March 7, 2024
Chapter 4
Applications Of Causal AI In Marketing
Dr. Frank Buckler
Founder, Success Drivers
AI-Powered Insights: The Liberation From Spurious Correlations
AI-Powered Insights: The Liberation From Spurious Correlations
I will touch on a few fields of application as examples. You may get the impression that Causal AI is really universally applicable. You may think: “Yes, of course, that just makes more sense”. This realization alone will inspire you.
Just as the benefits of mantras such as “lying doesn’t pay” or “humanity wins” are not obvious in the short term, life experience shows that they work. Just as human character with integrity “pays off”, more logically consistent approaches are also more successful in the long term.
So what can you do with Causal AI? I distinguish between three activities:
It is possible to pursue several activities with one model. However, differentiating between goals leads to greater clarity of thought.
Artificial intelligence has been used from the outset to automate decisions. Which product should be offered to the customer now? Which customer should be sent a mailing to discourage them from canceling? Will this customer be able to repay their loan?
For these questions, it is not necessarily problematic that AI is a black box. After all, the most important thing is to make the right decision, not to understand how it arrives at it.
But if we want to optimize the marketing mix, if we want to decide which initiatives we can use to improve the customer experience, then we need to explain what drives success. If we want to understand how we can improve our products so that they sell like hot cakes, if we want to fundamentally uncover the hidden reasons and relationships that lead to customers buying or not buying, then we should ask “why”.
Today, this “why” question is answered primarily through qualitative market research and the qualitative exchange between experts. Statistical modeling is also used occasionally, but only in borderline areas due to its limitations. Each method has its value and its place. But causal AI can now cover an area that neither qualitative research nor statistical modeling could serve well.
The situation of a gambling provider was tense. Turnover had been falling for years and the managing director rightly wondered whether the millions spent on advertising had been well invested.
The advertising channels in the lottery business include traditional media such as print, posters and radio, as well as shopping radio in supermarkets and advertising at the point of sale and on websites. The media planners determine the distribution of advertising according to general principles.
Radio advertising, for example, has the fastest impact. This is why it is used massively for short-term topics such as jackpots in order to increase sales. However, these beliefs do not say exactly how much advertising pressure is needed.
Our team differentiated the use of media according to the content of the advertising. Was a jackpot advertised, was it brand advertising or a special promotion? Other important influencing factors were the size of the jackpot, the current media coverage, the specific days of the week and the time of the month.
It is always important to include all influencing variables in a model, regardless of whether they are controllable or not. This is because false results (Keyword: confounders) can only be avoided if the model is as complete as possible. If, for example, radio advertising is mainly placed on weekdays, but sales during the week are generally lower than at weekends, a negative effect would be wrongly attributed to radio. For this reason, a careful survey of possible success variables plays a central role.
The result of the analysis was surprising. Non-classical media were many times more effective than classical media. Many media channels interacted strongly with each other, so that a combined use was recommended. It was also surprising that short-term media such as radio did not have a short-term effect, but had to be used for as long as possible. However, this only seems to contradict the credo. The analysis shows that radio only creates cognitive awareness. To buy a lottery ticket, the customer must first go to the sales outlet. Hardly anyone does this just to buy a lottery ticket. As a rule, buying a lottery ticket is an impulse purchase that is merely facilitated by the awareness that has been built up beforehand. It is therefore the movement habits of the target customers that determine how quickly the advertising effect occurs and not the short-term effect of the radio.
As a result, it was shown that advertising for these gambling products is so profitable that a reallocation of the budget leads to considerable increases in efficiency. The optimization of the media plan led to continuous growth again in the following years.
In addition, the causal analysis approach provided further strategic insights. The size of the jackpot was seen as a central lever for promoting the lottery business, as the correlation between jackpot size and sales was clear. However, communication focused on jackpots leads (causally) to a drop in sales after the jackpot peak. Customers learn that it is not worth buying a ticket without a high jackpot. Thus, the communicative focus on the jackpot leads to the destruction of what lotto is all about – a beloved weekly habit.
