The industry has spent billions capturing claims data: extraction, fraud scores, estimates, core platforms. Almost none of it captures the thing that actually decides cost — why a decision was made. That reasoning lives in the adjuster’s head and a two-line note, so it is inconsistent, unauditable, and gone the moment they leave. It is the most valuable data in the claim, and it is the data nobody records.
Walk the claims tech stack and you find a tool for every artifact in the file. Documents get extracted. Fraud gets scored. Damage gets estimated. The system of record stores what was decided. None of them store why. And the why is the part that decides cost — it is what makes a decision consistent, auditable, and defensible. Capturing everything except the reasoning is how an industry can spend billions on claims data and still leak money it can’t see.
Every tool captures the data — none captures the why
The stack keeps every artifact and the outcome. It drops the reasoning in between.
A file that tells the story of the reasoning — what was considered, what was ruled out, and why the call landed where it did — is the exception. Most of the time the reasoning was never written down at all. It sits one layer above everything the stack records, and it is the layer that pays:
| What every tool captures | The one thing none of them keep |
|---|---|
| Extraction — the documents in the file | Why a document changed the call |
| Fraud scoring — the risk flags | Why the adjuster trusted or overrode them |
| Estimating — the repair numbers | Why the number landed where it did |
| System of record — what was decided | Why it was decided that way |
The industry pays roughly 96 cents of every premium dollar on claims and expenses, so the underwriting margin is thin — and leakage runs straight through it. On most books it lands at five to ten percent of claims paid, much of it from decisions that were defensible at the time but never documented well enough to defend later. Estimate it on yours:
- Liability set a touch generous
- Subrogation never pursued
- Reserves corrected late
- The same file decided differently by different people
What moves the number is consistency — the same well-reasoned call whoever works the file. Which of these is costing you, and how much, is the specific breakdown a mapping call shows you.
Map exactly where your book leaks →A 30-minute mapping call · no data from you to start.The 1-10-100 rule, applied to reasoning
Capture the why at the decision, or pay a multiple to recover it later — or the most to lose it.
Data-quality practice prices this cleanly: fixing an error at entry costs about a dollar, correcting it later about ten, and working with bad data about a hundred. Reasoning follows the same curve. Capture the why at the moment of the decision and it costs almost nothing. Reconstruct it during an audit and it costs more. Lose it and pay downstream — in leakage and avoidable litigation — and it costs the most.
The adjuster is the integration layer
Every system hands over a fragment. One human holds the whole thing together — and then leaves.
This is the structural reason the why goes uncaptured. Every tool in the stack produces a fragment — a document, a score, an estimate — and the adjuster assembles those fragments in their head and issues a decision. When the integration layer is a person, the integration walks out when the person does. It also explains why efficiency drives in claims so often land as more files rather than better ones: the system was optimised to feed the human faster, not to keep what the human knew.
The layer that finally captures the why
Mysa records the reasoning behind every decision — the part nobody else keeps, and the part that compounds.
Every other tool in the claims stack stops at the data. Mysa captures the reasoning behind it — the rule a decision rests on, the facts that mattered, the confidence in the call — attached to the decision and accumulated across the whole book. That decision corpus is the part nobody else has, and the only part that compounds: a new adjuster can learn from it, an audit can trust it, and a pattern across a thousand files becomes visible in it. The most valuable information in a claim was never in a document — it was in the adjuster’s head. The entire point is to get it out before it walks out the door.
The hidden cost of claims is not a missing tool. It is a missing layer — the reasoning that every other system throws away. Capture it, and the cost stops being hidden.
Sources
- Industry estimates on claims leakage — consistently 5–10% of claims paid; conservative internal tracking of 2–4% versus 20–30% documented in rigorous audits. Regure; The Lab Consulting.
- Insurance Information Institute / industry results — 2024 U.S. P&C combined ratio ~96% (about 96 cents of losses and expenses per premium dollar).
- CAS Loss Reserve Seminar (CLRS) 2013 — claims-leakage data-mining: overpayment concentrates among lower-performing adjusters, who overpaid each claim by over half its expected loss.
Common questions
What do claims systems fail to capture?
They capture the artifacts and the outcome — documents, fraud scores, estimates, and what was decided — but not why the decision was made. That reasoning stays in the adjuster’s head, in hallway conversations, and in short case-note entries, so it is inconsistent and disappears when the adjuster leaves.
How much does claims leakage cost?
Industry estimates put leakage at 5–10% of claims paid, and rigorous audits often find 20–30% — well above the 2–4% most insurers track. At an industry combined ratio near 96%, even a few points of leakage can decide whether a book is profitable.
What is the 1-10-100 rule in claims?
It is a data-quality principle: fixing an error at entry costs about $1, correcting it later about $10, and working with bad data about $100. Applied to claims reasoning, capturing the why at the moment of the decision is cheap, while losing it and paying later in leakage and litigation is by far the most expensive path.
Tiago is Co-Founder and COO of Mysa, where he works with claims teams on how liability, subrogation, and leakage decisions actually get made — and how to keep the reasoning behind them from walking out the door.