This loss run would be a pain to extract manually. Dense tables, inconsistent formatting, and abbreviations that vary by carrier—just looking at it gives you a headache. And this is just one example. There are hundreds if not thousands of variations depending on the insurer, region, and even the department that issued it.
For brokers, underwriters, and commercial insurance teams, manually pulling data from loss run reports is one of those annoying but necessary tasks. It takes time, requires eagle-eyed attention, and can still result in errors.
Why Are Loss Runs So Difficult to Parse?
Loss run reports are meant to provide a clear record of a policyholder’s claims history—but in practice, they’re anything but standardized. Some are exported from outdated legacy systems, others are scanned documents with inconsistent formatting. A single report might include multiple policies, various carriers, and several years of data—all bundled into one file.
The essential data—claim numbers, loss dates, reserve amounts, total incurred, and claim status—is often buried in cluttered layouts, surrounded by inconsistent headers, footnotes, and varied terminology.
On top of the usual complexities that come with parsing PDFs, loss runs present their own unique challenges:
- Tables often aren’t real tables. What looks like a simple row may actually contain multiple embedded data points, or even nested structures where one "cell" holds several fields.
- The number of claims per report is unpredictable—it could be zero, five, or fifty. An effective extraction process needs to dynamically adapt to capture every claim.
- Some loss runs include claims from several policies held with the same carrier. In these cases, ensuring each claim is correctly matched to its policy number is critical for accurate analysis.
Enter Parsie.pro: AI That Understands Insurance Documents
Parsie.pro is built to transform messy, complex insurance documents—like loss runs—into clean, structured data in just minutes. Whether it’s a PDF, a scanned image, or a multi-page report, you simply upload the file and define what you need using a schema. Parsie does the rest.
With a schema-driven approach, you tell Parsie what to look for—for example, defining a claims field as a list of objects to account for zero or multiple claims in a report. Parsie then:
- Detects tables regardless of layout or formatting
- Identifies and extracts key fields like claim number, loss date, total incurred, paid to date, and claim status
- Handles complex formatting such as nested tables, merged cells, and multi-page layouts
- Delivers clean structured data in CSV, Excel, or JSON—ready to integrate with your systems via API
No more manual entry. No more cleaning up broken tables. Just accurate, schema-aligned data you can trust.
Below is an example of the structured JSON you’ll get back—fully aligned with your defined schema and ready to flow directly into your pipeline:
{
"row": {
"doldcm": "7/23/2007",
"coverage": "Excess Automobile Liability",
"loss_city": "ROSWELL",
"recovered": "$0.00",
"loss_state": "GEORGIA",
"closed_date": "02/25/2008",
"losses_paid": "$0.00",
"open_claims": "0",
"claim_number": "ATL27201",
"loss_reserve": "$0.00",
"claims_status": "CLOSED",
"expenses_paid": "$0.00",
"reported_date": "2/11/2008",
"policy_numbers": "",
"total_incurred": "$0.00",
"expense_reserve": "$0.00",
"loss_description": "MVA - Insd Driver was making a right turn, OV passed insured veh on the right at a high rate of speed, hit his brakes, lost control then hit the drive.",
"number_of_claims": 1,
"policy_terma_date_range": "4/1/2007 - 4/1/2008"
}
}
Built with the Insurance Workflow in Mind
Unlike generic OCR tools, Parsie understands real insurance documents. That means:
- It knows what a loss run looks like
- It understands insurance-specific fields
- It learns and improves over time based on feedback
- It’s not just for loss runs—COIs, ACORDs, policy dec pages, and more are supported
Try Parsie.pro
Whether you're a solo broker or part of a large MGA, Parsie.pro offers flexible plans to help you automate document extraction—starting with your most painful process: loss runs.
Ready to see it in action? Book a demo with us today.