So what is the difference to classic marketing mix modeling – a discipline that has been practiced for over 100 years (Unilever built its first MMM model in 1919)?
If there is one area where statistical modeling is mature, it is MMM. A lot of empirical knowledge and intuition is used to bend the models so that they “somehow” fit.
An MMM with causal AI naturally offers these advantages in particular:
What makes customers happy? What annoys them? How can you retain them? These are questions that an entire industry is now dealing with.
Customer satisfaction surveys have been around for a long time. However, the NPS concept has introduced a very simple methodology. Its advantage: all you need to do is ask a question at each touchpoint and, if necessary, receive an open answer. Today, CX software providers such as Qualtrics, Medallia and InMoment support the conception, implementation and evaluation. However, success is sparse and CX experts argue about why this is the case.
In my experience, one of the main reasons for this is that customer feedback is analyzed too superficially, resulting in crucial mistakes being made.
The streaming speaker brand SONOS decided to have its touchpoint surveys analyzed with Causal AI. The NPS had been declining for some time. This was explained by the fact that with increasing market penetration, other, less enthusiastic customer types were changing the customer mix. However, they were not sure.
Using Natural Language Processing, we analyzed the open responses from customers and assigned each topic that they mentioned to one of 80 categories. Tens of thousands of responses can thus be categorized with maximum precision. Today, the accuracy is even higher than when categorized by a human being. This is because humans get tired and tend to deviate from their own definitions depending on their mood.
This type of AI application in the CX area is now standard, even if many providers still use very rough, unspecific text AI models.
A look at the frequencies paints a seemingly clear picture. Almost every second customer justifies their rating with the good sound of the speakers. Unsurprisingly, there is a consensus within the company that sound quality is the key factor for customer satisfaction and customer loyalty.
Many other companies I have met are subject to the same misconception. Restaurant chains think “tastes good” is crucial, washing machine manufacturers think “washes well” and insurance companies think “has good service” are the decisive factors.
Using the categorization data, we created a causal AI model and the result was that sound quality was “nothing more” than a hygiene factor. What really engages customers is the experience of the device working properly. This could be disrupted for various technical reasons. However, this experience of smooth operation was not mentioned so often. No wonder, as there was potential for improvement.
The following graphic provides an insight into the Key Driver Matrix, which shows the frequency of naming on the vertical axis and the significance on the horizontal axis (positive on the right, negative on the left).
The tool’s impact simulator can be seen on the right. Improving the software architecture for smooth operation promises an increase in NPS of 4 points. Based on the reference values determined by the modeling, these 4 points can now be converted into additional sales.
The analysis provides many other valuable insights that would have remained hidden by conventional means. For example:
The topic of customer experience is a useful introduction to the topic of Causal AI. As a rule, companies are sitting on a lot of data that just needs to be analyzed in a more meaningful way. In my books “The CX Insights Manifesto” and “CX Insights Playbook”, I describe in detail how this can be achieved.
One of the advantages of using causal AI in the area of customer experience is that
Which product features are important to customers? What should be considered when developing new products? Which product features should product communication focus on? These are the classic questions that arise in the product innovation process. Interestingly, they also crop up again and again during the life cycle of the finished product, because the consumer changes and the competition changes with it.
For such questions, we at Success Drivers have created the Supra.tools platform, in which modern standardized solutions for price and product optimization as well as brand and touchpoint optimization are available. All of these solutions use causal AI. In addition, they use an innovative market research method developed in the field of neuroscience. The “Implicit Association Test” measures the unconscious opinions of consumers in a very simple way using a reaction time-based query. It is precisely this information that is decisive for purchasing decisions.
The price and product optimization of the APPLE VISION PRO was one of the first application examples that we carried out a year before the market launch. All that was needed to create the study was a product image and a description that briefly described all the relevant features in the brand’s language. The tool suggests a price range and product features on an LLM basis. The latter were then verified by experts.
