There are facts in your claim files that your reserves don’t reflect.
- Mar 4
- Adjuster note — “Received LOR from Kessler & Wolfe.”
- Mar 4
- Structured attorney flag — none
- Jun 4
- Case reserve set — $85,000
- Oct 22
- Case reserve — $85,000, unchanged 140 days
Illustrative. Not a real claim.
Incurra reconciles your open commercial-auto claims against what is already documented in them, and shows you the ones where the file and the number disagree.
Who it’s for
Commercial-auto carriers, program administrators and fronting carriers.
The tier that carries the reserve risk without an in-house data science team — and that the enterprise vendors cannot serve profitably.
If you write personal lines or workers’ compensation, this isn’t for you yet.
How it works
- You send a nightly export. Claims, reserves, payments, adjuster notes. A flat file out of Guidewire, Duck Creek, Sapiens — or a spreadsheet. No API project, no integration timeline.
- We reconcile the file against the reserve. Representation, suit, escalating treatment, venue — read out of the notes and checked against what the structured data and the reserve actually say.
- You get a ranked list. Every flag cites the note that produced it, with its date, beside the reserve that didn’t move. Ranked by how far comparable claims in your own book went on to develop.
What it isn’t
We don’t set your ultimates and we don’t touch your IBNR.
Those are yours, and they are an actuary’s job.
We don’t predict what a claim will settle for. We reconcile what your file already says against what your reserve says — so every flag is checkable against a document you already hold, rather than against a model you would have to take on faith.
There is no pooled data consortium. Your data trains nothing that leaves your book without a signed agreement saying so.
Who I am
I’m Evan Maus. I built Incurra, and I’ll be the person you deal with.
I’m finishing a dual B.A. in Economics and Data Science at Berkeley in December 2026, and I didn’t come up through insurance — you’d find that out anyway, so I’d rather say it.
What I have done is build systems where fooling yourself with data is the main failure mode: a systematic trading engine running real money on a survivorship-bias-free backtest of 66,753 trades, judged on fresh out-of-sample windows rather than on the fit. Reserving punishes the same mistake. That part I understand.
Talk to me