By Berend Booms | Associate Editor | Future of Assets

The Eurovision Song Contest is one of my guilty pleasures. I grew up watching the show every year, and have many fond memories of evenings spent sitting in front of the television. Getting to stay up late enough to watch the full final was exciting as a kid, but not nearly as exciting as the weird, wonderful, dramatic, strange, and funny spectacle itself. Eurovision is pure chaos, but in the best possible way.  

Now that I am older, I still watch Eurovision, albeit in a different way. I no longer sit together with my siblings in the same room to watch it, but the experience is still immensely enjoyable. Every year, my sister and I message each other throughout the semi-finals and final, reacting to every performance as it unfolds. Some artists are praised, others are met with complete disbelief. We revisit inside jokes that keep returning every year and generally enjoy the kind of absurdity that Eurovision seems uniquely capable of producing

Predictive Analytics in Eurovision 

Beneath all of the spectacle is another dimension I have increasingly started to enjoy, probably much to the amusement of my former statistics professor: Eurovision is an incredibly interesting exercise in prediction. The amount of data surrounding the competition is mindboggling, with most of it leading up to who various pundits tip to be the winner. The odds shift continuously in the lead-up to the final, as semi-final performances propel candidates to new heights or dampen the spirits of pre-show favorites. Rehearsal footage is analyzed frame by frame, almost religiously. Streaming numbers, social media engagement, running order, historical voting patterns, jury preferences, and fan favorites all become part of an enormous ecosystem of signals that people use to try and predict the outcome.  

In anticipation of tonight’s first semi-final, I decided to take a deeper look at this predictive landscape. I analyzed bookmaker predictions across the last ten years of Eurovision, from 2015 through 2025, focusing on two areas: how accurately bookmakers predicted which countries would qualify from the semi-finals, and how accurately they predicted the top three and eventual winner in the grand final.  

What the Data Actually Reveals 

What stood out immediately was that across the board, bookmakers were consistently better at broad classification than precise prediction. Across the last decade, bookmaker accuracy for predicting semi-final qualification generally sat in the mid-to-high 80 percent range. Predicting the final podium proved to be more difficult, while correctly identifying the winner happened only about half the time.  

It’s clear that the prediction market is usually very good at identifying the cluster of likely contenders. It can separate the strongest acts from the weakest with a fairly high degree of reliability. Predicting the exact order in which those top contenders would finish is much more difficult. These same challenges pop up when we look at predictions in asset management. 

Prediction, Risk, and Asset Management 

One of the most common misconceptions surrounding predictive analytics in asset management is the belief that predictions should be exact. There is often an implicit assumption that if enough data is available, we should be able to identify precisely which asset will fail, exactly when it will fail, and what the conditions are under which that failure will occur. Besides being far from reality, this is rarely how predictions create value.  

Most predictive models become valuable much earlier in the process. Their strength lies in helping organizations distinguish elevated risk from normal operating behavior. They offer a leg-up in prioritization, help establish where attention should go first, support triage, planning, strategy, and resource allocation. This is exactly what the bookmakers were best at as well.  

Even in years where bookmakers failed to predict the victor, the eventual winner almost always came from that small cluster of top contenders. This shows that the prediction was directionally strong, even if it was not perfectly precise. What I found most interesting is that this also shows how more data does not automatically improve prediction accuracy over time. 

Eurovision’s digital footprint has exploded: social media engagement is enormous, rehearsal footage is available from every angle imaginable, and there are thousands of fan reactions. The gigantic amount of live stream data, betting-market liquidity and algorithmic sentiment analysis create exponentially richer datasets than were available ten years ago. If we follow the common narrative surrounding analytics, especially with the recent AI boom, we would expect bookmaker predictions to become steadily more accurate with each passing year. But somehow, the upward trend you would expect to see is missing. 

In fact, 2025 turned out to be one of the weaker prediction years in the sample I reviewed. The market favorite failed to win, and the widely expected top three only partially stood on the podium at the end of the evening. My takeaway is that while more data helps improve visibility, visibility in itself does not eliminate uncertainty.  

This is something I think organizations underestimate when discussing how to combine AI, predictive analytics, maintenance strategy and connected assets. Our industry is still governed by the assumption that collecting more data naturally and logically leads to better decision-making. While additional information absolutely helps create better contextual understanding and improves situational awareness, it can also introduce noise and overconfidence.  

The narratives surrounding Eurovision emerge and explode rather quickly. Once an artist becomes the favorite, the surrounding conversations tend to reinforce this belief. Analysts, fans, bookmakers, social platforms: they all seem to sit in this echo chamber, amplifying the same signals. Consensus builds momentum, and the same thing can happen in industrial asset management. 

When all of your dashboards, top-of-the-line anomaly detection systems, high-tech condition-monitoring tools, and predictive models all point in the same direction, it becomes very easy to treat the output as deterministic truth rather than probabilistic guidance. The danger is not necessarily the technology that enables these predictions, but the confidence we attach to them.  

What I found particularly interesting in the Eurovision data I looked at is that prediction errors usually occurred near the cutoff points. The bookmakers rarely struggled with obvious favorites or obvious outsiders. The inaccuracies emerged around the borderline cases: the countries sitting around tenth or eleventh place in semi-final qualification odds, or the artists competing for the lower end of the podium.  

This same distortion rears its head in asset management. Most organizations are reasonably capable of identifying assets in poor condition, and truly critical failures rarely emerge without warning. The more difficult challenge lies in interpreting these so-called borderline cases correctly. Which assets require intervention now, to prevent more serious failure from happening in the future? Which deviations are noise, and which are early indicators of degradation? 

The Difference Between Prediction and Judgment 

Put this way, I would argue that useful prediction models do not remove uncertainty, nor is this their intended design. Instead, they help organizations navigate uncertainty more effectively. That’s why human judgment remains so important, regardless of how sophisticated predictive systems become. Data can surface patterns; models can rank probabilities; AI can process far more information than you or I ever could. But context, interpretation, experience and accountability still sit with people. 

Watching Eurovision every year reminds me of that in a very entertaining way. You can study and track the odds, watch every single rehearsal clip, analyze the historical voting blocs, and interpret every signal imaginable; once the performances begin, millions of people vote and uncertainty re-enters the system in full force. 

In the same way, we can instrument our assets extensively, we can build increasingly advanced predictive models, we can monitor vibration, temperature, pressure, lubrication quality, runtime behavior, and operational context in real time. But in industrial asset management, complex systems remain just that – complex. Human behavior changes, operating conditions shift, external variables introduce compounding uncertainty, as priorities continue to evolve. 

Whether you are sitting on the couch watching Eurovision, or running a multi-billion dollar asset-intensive operation, the lesson is the same: prediction is not about eliminating uncertainty altogether; it is about creating better conditions for decision-making. The real strength of predictive systems does not lie in their ability to guarantee outcomes. Their value lies in helping us better understand where attention is needed most, where risk is increasing, and where intervention is likely to create the greatest impact. In environments shaped by uncertainty, the real advantage does not come from predicting every outcome correctly, but from responding intelligently when reality diverges from expectation. That’s why this year, twelve points go to human judgment.