Subsequently, 250 computer users in the USA were surveyed, half of whom were so-called early adopters. These are people who are generally the first to buy innovative products without waiting for the product experiences of others. The willingness to pay was initially measured implicitly and the extent to which the product can actually fulfill the function of the twelve features was also implicitly queried. Features included “long battery life”, “ultra high display resolution” and “easy intuitive gesture user interface”.
The result was a profit-maximizing price of USD 1,999 for all consumers and USD 3,499 for early adopters. The price-profit function results from multiplying the calculated price-sales function by the price, minus the trade margin and the estimated unit costs.
Causal AI was used to determine the leverage effect of features on the willingness to buy (which the price test measures). Features that have a high leverage, but are not perceived as given or even doubted, should be improved communicatively or technically.
The following illustration shows the dashboard of the tool. The revolutionary gesture-based user guidance and the 3D applications in particular had a huge impact on the willingness to buy. At the same time, however, there was clear skepticism as to whether the user guidance would really work so intuitively.
At the launch, Apple communicated precisely these aspects and showed in videos how the user guidance works and familiar people in their usage situation.
Since then, the methodology has been successfully transferred to other markets in many cases. For example, a major shoe brand has recognized that despite all the interesting features that product development has given its shoe variants, the main reason to buy a shoe is a completely different one: the design must match the customer’s own style. The customer simply has to like the shoe. That sounds banal. But sometimes it is precisely this banal evidence that experts need in order to see the wood for the trees.
In traditional market research projects, conjoint measurement or MaxDiff are primarily used for such questions. The use of Causal AI combined with neuroscientific measurement methods offers the following advantages:
An alternative is the MaxDiff method. It forces the respondent to choose between properties. It is an intelligent form of direct questioning to find out which attributes are important. As it asks directly about importance, this method is rationally biased and blind to unconscious association processes that are crucial for purchasing decisions. The new method (Supra Product Optimizer) takes both approaches into account – implicit and explicit.
Ultimately, the most important thing is the validity of the method. Is what the method determines correct? This is exactly what Causal AI is designed for – the causal attribution, the justification of what caused the purchase.
What makes good advertising? What should advertisers pay attention to so that advertising works? What guidelines and rules of thumb help to avoid the risk of an advertising flop?
Questions that have long been left to qualitative research. However, advertising test procedures have always shown that there is an enormous discrepancy. Most commercials still today are only moderately effective, only a few are enormously effective. However, the unknown formula that guarantees advertising with a high ROI is still controversial today.
In 2016, insurer Metlife asked us to take a deeper look into its advertising test data to better understand what makes advertising successful. In 2017, we then developed the method further for this purpose and tested it extensively in six sectors.
Subsequently, I was invited to speak at several conferences. While my presentations to market researchers were received with (restrained) interest, the feedback from the “creative” audience was rather poor.
I still remember the final presentation at the Shoppers Brain Conference in Amsterdam. The presentations were evaluated afterwards by the participants using an app and as I was the last one, I was hoping for particularly good feedback. After all, in my presentation I showed a clear method of how advertising can be optimized holistically and systematically for the first time.
The result was a “kick in the butt” for my ego. It was the worst feedback I’ve ever received. The slightly agitated questions from the audience after the presentation should have given me pause for thought.
If I had told on stage that there will soon be an AI that could write great texts and produce photos and videos without being distinguishable from the real thing, the audience would have laughed out loud.
It is the self-image of creative people that they cannot be replaced by mechanics. Interestingly, I even agree with this in essence. But you have to understand what is really “creative” in the creative process and what is just the application of beliefs, smoke and mirrors and outdated pseudo-knowledge.
The creative spark and holistic inspiration will define us as humans for a long time to come. I believe that people who know how to use AI can be more creative and more effective – not less.
