Chief Executive Officer / Europe
Choices that cost us something
- Insurance revenue
- €5.1bn
- Talvermark SE, 2025
- Solvency II ratio
- 192%
- Year-end 2025; board range 170% to 210%
- Undiscounted combined ratio
- 95.8%
- 2025; 92.3% after discounting
- Insurance and reinsurance
- 28 years
- Actuarial, treaty underwriting and general management
Decisions & consequences
01Keeping flood cover available by pricing it honestly
The July 2021 floods in Germany and Belgium cost Talvermark's commercial property book €310 million gross and €118 million after reinsurance. Our small-business package included natural hazard cover with a flat loading. Hazard zones were used only to accept or decline new business, so existing customers in the highest zones paid much the same as customers on higher ground. After the floods, our reinsurers asked harder questions at renewal, and the honest answer was that we did not know what our own flood price was.
The stakes were reputational as well as financial. Several competitors had begun excluding flood for high-zone renewals. Our chief financial officer proposed that we do the same for the highest hazard band, which would have removed most of the exposure quickly. Our head of distribution wanted any re-rating phased over four years to protect broker relationships. I rejected both. Excluding flood would leave businesses uninsured for the risk they most needed covered; a four-year phase would leave us knowingly underpricing for three more seasons.
We re-rated 41,000 policies in high-hazard zones over two renewal cycles, with an average increase of 34% in the highest band. Alongside it we introduced a prevention credit of up to 15% for verified measures such as flood barriers, raised plant and water-resistant fit-outs, and spent €9 million funding surveys. Brokers received the zone, the loading and the credit criteria for every affected policy.
The cost was visible. Retention in the highest band fell from 89% to 81%, and we lost about €21 million of annual premium. By the end of 2024, 6,300 businesses had installed measures that qualified for the credit. In a flood affecting our Austrian portfolio that year, the average claim for protected sites in a sample of 212 claims was 58% lower than for comparable unprotected sites.
What changed was the conversation: flood became something a customer could act on. I would involve chambers of commerce and municipalities earlier, because many tenants could not install measures without a landlord's consent, and we discovered that too late.
Selling a motor book we had been told to defend
In 2021 Talvermark still wrote €780 million of personal motor premium, a legacy of an earlier merger. It was 13% of our premium and employed 410 people. The book had produced an undiscounted combined ratio above 104% in four of the previous five years. We lacked the scale in claims and repair networks that the strongest direct motor insurers had, and motor consumed management attention and capital that commercial lines needed.
Selling it was not an obvious decision. It made us smaller in a market where size is often equated with strength. Our head of retail distribution argued that motor brought us small-business owners who later bought commercial cover. Two employee representatives on the supervisory board opposed the sale because of what it meant for colleagues. An analyst I respected described motor as our only consumer brand. The trade-off was real: we would give up premium, the cross-selling route and a measure of public recognition.
I commissioned an analysis of the cross-selling claim, which found that 6% of commercial customers had first come to us through motor. That was meaningful but not decisive. I then took a paper to the supervisory board that set out the arguments against the sale in full, including the employee representatives' concerns in their own words. We ran a sale process in 2022 and agreed terms with Ardwyn Direct, a direct motor insurer, for €265 million. Colleagues transferred with the business on their existing terms, and we negotiated an eighteen-month service agreement to protect customers during migration. The transfer completed in January 2023 after supervisory approval.
The sale added 14 points to our Solvency II ratio. We allocated the released capital explicitly: €150 million returned to shareholders through a buyback, €90 million to commercial underwriting tools and data, and the remainder held within our range. The retained business has run below a 97% undiscounted combined ratio in each year since.
If I did it again, I would start the employee consultation earlier. The representatives were right that colleagues heard rumours before they heard reasons.
Reporting the combined ratio before the discount
IFRS 17 took effect for Talvermark's 2023 financial year. During preparation in 2022, our finance team showed me that discounting incurred claims at the prevailing interest rates would reduce our reported combined ratio by between three and four points. Nothing about our underwriting would have changed. Several peers were preparing to lead their reporting with the discounted figure, and our results would look weaker by comparison if we led with the other one.
The disagreement was direct. Our chief financial officer and our head of investor relations preferred to lead with the discounted ratio, because it was the figure the standard produced and because comparability with peers mattered to analysts. They were not wrong about either point. My concern was that a discount benefit worth several points would shrink if interest rates fell, and that investors would read that shrinkage as underwriting deterioration. I also did not want our own underwriters to start reading their portfolios through a figure that interest rates were improving for them.
We agreed a compromise that was mostly mine: lead with the undiscounted combined ratio, publish the discounted figure alongside it, and show the bridge between the two every reporting period, split into the discount effect, the risk adjustment and other items. We also published the confidence level of our risk adjustment, which the standard requires, with a short explanation of what it means. I presented the approach at our capital markets day in November 2022, before the first results under the new standard.
The cost was some months of awkward comparisons. In our first IFRS 17 year-end results, published in March 2024, our headline ratio was higher than most peers'. By 2025 our undiscounted ratio stood at 95.8% and the discounted figure at 92.3%. Analysts' estimates of our discount effect had ranged from 1.9 to 5.2 points before we published the bridge; afterwards the range narrowed to within half a point.
What I would change is the timing of our internal training. Our underwriters learnt the new numbers from the press release.
Taking back the pen from partners who would not share data
In 2022 Talvermark wrote €1.1 billion of premium through 64 partners with binding authority, mostly managing general agents and broker networks. The loss ratio on that delegated book was 71%, compared with 61% on business we underwrote ourselves. Risk-level information often arrived months late in summary form, and in some property schemes we could not tell which buildings we insured until after a loss. That made our catastrophe accumulation figures less reliable than our board papers implied.
The stakes went beyond profit. Under the Insurance Distribution Directive we remain responsible as manufacturer for the products these partners sell, and I could not explain to our supervisor how we oversaw business we could barely see. Our head of broker distribution warned that tougher terms would push volume to competitors. One of the largest consolidated brokers in our markets asked for a higher commission and a profit share in return for continued placement. Accepting would have kept €140 million of premium on our books.
I decided that binding authority would depend on information. Partners had to provide risk-level data within thirty days and accept our pricing for property risks above a set size. We moved commission away from volume tiers towards a structure linked to data timeliness and loss experience. The consolidator declined the terms, and we let that business go. We gave smaller agencies a twelve-month transition and built a shared data portal with them.
By the end of 2025 we had 38 binding partners. Delegated premium fell to €860 million and its loss ratio improved to 63%. The share of risk data received within thirty days rose from 54% to 93%, which allowed our accumulation reporting to cover the delegated book properly for the first time. Some of the agencies that stayed have grown with us since.
What changed was the principle behind delegation: we now delegate judgement, not ignorance. I would have built the portal before announcing the conditions. Several capable small agencies nearly left because the systems lagged behind the policy.