Reinsurance pricing platform: what to look for

Speed at the moment it matters

The test of a pricing platform is not the annual report; it is the phone call with the broker. Can an underwriter change a layer or an attachment and see the new price, expected loss and marginal PML before the broker has finished the sentence? Anything measured in minutes rather than seconds pushes pricing back into spreadsheets, where the platform's value evaporates.

Live portfolio impact, not batch roll-up

Every deal should be priced against the book as it stands: marginal 1-in-100 PML, capacity used, return on capital, by zone and peril, updating as terms change. Platforms that roll up overnight tell you what you did yesterday. Deal selection rules — appetite encoded by zone, peril, counterparty and return threshold — should screen submissions automatically, and a what-if scenario generator should let you stress the portfolio before you commit.

Analytics you can actually see

Exceedance curves gross and net of retro, exposure maps with peak zones, event-level portfolio analysis, cat bond tracking where relevant. These should be interactive and customisable, and the underlying data should be yours to query directly, not locked behind the vendor's reports.

Headless by design

The most important question in 2026: does the platform run without its own user interface? A pricing engine, loss models and roll-up exposed through an API can sit under your own underwriting workbench, a low-code tool, Power BI, or a Foundry Workshop app. Your underwriters keep the screens they designed; the platform does the mathematics. A platform that insists on its own UI will fight every integration you attempt.

An AI assistant that can do the work, not just describe it

The new benchmark is an MCP server. With one, any MCP-capable assistant — Claude, ChatGPT, Copilot or an agent you build — can create and renew deals, change terms, run analyses, roll up the portfolio and produce the report, as the signed-in user with a full audit trail. Ask vendors whether their AI assistant can take actions or only answer questions about the help pages.

Ownership and customisation

Can you add your own risk metrics, fields, screens and loss distributions? Can you export everything? Does it support custom add-ins in a real language such as C# or Python? Treefrog's Foundation platform was built around these answers — headless, with an MCP server, live portfolio impact and a database you can query — because we spent years building bespoke pricing systems for Bermuda reinsurers and ILS managers before we built a product.

Frequently asked