But back to the process, which uses Causal AI to distil what makes advertising successful and which specific guidelines can be used to significantly increase ROI. The methodology, which we have named Creative.AI, consists of three components:
It turns out that causal AI can explain the willingness to buy about three times better than statistical modeling. Both the way in which we test advertising and the way in which we profile it could still be improved today. What is crucial about the approach, however, is that it does not stop at perception as before, but takes up the actual characteristics of advertising.
Most exciting are the findings that the methodology has brought to light in many industry studies. Generally valid, recurring and industry-specific findings emerged:
It is universal that advertising works by giving pleasure in some way. Advertising that triggers negative emotions such as anger, disgust or contempt is usually a waste of money. Such emotions can easily be triggered unintentionally. For example, when a meat eater sees someone biting into a vegetarian sausage (which back then was not as meat-like as it is today), many of the meat eaters watching feel “disgust”. This reaction prevents any positive advertising effect.
Many advertising techniques are also universal in their effect. Here are a few examples: The voiceover technique confuses the viewer. If the consumer is supposed to learn something specific in the commercial, it is best to show a speaker who speaks the message in clear words into the camera. The most effective technique to make the viewer happy is the “looser trick”. This is the same trick that Laurel & Hardy or Tom & Jerry have been using for 100 years. There is one person in the commercial who is laughed at because something happens to them or they are clumsy.
However, the emotional messages of the commercial are industry-specific. For example, spirits are usually advertised in a way that emphasizes their quality or the fun in the company of others. However, the promise of indulgence is the most effective theme spirit brands needs to play. For investment products, the message “We are like a friend at your side” is the most effective.
From this very brief description it is already clear that this is by no means a blueprint for an advertisement. They are guidelines that a creative person can fill in with their work. It is more of a lighthouse that ensures that you navigate safely in the right direction.
Our research approach based on causal AI has never gained wider acceptance. When I met Jon Puleston again last year, who is Director of Innovation at Kantar, I realized why that is. Kantar itself runs an advertising testing methodology called LINK and has a large department of data scientists. They have now developed a methodology they call LINK.AI. They use deep learning AI systems to automatically categorize advertising into several hundred properties. An AI model is then trained based on a huge database of past advertising tests. With this AI model, Kantar can now predict the results of advertising test surveys quite well, making the advertising test redundant.
The product sells “like hot cakes”. It sells so well that Kantar has made advertising tests mandatory for all those who want to use the prediction in order to obtain more learning data.
When I heard about this, I realized where the need really lies. Decision-makers want a forecast. They want to know whether an ad is good or not. Kantar delivers that. What it doesn’t do is describe what would increase success instead. This is not (yet) a good sell. Because these statements are then in conflict with what the “creatives” say. But who knows how things will develop? We will come back to this in context of Gen AI.
After several successful Causal AI projects at Deutsche Telekom, I received an email from T-Mobile USA in the summer of 2013. David, the Insights Director at the time, told me that the new brand and product strategy was working wonders and spurring growth. What was giving them a headache, however, was that T-Mobile didn’t know what exactly was attracting customers. Was it the then innovative flat rate? Was it the fact that T-Mobile had completely removed the contract commitment? Or was it the fact that it came with an iPhone for USD 0?
We had the opportunity to take a closer look at the extensive brand tracking data to better understand what drives customer behavior. On the day I presented our findings, I was a little unsure whether David would be satisfied with the depth of the insights. I didn’t realize that the results would have such a big impact on the future of the company and the group as a whole. Just before Christmas, I received this thank you email from David:
What had happened? Our analysis had shown that none of the presumed success drivers were directly responsible for the growth. Instead, it turned out that the company’s positioning as the “Robin Hood” of the industry was the key lever for success. The innovations “no contract commitment”, “flat rate”, “good and affordable devices” were the perfect arguments to make this positioning credible. They had an indirect effect and reinforced the positioning. Each of these elements could be copied. The Robin Hood status was not. So the brand decided to develop a continuous stream of unusual features and launch new ones every quarter. One example was the “Free Global Roaming” feature. The approach became known as “Uncarrier Moves”.
