Why Shark Attacks Won’t Sell More Ice Cream – And Other Data Pitfalls with Senior Data Scientist Ola Lindeberg ex Telia, Ivbar, Swedish Armed Forces

🐟🍦 If we just increase the number of shark attacks, we’ll sell more ice cream! Sounds ridiculous, right? Yet, many businesses unknowingly fall into similar data traps when working with predictive analytics. While the two graphs might correlate - one is not causing the other. Could the sun be at play as the common determining factor?

The shark anecdote is one many great anecdotes Ola Lindeberg shared in the latest episode of the Hello $Firstname podcast. Ola, an indeed very senior data scientist with experience from Telia Sweden, Ivbar, and Swedish Armed Forces, joined me to discuss the flawed assumptions businesses make when using predictive analytics—and why prescriptive analytics would often be better if you want to drive real behavioral change in personalization. (OMG also just remembered the anecdote with the genAI powered spoof caller trap!!! It’s towards the end of the episode!)

🔗 How This Ties Into the CX Layers

At the core of our conversation was the Data Layer in the CX Layers framework. While most personalization setups heavily rely on data-driven insights, many fall into the trap of using predictive analytics without considering how to act on those insights. In other words, they’re forecasting the weather instead of deciding what to do about it.

Ola’s perspective perfectly illustrates why the Data Layer isn’t just about insights—it’s about decisioning and action. And that’s where prescriptive analytics comes in: Instead of just predicting what a customer might do, it helps businesses determine what action should be taken to actually influence customer behavior.

🔍 Key Insights from the Episode:

📊 Predictive vs. Prescriptive Analytics: Why Knowing Isn’t Enough

• Predictive Analytics tells you what is likely to happen (e.g., “This customer might churn”).
• Prescriptive Analytics helps determine what action should be taken to influence the outcome (e.g., “This content or offer could actually prevent churn”).
• Businesses often mistake correlation for causation—leading to wrong conclusions and poor personalization decisions.

🌦️ Why Wearing a Jacket Won’t Change the Weather (And Why Your Marketing Might Be Doing Just That)

• Another great analogy from Ola: You check the weather, see it’s cold, and decide to wear a jacket. But that doesn’t change the weather - which might be what you’re actually after… figuratively speaking of course.
• Many companies use predictive models the same way—reacting to insights without considering how to actually influence customer behavior.

⚡ The Booking.com Experiment: Why One Day in Paris Sells More Trips

• Booking.com tested an AI-driven recommendation model to suggest trip lengths based on user behavior. When they suggested three days (which was indeed the most frequently bought option), it had no impact on conversions.
• But when they suggested just one day, it triggered higher engagement and bookings, even though most people ultimately booked longer stays.
• The takeaway? Sometimes, the best way to drive action isn’t the most "logical" data-driven assumption—but rather an approach that nudges customer behavior effectively.

🤖 The Role of Generative AI in Decisioning

• Generative AI is already streamlining content production, but its impact on decisioning is still developing.
• Ola sees potential but argues that businesses need to test and quantify the impact of AI-generated personalization efforts rather than assuming they will automatically improve engagement.

🚀 Final Takeaway: Stop Predicting—Start Influencing If you’re only using predictive analytics, you might just be f