How to Read a Loss Run Report and Automate Extraction

To read a loss run report, start with the valuation date, confirm the policy periods, then work through each claim's status, paid amount, reserve, and incurred total. A loss run is a carrier's report of claims on a policy, and underwriters treat it as the clearest evidence of how an account has actually performed. Reading one well means checking that the numbers are current, that the math adds up, and what is still open.

Reading one loss run is quick. Reading forty from a dozen carriers, each in its own layout, is a different job. This guide covers both: what every field on a loss run means and what to check, then how teams automate loss run extraction so underwriters and account managers start from clean, structured claims data.

Key takeaways

• The valuation date tells you how current the figures are. Paid and reserve amounts change after that date.
• Incurred equals paid plus reserve. Check the math on every row and confirm how the carrier treats expenses and recoveries.
• Open and reopened claims can still grow, so they deserve more attention than closed ones.
• Totals rows are check figures, not claims. Counting them as claims inflates frequency.
• Automated extraction works when it validates rows against the document's own totals and keeps a person in the review loop.

What a loss run report is

A loss run report is a claims history that an insurance carrier issues for a policyholder. It is a periodic report of claim information that insurers provide to their insureds. According to The Hartford, it shows the date and description of each claim, whether it's open or closed, settlement costs, and reserves for open claims.

Brokers usually request loss runs from the incumbent carrier when they market or renew an account, often with the ACORD 611 Claims History / Loss Run Request form. Several states require carriers to supply them on request. Florida sets a 15-day deadline, New York 10 days, and Illinois 30 days. Our broker submission prep checklist covers how many years to request and where state law sets the look-back period. For where loss runs sit in a full submission, see our explainer on commercial insurance submission documents.

How to read a loss run report, field by field

Carriers lay out loss runs differently, but most carry the same core fields. Read the header first, then the claim rows, then the totals.

‍

‍

Five checks to run on every loss run

1. Is it current?

The valuation date is the cutoff for the paid and reserve figures on the report. After that date, open claims keep moving, a process IRMI calls loss development. Big "I" submission guidance asks for currently valued loss runs for the past five years, and a run valued a year ago can understate an account with large open claims.

Development is not a small effect in liability lines. Insurers reported $7.3 billion of adverse development in other liability (occurrence) during 2025, the largest of any line, according to S&P Global Market Intelligence data reported by Carrier Management. AM Best estimates that commercial auto liability is under-reserved by $4 billion to $5 billion.

2. Does the math add up?

Paid losses plus outstanding losses equal incurred losses, so every claim row should reconcile. Two details change how you read the totals. First, some carriers fold allocated loss adjustment expense (ALAE) into paid and incurred, while others show it in a separate column. Second, recoveries from subrogation and salvage may be shown net or gross. A Casualty Actuarial Society study note on estimating unpaid claims also warns that some third-party administrator reports use "incurred" to mean the outstanding reserve only. When the numbers look off, check the column definitions before you question the claims.

3. Is the problem frequency or severity?

Frequency is the likelihood that losses will occur, and severity is how much damage they cause. A run of small, repeating claims suggests an operational issue the insured can fix, such as a loading dock or a driver. One or two large losses may be outliers or a sign of exposure the price doesn't reflect. Severity matters more in liability lines than it used to. A Casualty Actuarial Society and Triple-I analysis estimates that legal system abuse and related litigation trends added $231.6 billion to $281.2 billion to US liability insurance losses over the decade to year-end 2024.

4. What is still open?

Open and reopened claims carry reserves that can still rise. Look closely at the reserves on recent open claims, since those figures are the carrier's current estimate and will change. A loss run also leaves out claims the carrier doesn't know about yet: incurred losses ordinarily exclude incurred but not reported (IBNR) losses. Big "I" guidance suggests asking the insured whether any claims are pending but not yet on the loss run.

5. How does it compare to premium?

The NAIC defines loss ratio as incurred losses as a percentage of earned premium. Many loss runs don't show premium, so pull it from the policy or the carrier and calculate the ratio by policy year. For workers' compensation, also compare the loss run with the experience modification factor. NCCI builds the mod from unit statistical reports, usually covering the latest three years, so the two documents draw on the same claims and should tell the same story.

Why loss runs are hard to read at scale

Loss runs come from hundreds of different sources, and no two look alike. When FurtherAI's engineers built loss run extraction, they found documents from one page to more than 200, with around 30 fields to capture per claim. In one file, the policy number appeared once as a section header and applied to about 100 claims below it. Blank cells meant "same as the row above." Summary rows looked exactly like claim rows, and in one document they inflated the claim count by 15%.

Most of that work still lands on people. Underwriters in commercial and personal lines spend 41% to 43% of their time on administrative activities, according to Capgemini. Accenture's 2025 underwriting research found that only 9% of the underwriting executives surveyed say their organizations use intelligent ingestion to extract submission data to a large extent today, and 68% expect to within three years. Loss runs are a good example of why: as our guide to mixed-format submissions explains, scanned loss runs are among the worst offenders because they aren't machine-readable.

How to automate loss run extraction

Automating loss runs means turning each report, whatever its layout, into the same structured claims table, then checking it before anyone relies on it. Most teams follow five steps.

‍

‍

The validation step decides whether automation can be trusted. In FurtherAI's testing on difficult real-world loss runs, giving the extraction agent a tool to check its output against each document's own totals raised row count accuracy from 80% to 95%. The gain came from catching missed claims and misread totals rows, which a simple extraction pass returns without any warning.

Normalization matters most when one account has several carriers or administrators. Our guide to consolidating TPA loss runs covers the schema side, and for property accounts, loss run data still has to be reconciled with the SOV and inspection reports, as our guide to multi-source risk data explains. Loss runs usually arrive alongside ACORD applications, so the same pipeline often handles ACORD form data extraction too.

What this looks like in practice

At Leavitt Group, one of the largest independent insurance brokerages, a single loss run of more than 130 lines of unstructured data used to take hours of manual work. The carrier couldn't supply it in a sortable format, and attempts with general AI tools didn't produce usable results. Processed through FurtherAI, the same file came back structured and ready for analysis in minutes.

On the underwriting side, one of the largest MGAs in the US used FurtherAI for commercial property submission intake, with loss run data extracted, summarized, and combined with the SOV and ACORD data for each submission. The average time to clear a submission fell from about 32 minutes to about one, and the team saved more than 2,000 hours of manual effort in three months. To see how FurtherAI handles the full workflow, from ingesting any carrier format to producing broker-ready summaries, visit our page on loss run processing automation.

Frequently asked questions

‍

DISCLAIMER 

This article is for general informational purposes only and does not constitute legal, regulatory, compliance, underwriting, or other professional advice. The content reflects information available as of the date of publication, and FurtherAI undertakes no obligation to update it as laws, regulations, or AI technologies evolve.