The strategy worked. T-Mobile grew year on year and took over its competitor Sprint seven years later. Today, the brand has risen from a small, loss-making provider with inferior mobile networks to become the market-leading, highly profitable mobile communications company in the USA.
In 2022, I met Tim Höttges, CEO of Deutsche Telekom, at his keynote speech (picture below). What he showed there put the icing on the cake. The parent company has risen from being the seventh-largest telecommunications group in the world to number one largely thanks to the development of T-Mobile USA.
It’s amazing what the results of a Causal AI analysis can do. Today I think: We should have invested our fee in T-Mobile shares. But well, what could be better than having your recommendations implemented?
This is a prime example of how marketing strategies should be well-founded. I had already shown similar examples in the previous chapter, e.g. from Kindernothilfe. These examples show that it is worthwhile for every sector to take another look at its own understanding of the market.
For the German beer market, for example, we have found that it is not the quality or the special taste, but the promise of refreshment that makes the basic benefit of a pilsner. We also found out that the purchase of body care products (shower gel etc.) is most sustainably influenced by the fragrance experience. Both are plausible findings, but they are not mainstream.
There are probably thousands of methods and variations to understand what motivates the buyer of a product category. Many of them complement each other, some are more useful than others. What most of them have in common is that they deliver plausible results. Therein lies the danger. Because people tend to use plausibility as an indicator of truth. However, plausibility only expresses whether something is congruent with existing prior knowledge. That doesn’t really help if you want to learn something.
There are many methods that claim to uncover the “why” behind customer behavior. Very few of them have the scientific ideal of causality in mind. That is why there is no perceived lack of “why” explanations. Our brain generates them automatically anyway. It’s like with sports. The fans always know why things aren’t going well. That’s just how our brain works.
However, I hope that these examples give some indication that it is worth approaching the question of “why” in a exploratory but quantitative way. The benefits of Causal AI for marketing strategy can be summarized as follows:
The greatest leverage lies in explaining the reasons behind success and failure. Causal AI delivers this “why” in a quality that no other methodology can. Unfortunately, many decision-makers do not realize that plausibility is a weak predictor of truth. This is why there is no urgent demand for answers to the “why” in many companies. Because “answers” are a dime a dozen. Whether they are valid is another matter. Decision-makers who can separate the wheat from the chaff here have a strategic advantage, as the examples have hopefully illustrated.
The situation is different when it comes to operational decisions. Here, the success of the decision can often be measured in real time. When AI came back into fashion after another “AI winter” in the 2010s, it was almost always about predictive analytics. The aim is to use AI to make better and more flexible operational decisions.
What are operational decisions? In marketing, we can distinguish between two types of decisions: Firstly, decisions that affect communication, product, packaging and distribution. These decisions affect ALL customers. On the other hand, there are customer-specific decisions that can be found in direct marketing, sales, customer service and customer management, where a separate decision can be made for each customer.
Which advertising idea should be implemented? Which new product concept should be introduced? Which packaging design is best received? Which price maximizes profit?
Decisions of this kind are primarily evaluated using test surveys. AI can be used to predict the outcome of the survey. By applying causal AI, such a model can be made more stable so that the predictions are actually more valid and more likely to occur. Some examples have already been mentioned.
For example, Kantar’s “Link” model, which predicts the results of an advertising test. With the help of Causal AI, such a model could achieve better forecasting performance on live data in the future. This is because Causal AI models suffer less from model drift, as they not only interpret the characteristics of the advertising causally better, but also include or control confounders.
One example is the New Product Forecasting Model that we built for Mintel. It enables an initial screening of new products before the actual market launch. With a success rate of 81 percent, the model is very useful.
Deep learning systems (providers such as Aimpower or Neurons) that use image information to predict the results of real eye tracking studies have already been established for testing ads and videos in particular. It is already possible to predict with an accuracy of 90% where consumers will look at images and for how long. Of course, there are borderline areas where this does not yet work so well.
