A fact lands in the file. Your standard says review the reserve. Does the file show anyone did?
Incurra reads the adjuster notes on every open claim. A demand lands with a number on it. Specials climb. Treatment escalates. It catches those facts and their dates, then lists the files where nothing shows anybody looked.
It never says the reserve is wrong. A reserve that hasn’t moved is not a defect, and arguing valuation with the person who set it is an argument you lose. Your own standard asks the narrower question. Did the review happen, and is it in the file?
The output
Your auditor already publishes this format. I didn’t invent it.
A claims auditor’s per-file worksheet, printed in a public board packet, runs columns headed 90 Day Review and Reserve Change next to the claim identifiers. Six of the seven below are the auditor’s own. The last is mine: the fact that started the clock, and when it landed.
| Claim number | Insured | Claimant | DOI | 90 Day Review | Reserve Change | Trigger, from the note |
|---|---|---|---|---|---|---|
| CA-24-118337 | Marfield Transport LLC | R. Delgado | 2024-03-11 | Not evidenced | None since 2024-09-14 | Demand received, 2025-06-02 |
Carried reserve that day: $75,000, last changed 2024-09-14.
Reserve review recorded since: none found.
Example only. Synthetic claim, synthetic insured, synthetic note. Incurra has never processed a real claim file.
Every line is checkable against a document you already hold. Nothing here asks you to trust a model.
Three practitioners told me this month that an attorney letter neither does nor should move a reserve on its own. Nothing new has arrived to price until the records do. They’re right, and that’s why the trigger above is a document with a number in it.
The standard
The benchmark is published. It isn’t my opinion of what a claim is worth.
A carrier can argue a valuation forever. It can’t argue whether it followed its own procedure. PRISM’s claims standards, amended 1 July 2026 and binding on its general liability members:
“Reserves shall be evaluated for adequacy at least every ninety (90) days. All reserve changes shall be documented … providing explanation of the reason for the reserve change or notation that the current reserve is adequate and why.”
That last clause is the one that matters. It wants something written down even when the number doesn’t move. ICRMA and AIMS start the same clock on the fact instead of the calendar, though neither asks for a notation when the reserve holds. Only PRISM does, and your counsel would find that, so it may as well come from me.
All three bind their own members by contract. None is a carrier standard, and no property and casualty regulation requires a documented reserve review at all.
The evidence
How often it fails, counted by somebody who isn’t me.
| What failed, and out of how many | Count | Auditor |
|---|---|---|
| No reserve review inside the required window | 25 of 175 | Marsh, 2024 |
| No supervisory review of reserves inside the required window | 41 of 172 | Marsh, 2024 |
| Reserves that needed revising and never got it | 6 of 97 | North Bay, 2018 |
A fourth audit revalued the files instead of scoring process, and found 23 of 173 case reserves that “failed to reflect probable outcome,” 22 of them under-reserved, worth $729,667 across a $16.3M book. That is the only dollar figure anyone has put on this.
How were the files picked? The 2024 rows say “randomly selected,” verbatim. The 2018 row doesn’t. Its auditor calls the draw “a carefully selected and structured sample rather than a random sample,” 48% of it in litigation, so it runs high by an amount nobody has sized. None of this is a rate for your book. It is workers’ compensation and general liability at public entity pools, because pools publish their audits and carriers don’t.
And rows one and two are the year those metrics failed. The next year they failed on 6 of 175 and 2 of 174, after a board saw the numbers. That is the best argument on this page against buying anything from me. Below is where I answer it.
Two objections
“We already do this in our claims audit.”
Probably you do, and if it reads every open file every quarter on this one question, I have nothing to sell you. But AGRiP sets the standard at an audit every three years, and CAJPA’s two-year clock binds only its Excellence tier. At a lot of pools nobody looks for two to three years.
And it doesn’t hold in between. At one program, two metrics with identical wording were scored, reported to a board, and came back worse the next year: 11 of 14 down to 6 of 15, and 11 of 19 down to 10 of 28. Both on a census, not a sample. What improved there was one person working their own diary. What slipped needed two people to talk and write it down. One program, fourteen files. A hypothesis to test against your own history, not a law.
“You’re generating a document plaintiff’s counsel will want.”
Your counsel decides that, not me. The output gives the fact, its date, the note it came from, and whether the file records a review. It never says a reserve was wrong. The Tenth Circuit looked at a reserve that didn’t match the file and declined to call it bad faith, putting it down to “lack of care” rather than “any bad-faith attempt.” No. 16-6336, decided 2 July 2018. One circuit, its own facts, and I am not a lawyer. What I can promise is narrower. Nothing here treats a reserve gap as evidence of bad faith.
On the record
A federal judge has already written this one up.
A case reserve was set at $8,000 in June 2006, “prior to receiving any medical records.” A year later a demand package arrived with a 13% impairment rating and $10,210.10 in specials. The court’s next sentence: “After receiving Plaintiff’s demand package, Defendant continued to maintain its reserve of $8,000.”
It moved to $25,000 nineteen months in, when the adjuster asked her manager to look. The judgment was $1,696,826 against a $50,000 policy. One more person reading the file was the only thing that caught it. Doing that reading on every open file, on a schedule, is the product. M.D. Fla., docket 6:17-cv-02026, order of 23 June 2019.
Fit
Who it is for, and who it isn’t for.
For third party administrators running liability and trucking programs, and the pools that hire them. You are already graded on this by an independent firm, in a report that goes to a board in open session. Also fronting carriers and program administrators holding reserve risk on files somebody else handles.
Not for you if one person can approve every reserve change. A public entity running sixty claims a year told me its risk manager approves every one, and that its peers that size do the same. Below a certain caseload one person is the review. Same answer if reviews happen out loud and never get written down, because I can’t see those and neither can your auditor.
Who I am, and what is true today
I’m Evan Maus. I finish a dual degree in economics and data science at Berkeley in December 2026, and I didn’t come up through insurance. Every practitioner quoted here, I got by asking.
No customers. Not incorporated. No SOC 2. No carrier data has ever been processed. One person, and it’s me.
Auditors have measured how often this gap happens. Nobody has measured what it costs, and until someone does the financial case is an argument rather than evidence. Measuring it takes a book of closed claims.
One question
I’m not asking for a meeting. I’m asking for a story.
Nothing to sign up for, no demo to book. One question is worth more to me than a call:
When a reserve on a quiet file finally moves a lot, what set it off, and how long had that fact been sitting there?
A few lines is plenty. And if you think the premise is wrong, that is more useful to me than agreement.