At the Allphins Spring roundtable, reinsurance professionals from across the market gathered to explore how Exposure Management (EM) is used not just to identify peak exposures, but also to align capital with risk in a more responsive way. As Laurent de la Porte, Allphins' CEO puts it, EM has a role in "evaluating concentration, gaps and exposure so that you can optimise allocation."

Peak Zones vs. Hidden Aggregations

Our conversation began with the familiar territory of peak peril regions: where does EM add most value as traditional modelling starts to falter, and equally, what are its limits? One senior EM Manager highlighted the 'usual suspects':

"We're always asked to look at Gulf wind storms, California quakes. These are areas where we're closest to tolerances and see the greatest concentration of risk; but where, if we added spatial geocoding, EM could make the most difference to the model."

But while these zones remain a key concern, the group agreed that the bigger challenge is spotting hidden aggregations and cross-class accumulations.

Vanessa Jones, Head of Exposure Management at Dale Underwriting Partners, says "Earthquakes are the one that really keep me up at night. You've got huge levels of aggregation on the West Coast, for example, and basically an untested model. Add in the contagion risk from fire-following, and the exposure potential is enormous. We've just seen in the wildfires, the contagion risk, once you're in a built-up area, is huge. That's not necessarily as well captured as we think it could be in our modelling and aggregation."

James Simpson, Head of Exposure Management at Blenheim Syndicate added, "It's the event we haven't thought about - the one that connects across everything, that worries me; or some of the contract terms being tested and not holding," he said. "It's those connecting events - those surprises - that are most dangerous."

Another of our guests agreed, citing one of many recent examples of hyper-connected risks: "The Ukraine conflict showed us how quickly things can link together – the war portfolio itself, trapped ships for the cargo portfolio, aviation losses… add something like cyber to that and it's a perfect storm. That's where, as an Exposure Management professional, I see whole-platform exposure."

Overexposure and Market Share

The purpose of the analysis, of course, is to assess the quality of the portfolios covered, based on perceived risks and to similarly query the value of modelling data. Says Jones, "We're always trying to look at how well-weighted the portfolio is. Are we overexposed to one type of event across multiple books - treaty, insurance, etc.? Are we diversified enough to withstand something going wrong in a single area? Terror is another one that I worry about because as well as cross-class potential, there's often slightly poorer data than you'd see on the Nat Cat side."

But another theme that emerged was relative exposure, compared to the broader market. "'What are the peak risks' is a simple question – it's the same as everyone else's! If there's a big event and everyone takes a hit, that's fine - if you're well-enough capitalised, you will cover yourself", says Mathias Borjesson, SVP Underwriting at Renaissance Re. "But if we have a disproportionate loss because we were unknowingly overweight, that's much harder to explain."

These concerns have prompted many reinsurers to attempt internal assessments of their market share exposure—though most acknowledged that it's more art than science. "Increasingly, we're trying to identify what the constraints really are. But we don't have the historical benchmarks", Borjesson continues.

If EM can highlight where a reinsurer is overexposed, can it also point to books where you're under-exposed – and thus identify a commercial opportunity? Says Simpson, "I think we know where we're not exposed. And it's normally for a reason! But in a softening market, for example, when rates are falling, you have to ask: do we stay where we're comfortable, or expand into new areas for diversification? So EM is only one of several components in these decisions. A few years ago, you could just stay with one position, but that's not the world we live in anymore. We do look at the holes in our coverage, and it's quite a topical discussion to keep our position."

But Simpson warns that caution remains paramount. "It's easy to be tempted into areas that look attractive in the model," he says. "But that's how you find yourself overexposed to a region you don't like, just because the model said it was cheap. Now's the time to maintain discipline. We certainly try to."

Diversification and the challenge of return periods

Of course, the key purpose of modelling for a reinsurer is to quantify the diversification of the portfolio, but our panellists agreed that there is no single or systematised approach. Jones described a "logic tree" approach, starting from top-level splits and working down to regional exposure and peril mix. "We try to keep a consistent balance – which for us is US critical cat, wildfire, retro - and smooth it out across the capital model. In the property line, we have the most intense capital use."

