Bermuda's delegated-underwriting push hits a bordereaux bottleneck
Convex runs a fifth of its business through MGAs and MGUs and says it would write more — the ceiling is administrative, not appetite.
At the Rendez-Vous de Septembre in Monte Carlo, Convex chief underwriting officer Doug Howat put a figure on the channel Bermuda reinsurers keep naming as their route to growth beyond the traditional catastrophe market: about a fifth of the group's business comes from what it calls high-volume underwriting, most of it delegated through managing general agents and managing general underwriters. A fifth of a Bermuda balance sheet running through third-party pens is a concrete sizing, and yet Convex caps the volume it will write through delegated authority even though the appetite is there, because ingesting coverholders' bordereaux data is, in Howat's word, quite clunky.
What the channel buys is reach: business in markets where the group has no direct presence, specialist underwriting expertise, and policy-administration systems better than its own. What it costs is administrative, and the fix is technology that gives Convex a view of the underlying portfolio and, in time, gives coverholders something back; if it lands, Howat said the group would look to write more of that business.
Ariel Re attacked the same problem from the segment rather than the pipeline. Ryan Mather, its chief executive, pointed to the continuing migration of business from the admitted market into excess and surplus, a shift he said the company identified several years ago and met with products new to the market. What Ariel gets back is twofold: catastrophe risk for its investors and data that sharpens the models behind its underwriting — the segment is doing a job for the investor base, not just diversifying the group.
Conduit Re's Neil Eckert argued from the infrastructure side, saying newer balance sheets are not carrying the systems their older peers do. Conduit has built a data lake and is prototyping systems that Eckert said came together faster than traditional builds allow, with the payoff running to operating efficiency and to the quality of pricing data. The point about legacy systems is fair, but the moat is thinner than that phrase suggests: a data lake is a starting position, and whatever edge it buys is measured in implementation time rather than in underwriting judgment.
ICD's records have Willis Re naming Conduit cofounder Eckert as its chief executive on September 9, five days before this Monte Carlo dispatch was published with Eckert identified at Conduit.
The discipline test in a softening market has moved from price to attachment points, and delegated authority opens a second front that gets less scrutiny. If that share of a book is written by third parties whose data arrives late and in the wrong shape, the reinsurer's picture of its own aggregate is only as good as its ingestion, and a group that cannot see its delegated portfolio cannot price its retentions against it. Convex is spending on exactly that and says it will write more when the fix lands; Ariel is buying data alongside risk through E&S. The binding constraint on the next slice of growth in this channel is operational, and the test arrives with the next local loss that pierces a retention inside a delegated book.