Photo by Nadir sYzYgY on Unsplash

Revolutionizing Marketing with Causal AI

5 min. Read

The Marketer’s Dilemma:

In the high-stakes arena of customer marketing, the game is changing rapidly. Gone are the days when gut instinct and a spreadsheet could successfully carry a campaign. Today’s marketers face a formidable challenge: pinpointing the elusive sweet spot where maximum conversions meet minimum costs. Whether choosing exactly the right level of discount, which message to send, or just whether to contact the customer or not, marketers are not just seeking potential customers; they’re on a quest for the persuadable ones, those who will pivot when nudged by the right message, and at the right time.

But here’s the rub: traditional data science, as advanced as it is, often misses the mark in answering the real questions marketers are asking.

Imagine this: A marketer asks, “Who will upgrade their subscription, only if I target them?”

Instead of a direct answer, they would usually receive propensity scores from their data science teams – insightful, yes, but not addressing the critical “only if I market to them” part. It’s akin to asking for directions and being told which routes are paved, gravel, or dirt. This is useful information, but it does not answer the question.

This gap in understanding isn’t due to a lack of effort; it’s because the question is too complex for the available tools. We’re talking about distinguishing between people who will respond positively to marketing efforts and those for whom such efforts are a waste or even harmful. AI is still racing to catch up with this nuanced challenge. Despite consistently capturing the headlines, LLMs’ are not that much more helpful in answering these types of questions either.

Today’s Marketers are grappling with complex, multifaceted questions:

  • Who will change their behavior in response to our outreach?
  • What specific offer or channel will trigger this change?
  • Which high-risk individuals should we actually avoid?
  • What’s the magic number in terms of discounts to sway their decision?
  • How would our results vary with a more targeted approach?
  • Which tweaks in our imagery or messaging make a real impact?
  • Out of our arsenal of creatives, which one will hit home with each customer segment?

This isn’t just about data, it’s about understanding the human element behind the numbers.

Causal AI: The Game-Changer in Marketing Strategy

“Causal questions can never be answered from data alone.” – Judea Pearl.

This quote from Judea Pearl, author of “The Book of Why” encapsulates perfectly why something new is required and why no amount of Large Language Modeling or Deep Learning is good enough. Causal AI, a relatively new entrant into the AI space, is revolutionizing how we approach these challenges. Unlike traditional data analysis, which often stops at finding patterns, causal AI digs deeper. It’s about understanding the effect of specific actions, isolating the impact of individual variables in a complex ecosystem, and understanding the true power of Cause and Effect.

One of the simplest descriptions of this challenge comes in the form of Simpson’s Paradox, a statistical conundrum that can, at first, confuse analysts.

Simpsons’s Paradox – why data alone can mislead and confuse

Imagine we want to evaluate the effectiveness of a vaccine and we have some efficacy data. What does it tell us?

It appears the vaccine doesn’t work, as the mortality rate is higher in the vaccinated population, right?

But what about if we reveal some more information?

Ok, so the vaccine appears to work really well in the 10-29 cohort AND the 30-59 cohort, but not when you look at the population as a whole. So does the vaccine work or doesn’t it?!

It doesn’t take much for a seasoned experimentalist to unpick the problem (you have to control for age because older people simply have higher mortality rates and are being vaccinated at a much higher rate, which is biasing the overall result. In this example, age is an example of a ‘common cause’ of death that, if not understood, would lead to inaccurate results).

So yes, in this case, vaccines might appear to work, at least given what we can see here. But even then, one can’t really be sure whether there are other confounding variables at play.

Simpson’s paradox illustrates why data alone can be misleading and underscores the need for causal reasoning to discern the actual impact of interventions.  In the marketing scenario, there is a similar need to identify sub-audiences who behave differently and might not receive the correct treatment if the nuances of cause and effect are not understood. Mistakes can lead to customers who might have made a purchase without the need for an offer or discount, equating to a real waste of marketing budget. 

Imagine if there was a way to systematically determine a way to exclude these people from a campaign and focus your budget where it can really make a difference. The challenge is that to do this requires some subject-matter expertise on the sort of data that you need to capture (and control for) in order to make the correct evaluation. Historically there has been no good way to systematise this process, so marketers have had to rely on their own expertise combined with manual A/B testing to try and answer these types of questions. 

Causal AI brings a solution to these challenges.

Brace for Impact: The Causal AI Revolution is coming

Causal AI is a relatively new way to tackle questions of cause and effect. It allows practitioners to inject subject matter expertise into the modeling problem by encoding that knowledge into causal graphs that AI can leverage to determine how to suitably control the data.

The models and methodologies provided by Causal AI bring an arsenal of new tools to the marketer’s toolkit, enabling strategies crafted with a precision and insight previously unattainable. For marketers willing to embrace this revolution, the possibilities are as boundless as they are exciting. Welcome to the new era of cause-driven marketing, where prescriptive recommendations replace predictions and propensity scores as the tools marketers use to spend their budget and achieve ROI.

The field of causal AI is soon to enter a meteoric rise. Indeed, Gartner has identified Causal AI as a technology to watch in their most recent AI Hype Cycle. We’re witnessing an explosion in sophisticated methodologies and a growing cohort of commercial entities specializing in this space. This isn’t merely a technological advancement; it will drive a paradigm shift in understanding how marketers and AI can partner together to achieve significant results.

At Intent HQ, we are deploying Causal AI at some of the world’s biggest companies – helping them optimize their campaigns (and their marketing budgets!) for true long-term value.

Our next AI blog will focus on why we believe the concept of ‘customer journeys’ is already obsolete and how Causal AI will change the paradigm from curated journeys to genuine relevance through True Personalization.

About the author:

Andy Cole, Global Director, Applied Data Science, Intent HQ

Andy Cole is an applied data scientist, analytics leader, and data strategist with 15 years of experience delivering bleeding-edge customer machine learning and analytics strategy in the Telco, Financial Services and Retail Sectors.  Andy specializes in delivering behavioral analytics at scale. 

At Intent HQ, he is spearheading the application of Causal AI, a branch of artificial intelligence that aims to understand and model cause-and-effect relationships in data.