Pricing teams have never had more internal data. Yet many still struggle to answer a simple question: Are we pricing risk deliberately or drifting out of alignment with the market without realising it?
We sat down with our Lead Data Scientist, Kalle Myllarniemi, to talk about how our new machine learning capabilities, Apollo and Atlas, came to life and what they now make possible.
Kalle:
The starting point was scale.
Over the last few years, we reached a point where we are collecting 600,000 prices per day across multiple sectors. That volume of data fundamentally changes what is possible. When you combine that with the increasing number of brands in the market and the growing diversity of pricing strategies, it becomes clear that traditional benchmarking only tells part of the story.
Competition is more complex than it was five or ten years ago. There are more underwriting appetites, more segmentation strategies and more nuanced rating interactions. Pricing has become more multidimensional.
Apollo and Atlas were born from that intersection of unprecedented data scale and increasing strategic diversity in the market. We realised we were not just observing prices. We were observing behaviours. If you can model behaviour, you can understand strategy.
That was the shift.
Kalle:
Pricing teams are exceptionally knowledgeable about their own books. They understand their rating structure, their claims performance and their underwriting appetite in depth.
But they often have limited visibility into how others are pricing, particularly in segments they are not currently writing.
That creates a structural blind spot.
If you are not writing a particular risk type, you do not have claims data for it. Without claims data, it becomes difficult to confidently price into that space. This is one reason insurers can hesitate to expand appetite. The uncertainty feels commercially risky.
At the same time, adverse selection remains a constant threat. If your pricing does not accurately reflect how the wider market is segmenting risk, you can end up getting selected against and attracting risks that others have deliberately priced away from.
Apollo addresses both issues.
It allows insurers to understand how the broader market prices risks that sit outside their current footprint. It also reveals the diversity of pricing strategies in play. Many pricing professionals know their own model inside out, but Apollo shows how weighting structures, feature interactions and segmentation approaches differ across the market.
It moves the conversation from “where are we priced?” to “how are we positioned?”
That is a fundamentally different level of insight.
Kalle:
Genuinely exciting and occasionally humbling.
When you first start analysing Apollo outputs, they can challenge assumptions. You might see an insurer showing significantly less weighting for licence tenure than the broader market, which initially feels counterintuitive.
But when you interrogate feature interactions and clustering, the story starts to emerge. Perhaps that insurer is placing much stronger emphasis on claims free driving experience instead. What looks like underweighting in isolation becomes a deliberate strategic trade off.
That is when it becomes powerful.
Apollo does not just surface differences. It helps explain them. You start to see the narrative of different pricing strategies take shape. Certain insurers lean into experience signals, others into tenure, others into postcode granularity or behavioural proxies.
Seeing those patterns crystallise was the moment it became clear this was more than a modelling exercise. It was strategic intelligence.
Kalle:
Several things.
First, it provides aggregated market models, a view of how the market prices risk overall.
More importantly, it shows how your pricing structure aligns with or diverges from the main pricing strategies operating in the market. You can see which strategic cluster you most resemble.
Second, Apollo highlights rating factors that others appear to be using heavily but you are not, and equally, factors where you are distinctive. That can trigger valuable internal discussion. Is this deliberate differentiation or an unintentional gap?
Third, it introduces a market lens on rating factor interactions. Pricing does not operate in isolation. Factors interact. Apollo helps teams understand how those interactions compare with broader market patterns.
Finally, it provides a test bed.
If the analysis highlights a potential opportunity, for example adjusting weightings in a specific segment, insurers can model the impact and view a before and after position relative to market behaviour. That reduces uncertainty and allows for more confident implementation of change.
It moves external data from descriptive benchmarking to actionable simulation.
Kalle:
In practical terms, it improves risk selection and pricing accuracy.
With Atlas, insurers gain more granular categorisation of risk, particularly in areas like postcode segmentation. That leads to more accurate pricing of location based risk and can directly improve loss ratios.
Apollo complements this by ensuring rating structures are aligned with observable market behaviour, reducing the risk of adverse selection.
The combined effect is sharper competitive positioning and better technical performance.
It is not about copying the market. It is about understanding it well enough to choose your position deliberately.
That is where the commercial value sits.
Machine learning in insurance often focuses on internal optimisation.
Our approach focuses on market intelligence.
Because sustainable pricing performance is not just about building better models. It is about understanding how your pricing logic sits within a live, competitive market.
Download the full information packs to see how market-grounded machine learning can strengthen your pricing strategy.