As insurers pull back from California’s fire-prone regions, one MGA writing excess and surplus (E&S) business is taking a different approach. Delos Insurance is underwriting property coverage using a wildfire risk model developed in partnership with the Spatial Informatics Group, a research think tank known for its work with Cal Fire and the California Public Utilities Commission.
“Spatial Informatics Group has been doing some of the most important wildfire modeling for civil governments for decades,” said Shanna McIntyre (pictured), chief data officer at Delos. “Through our partnership, we have exclusive access to all of that data for anything insurance related.”
Delos’s model blends this environmental science with machine learning. According to McIntyre, it has been rigorously tested and has successfully predicted the scope of major wildfires over the past seven years. “We innovated using machine learning on geospatial datasets,” she said. “It’s not just about historical loss. We include climate projections and focus on worst-case scenarios.”
The model is also aimed at addressing volatility in the market. Sharp pricing swings have contributed to carrier exits and non-renewals in high-risk ZIP codes. “We like to make sure that we don't have to have large swings between one year to the next in our view of risk,” McIntyre said.
That stability, she added, is partly what makes the E&S space viable for Delos. As a non-admitted carrier, the company is not bound by the same rate approval processes as insurers in the admitted market. Still, McIntyre emphasized Delos’s ongoing engagement with regulators. “We always want to make sure that we are communicating really well with the Department of Insurance,” she said.
Delos also uses its model to distinguish insurable properties in areas often redlined by traditional markets. McIntyre said their analysis finds that 65% of homes flagged as high-risk by legacy carriers are acceptable risks under the Delos model.
Rather than relying solely on theoretical modeling, Delos validates its approach against actual insurance losses. “The most important role that we use for testing our model are people who have access to loss data at insurers and reinsurers,” said McIntyre. The company also consults with wildfire experts and has a former Cal Fire chief of staff among its advisors.
This commitment to empirical testing, she said, is what makes the model viable for underwriting in today’s environment. “We undergo back-testing against historical fires, and the model has accurately captured the full extent of the risk in advance of the most significant fires in recent years, including those in Los Angeles.”
Before joining Delos, McIntyre worked in aerospace engineering, where she focused on predictive modeling. She sees strong parallels between the two fields. “Both are highly regulated and involve complex systems with many interdependent parts,” she said.
As wildfire behavior evolves – driven by climate change and events like dry lightning storms and urban conflagrations – Delos believes it can carve out a role in covering underserved markets. The company has already insured more than 25,000 policyholders in California and continues to grow its capacity partnerships.
Whether this model-driven approach will hold up under ongoing climate and regulatory pressure remains an open question. But in a retreating market, it offers one potential roadmap forward.