For example, the AI knows that people like to look at faces, but it has not learned that dog owners like to look dogs in the eye just as much. Another example is the “black square” painting. People are looking at the edges, while the AI assumes that the gaze is in the center. These examples show that this AI lacks the contextual information that a causal AI would ideally take into account. Further developments through causal AI can be expected here in the future.
The procedure is described easily. Lets take churn prediction again. In the first step, the customer data is prepared in such a way that the customer’s current characteristics and behavioral data from one year ago are used as predictor variables. The target variable examined is whether the customer canceled the contract in the following period. The model should learn to predict these churns. The AI can access thousands of customer data records to create a prediction model.
The trained AI model is then used to carry out the churn scoring of all customers itself. Only the same feature variables are updated for this purpose. The model was trained with characteristics dating back one year. The AI model uses this to calculate a pseudo churn probability – a value between 0 and 1. For various reasons, this value is not a “real” probability.
As the value does not represent a “real” probability, it is still necessary to define the threshold value above which an action is to be triggered that attempts to prevent churn.
As mentioned in the first chapter, every churn prediction model makes two types of errors: false alarms or missed opportunities. In the case of a false alarm, churn is predicted even though it will not occur. The costs for the campaign were therefore spent in vain. In the case of a missed opportunity, the model fails to predict churn. The customer is lost.
If we set the threshold value to just 0.1, we will consider most customers to be churners. This will minimize the error of the missed opportunity, but the error of false alarms will be very large.
If we set the threshold value to 0.9, we only have a few false alarms. But the missed opportunities are increasing.
The optimum threshold value can only be determined if I know how much an action costs, how likely it is to prevent a cancellation and how much customer value is lost if a customer cancels. By simulating the ROI for all possible threshold values, the ROI can be optimized.
Campaign management
Back to the example of the telecommunications company. The question arose as to how effective the cuddle calls actually were.
The properties of the model can be revealed using the simulation techniques described in the previous chapter. In this example, we were able to show that the cuddle calls were actually useful. The calls were increasingly answered by people who were at home during the day. It turned out that these people had a higher cancellation rate per se. However, of those who would have accepted a call, fewer canceled among those who had accepted the call.
How is the evaluation of measures (campaign management) usually carried out today? Exactly the same as with our telecommunications company. The number of callers is compared between those called and those not called. The example shows that this type of control works only accidently – at best.
Proper campaign management requires causal AI. This not only allows us to measure how much a campaign reduces the likelihood of churn. We can carry out this simulation for each individual customer. The effect can be greater for some customers than for others. Especially if the number of customers above the threshold is greater than the budget would allow. In this case, the customers for whom the planned measure is particularly effective are selected.
This shows that a forecast should ideally not be viewed in isolation from the planned actions. After all, the aim of the churn forecast is to carry out actions that avoid churn.
Customer Lifetime Value
Customer lifetime value has been a hot topic in marketing literature for thirty years, but hardly any company goes beyond a current sales or margin analysis. The reason for this is a lack of information about how likely it is that the customer will remain a customer in the future and how high the annual recommendation rate is.
A practicable customer lifetime value can be calculated if a number of churn forecast models are used in parallel. One that predicts cancellations in one year. Another for the second year. Another for the third year. And so on.
Alternatively, a forecasting model can also predict the sales or margin contribution that the customer will make in each of the future years.
The future years must then be discounted using the opportunity interest rate. The sum of all these values gives the customer value.
This customer value is in turn useful in many applications. In particular, it is needed when searching for the optimal threshold value of the churn model in order to correctly take into account the opportunity costs of the “missed opportunity”.
The advantages of causal AI for supporting operational decisions can be summarized as follows:
Generative AI is so much in the spotlight today that people often talk about “artificial intelligence” when they actually mean generative AI. The fields of application are diverse. The potential is only just being explored.