Simpson endorsed the capital metric and noted the importance of aligning with actuaries. "You've got to work with the actuaries to make sure the capital model reflects what's actually in the portfolio and that they understand the limitations of the exposure data. We have to work alongside them, not for them." In fact, Simpson says that aligned data, reporting and capital are worth more to a balanced business than moving fast: "You want to be timely, but not too timely," he said. "When a model version updates, we wait six to ten months before moving. Take your time. Get everything right."

But not everything is right. A recurring frustration for our contributors was the interpretation of return periods (which of course inevitably directly influences capital modelling). They felt that return periods are a rather corruptible commercial lens rather than a true reflection of probability. Says Jones, "You have to be really careful with the biases and anchoring to previous opinion in expert judgement." She points out that this is partly because many scenarios are over 20 years old and need updating. Simpson concurs: "I leave the return period field blank on some RDS forms. Everyone has a different number. We should focus on loss magnitudes and likelihood, not something that's made-up."

Borjesson adds: "People overfit to historical scenarios rather than thinking about the loss process and type of future events that might occur. It might be 1:250 for the specific circumstances of a Deepwater Horizon to happen again, but we're not interested in the specific event, we want to understand a representative event. 'What's the chance of a loss of that magnitude happening in your book' creates a very different set of answers."

Jones also says that return periods suffer from not being intuitive. "People don't think in 1-in-200 terms. That's much longer than anyone's experience. It's better to talk in terms of probabilities – 'there's a 0.5% chance that something will happen this year, now.' That's actionable." Simpson agreed. "That's the difference between exposure management and everyone else. Everyone else is used to return periods, because of the media. Spinning it around and educating internally to bridge the gap from return to probability is a challenge."

Challenges in Retro

All our participants reflected that retro is particularly opaque in contributing to understanding the diversification of a portfolio. The natural structure of the packaging of risk encourages a defensive posture because retrocession is priced as if the worst-case scenario is always just around the corner, rather than mitigated (or indeed conceivably accentuated) by factors like localisation.

And that attitude is self-reinforcing if you rely on the data alone. Simpson said: "I think there's a danger in being overly scientific. There was a product about 10 years ago that relied solely on modelling data. It looked clever, it was priced low—and it failed. It didn't make money because the models just don't tell the full story. Nowadays, you look at everything and price accordingly. You don't rely on a model you don't understand,"

Retrocession also tends to react faster than the reinsurance market it supports. "When uncertainty rises, rates go up quickly on the retro side - sometimes well ahead of the underlying reinsurance market," Simpson adds. "That can create real disconnects. Your clients may not have moved their pricing, because nobody likes paying more for their insurance. But your retro costs have risen sharply. It makes managing the economics very tricky."

Finally, because most underwriters treat retro limits as undifferentiated from other risks, they simply add the deployed limit to the portfolio total. This creates an overly cautious picture, so there's little pressure in the market to change. "It's more conservative," said Jones. "But that satisfies boards and capital providers. So the incentive to do it better is low—unless you're running up against your net." In this sense, retro is a very specific type of drag on market innovation.

From Monitoring to Strategy

Our roundtable made clear that the use of exposure management to balance risk in reinsurance is not just about tracking tolerances, but instead actively shaping the portfolio: anticipating concentration risk, finding overlooked correlations, and pushing back on false certainties.

It's also a voice of discipline. In a market full of shifting rates, evolving risks, and untested assumptions, exposure management provides the rigour to ensure that what looks good on paper doesn't create dangerous asymmetries in practice.

Like all data tools, what you get out is always a reflection of what you put in; so EM has imperfections. But insight shapes strategy, and EM offers robust insight in a discipline where the temptation to mould the theory to the narrative is always great.

‍