The workbench
Triangle in.
Drafted memo out.
Sixty seconds.
Paste a loss triangle — or start with the bundled benchmark — and the engine runs chain-ladder, Bornhuetter–Ferguson, Mack, and a 2,000-path bootstrap, then drafts your reserving memo. Change an assumption; watch every downstream number move.
Your data stays yours — pasted triangles are parsed right here in your browser and never uploaded. Check us against the textbook — the bundled sample is the industry-standard Taylor–Ashe (1983) benchmark, so you can verify our results against the published literature before you trust us with anything.
| metric | value |
|---|
Method references: Mack (1993), Distribution-free calculation of the standard error of chain ladder reserve estimates, ASTIN Bulletin 23(2) · England & Verrall (2002), Stochastic claims reserving in general insurance · Bornhuetter & Ferguson (1972). The engine's results on the bundled triangle match Mack's published totals; see Method & Governance.
SAVE YOUR WORK
Create a free workspace to keep your books, build a verified run history, and export drafted memos as PDF. Takes thirty seconds; your data policy is one paragraph long.
From workbench to platform
What you just used, at carrier scale
Same engine, your data
In the platform, intake agents assemble and reconcile triangles from your claims data — with lineage recorded at every join — instead of a paste box. platform · in development
Continuous, not on-demand
Actual-vs-expected drift runs as data lands, not when you open a page. The quarter becomes a checkpoint, not a discovery. platform · in development
Sign-off workflow
The drafted memo routes to a credentialed actuary. The sign-off line stays blank until a human fills it — that is the product's constitution, not a feature flag.