However, this sounding out reveals a peculiarity of us humans. We generate marketing texts, we generate advertising images and evaluate the results with our subjective impression. Is the result plausible? Does it make a “good impression”? This type of assessment is justified, but it also has many blind spots.
Anyone who has played with ChatGPT has had to realize that the results change greatly depending on how you “prompt” the system. Can we use causal AI to figure out how to prompt to generate marketing material that works better?
Dr. Steffen Schmidt is a marketing researcher like no other. He is constantly testing and combining the latest tools. When he sat among the participants at our first Causal AI seminar in 2009, it was immediately clear to me. Since then, he has been one of the pioneers in the application of our technology in market research. In 2023, he skillfully combined Causal AI with various Gen AI tools and massively improved the impact of Social Media Ads for Samsung. Here’s how he went about it:
Step 1 – Causal AI: He conducted a survey among smartphone users and measured which brand archetypes customers associate with the respective brands. He fed the data into a causal AI model to find out which brand archetypes increase the willingness to buy a smartphone. The strongest factor was “Explorer”, although (or precisely because) most brands are not perceived as Explorers.
Step 2 – Gen AI: He then asked ChatGPT to write a prompt for Midjourney to produce a social media ad for a particular SAMSUNG phone expressing the Explorer archetype. Without further choice, he accepted Midjourney’s suggestion. He generated the tagline using the Neuroflash app – an application specifically designed for generating advertising copy. The result was the slogan “The freedom to go further”.
Step 3 – Test: With the InContext market research solution, respondents experience Instagram, Facebook, TikTok or Amazon as if they were visiting them in real life. The solution uses a clone of the website and can therefore replace the advertising at will. After one minute, a survey that takes the response time into account measures the willingness to buy the Samsung brand. The comparison with conventional advertising showed an 18 percent higher market share for Samsung.
The result is amazing. The creation process did not require any advertising expertise. A very standardizable process combining Causal AI and Gen AI was used. Without further optimization, an increase in performance was achieved that would have required a lot of work, variants and test loops under conventional circumstances and would never have been implemented due to budget constraints.
The example is so powerful because the market researchers are not sufficiently successful in briefing the creatives so that creatives implement the results in the way market researcher did not intend.
I have experienced this myself time and again. Creative agencies often have a completely different understanding of what is written in the market researchers’ recommendations. What is an explorer archetype? How do you visualize it? Qualitative summaries often offer so much room for interpretation that the effect is lost.
Gen AI offers a standardizable and validatable interpretation mechanism. This is exactly what the manual process of a creative agency cannot provide.
While Causal AI is the more valid method for generating insights, Gen AI is the more valid method for creatively translating them into marketing material.
Will humans become reduntant? Not at all! Humans are taking the helm. With their knowledge and wisdom, they orchestrate both Causal AI and the control of Gen AI.
Marketing professionals quickly realized that although Gen AI can produce beautiful ads, the brand is not recognized in them in case of doubt because the design brand assets are not used. There are solutions to this and other weaknesses. For example, the addition in Midjourney –sref “URL” can provide the AI with an image and color language that the AI can use to create recognition.
Where Causal AI consists of AI, filter algorithms and specialist knowledge, a functioning creative system consists of GenAI, Causal AI and specialist knowledge.
Creative AI = Gene AI + Causal AI + Expert knowledge
The effectiveness of emails, letters and websites is largely determined by effective texts. This involves the right topics and arguments on the one hand and the right tonality and metaphors on the other.
To make Success Drivers’ own emails more effective, we experimented with the combination of Causal AI and LLMs in 2020. The latter already existed in the form of GPT2, among others. Our process should increase the open rates of our emails by no less than 500%.
Step 1 – Expert Judgement – We designed 50 variants for subject lines and had them evaluated by experts. The best and worst variants were tested in real mailings. As a result, the variants rated as poor were better than the supposedly good ones.
Step 2 – Trial & error: Now we got creative and tested the craziest variants in small random samples in real mailings. The resulting opening rates were to serve as the target variable for the next optimization step.
