thousand words.
Role model
In 2022, we had started to market our pricing method “Implicit Price Intelligence” as a DIY tool. Then I had an idea. We invited our target group to a webinar where each participant had to take part in a 5-minute survey as a prerequisite for access. Our tool itself was used in this survey and the target group’s willingness to pay for the tool was measured. I then presented the procedure and the results in the webinar – along with the agreed price change. The price was increased by 20 percent for frequent buyers, but reduced by 30 percent for newcomers.
“Drink your own wine”, as the saying goes.
It makes intuitive sense. But is that really necessary?
Convincing others is a difficult game. You can deceive and manipulate. Or you can adapt to your counterpart for nobler motives and present the information to them in the way they can best absorb it.
However, the decisive factor is the intention. Whenever your own interests do not fully coincide with those of your counterpart, persuasion can turn into manipulation. However, people usually smell manipulation attempts “three miles upwind”.
If we manage to free ourselves from the need to convince, if we allow our counterpart to form their own opinion, we unconsciously send a strong message.
This message is reinforced if you set a good example yourself, if you have the opportunity to pilot the proposal on a small scale and take risks yourself.
Actions are more powerful than words.
If you want to inspire rather than persuade, you can be sure that you will achieve more. By setting a good example, you make your attitude visible.