Step 3 – Optimization: At first, the subject lines were broken down into basic associations and categorized using natural language processing using the Neuroflash App. This categorization then served as input for the causal AI. This showed that a dominant language encourages recipients to open. We then prompted the LLMs (with the help of the then already existing Neuroflash tool) to suggest subject lines that were dominant but no longer than five words. The “straight to the point” subject line proved to be a direct hit. Open rates of 54%, previously thought impossible, became reality.
The same process is used to optimize newsletter headlines, texts, mailings or website headlines. The largest possible number of examples is required from which the AI can learn. The number of variants is more important than the sample size per variant. For example, we only sent 50 emails per subject line and were thus able to test 40 variants simultaneously with 2000 recipients.
ChatGPT is known for writing great texts and responding to the user’s wishes. But how do we know whether their requests lead to effective variants? This is precisely the evidence provided by a Causal AI-driven process.
Sounds complicated? Yes, it is not simple. But can you calculate the impact if a campaign improves by just 10%? What is the absolute additional profit? Probably more than ten times the investment in weeks. It quickly becomes clear that these AI-based processes pay for themselves very quickly.
Causal AI helps to become more efficient in all areas of marketing. It is a kind of AI upgrade that makes insights, decisions and creation more efficient.
Better insights
In marketing mix modeling, causal AI helps to get an analytical handle on the increasingly correlated channels. Their influence is determined more realistically and at the same time indirect effects, which are reflected in long-term effects among other things, are taken into account.
The drivers of the customer experience can be better understood. The application of Causal AI overcomes the fallacy that the topics most frequently mentioned by customers are also the most relevant. It makes it possible to simulate the financial impact of improving the customer experience.
Causal AI gives product innovation and optimization new methodological possibilities to find out which product characteristics and barriers play a central role in purchase intention and willingness to pay. This means that you are no longer limited by the small number of features of Conjoint or the explicit survey methodology of MaxDiff.
The impact of communication and advertising can be better understood through Causal AI. This deeper understanding goes far beyond the aspects that are queried in an advertising test. It is revealing to understand which emotions are “deadly” or what contribution brand building makes. It is particularly useful if the analysis can provide recommendations for the storyline, the emotional message, the choice of music or the type of humor.
A marketing strategy requires effective positioning, target group selection and segmentation. Causal AI can provide insights based on appropriate surveys to deliver effective brand positioning both in terms of content and association/emotion and to understand which customer types have an affinity for the product category.
Better decisions.
Causal AI is also used to create forecasts. However, these models are more stable and less susceptible to model drift. They also reduce the risk of discrimination against minorities.
When it comes to selecting products, advertising or packaging design, a Causal AI model can in some cases replace testing through market research or trial and error. This allows decisions to be made more quickly and cost-effectively.
When it comes to direct marketing or customer service campaigns, an individual decision can be made for each customer and target customer as to which marketing campaign should be used. This decision can be optimized using causal AI. Examples of this include reducing cancellation rates, estimating customer potential or customer lifetime value.
Better creation
Generative AI supports marketing in the creation of images, videos and texts. The contribution that Causal AI can make is to ensure that from the almost infinite possibilities, a variant is found that is highly likely to be highly effective, not just plausible.
More effective advertising motifs can be generated by feeding Causal AI on the basis of suitable market research, which then identifies which content and emotional-associative characteristics the image must express in order to be effective.
Effective messages are needed, for example, on websites, in subject lines or mailing headlines and in slogans. This can be optimized with causal AI by trying out many different variants in real experiments. By breaking down texts into their associative factors, Causal AI can identify the hidden characteristics of effective texts. With this knowledge, Gen AI can then be used to generate new, more effective texts.
Benefits = Better insights + Better decisions + Better creation
IMPRINT
Success Drivers GmbH
Johann-Heinrich-Platz 4
50935 Cologne
Germany
Success Drivers GmbH holds the copyright for any material, included translations, published at www.success-drivers.de
