Email Body Table Extraction for Reinsurers: Error Catalogue
On this page
- Key Takeaways
- Why Do Tables Pasted Into Treaty Emails Break When They Are Extracted?
- Which Errors Appear in Email Body Tables, and How Are They Detected?
- How Does a Loss Table in 000s Get Misread by 1,000 Times?
- Which Email Table Errors Should a Reinsurer Fix First?
- How Can You Test Last Year's Broker Emails for Table Extraction Errors?
- How Does Insurnest Help With Email Body Table Extraction for Reinsurers?
- Key Terms
- Sources
- Frequently Asked Questions
Last reviewed: November 2026
One thousand times. That was the gap between the large loss the cedent reported and the figure in the reinsurer's loss file.
The broker had pasted the table into the email body. Its header said "USD (000s)", and nothing downstream read the header.
Email body table extraction for reinsurers is capturing the pasted tables in treaty emails, such as loss and exposure lists.
Done well, every figure keeps its unit, currency and row.
This guide gives you an error catalogue with detection rules and fixes, a triage order, a test plan and a worked 1,000× example.
Key Takeaways
- Read the header first. Units and currency often appear once, above every figure they govern.
- Know the sender's locale. A full stop can be a decimal point or a thousands separator.
- Drop totals before summing. Subtotal rows double-count if they stay in the data.
- Ask for the source file. A table pasted as an image is a picture, not data.
Why Do Tables Pasted Into Treaty Emails Break When They Are Extracted?
Tables pasted into treaty emails break because the email keeps the look of the table, not its meaning.
In HTML, one cell can span several rows or columns. The HTML standard defines this with colspan and rowspan.
When a reply goes as plain text, spans and borders collapse into spaces.
Attachments are covered in excel submission data extraction.
Which Errors Appear in Email Body Tables, and How Are They Detected?
Errors in pasted tables fall into four families: structure, scale, number format and content. Each has a detection rule.
| Error | How it shows up | Detection rule | Fix |
|---|---|---|---|
| Merged cells | Class or year label only on the first row of a block | Blank key cell followed by filled value cells | Fill the label down before loading |
| Wrapped rows | One claim split over two lines | Row with a description but no amount | Join to the row above, then recheck the count |
| Units in the header | "(000s)" or "USD m" above the figures | Header text contains 000, k, m or mn | Apply the multiplier to every value in the column |
| Currency only in the header | "EUR" once at the top | No currency code in any value cell | Store currency as a field on every row |
| Separator confusion | 1.234.567 or 1,5 | Two or more full stops, or a comma before two digits | Parse with the sender's locale |
| Totals and subtotals | "Total" or "Sub-total" rows | Row label matches total; value equals sum above | Remove, then use as a control total |
| Footnote markers | 4,250* or 3,100(a) | Non-numeric character attached to a number | Strip the marker; keep the footnote as a note |
| Bracket negatives | (1,250) for a recovery | Value wrapped in brackets | Load as negative |
| Table pasted as an image | Screenshot in the body | No text in the table area | Request the spreadsheet |
How do merged cells and wrapped rows distort a broker loss table?
Merged cells and wrapped rows distort a broker loss table by breaking the one-row-per-record shape that loading needs.
A merged accident year label leaves the next four rows without a year. A wrapped description creates an extra row with no amount.
Check row counts against the broker's stated claim count. A mismatch is usually one of these two.
How do units and currency in a table header change every figure?
Units and currency in a table header change every figure, because they appear once and govern the whole column.
Miss "(000s)" and every value is 1,000 times too small. Miss "m" and it is a million times too small.
Currency in the header is lost when rows are copied out. Write the currency onto each row at extraction.
How are thousands and decimal separators misread in pasted broker tables?
Thousands and decimal separators are misread when the sender's locale differs from the reader's.
1.234.567 in a German-style table is 1,234,567 in a UK one. 1,5 is one and a half, not fifteen.
Microsoft's Excel guidance notes that Excel uses the system separators set in regional settings by default.
So the same paste can parse differently on two desks. A value like 1.234 is ambiguous; resolve it from other rows.
How should totals, footnotes and image tables be handled in treaty emails?
Totals, footnotes and image tables in treaty emails should be separated from the data before anything is summed.
Keep the broker's total as a control figure. If your sum of rows differs, a row was dropped or a subtotal was kept.
A footnote such as "* includes ALAE" changes what the number means. Store it against the row.
How Does a Loss Table in 000s Get Misread by 1,000 Times?
A loss table in 000s gets misread by 1,000 times when the header unit is dropped and each figure is read as whole currency.
Take an illustrative motor excess of loss layer of USD 5m xs USD 5m. The broker pastes this large loss list into the email.
| Claim | Gross incurred, USD (000s) | Read correctly | To the layer |
|---|---|---|---|
| 2021-014 | 6,200 | USD 6.20m | USD 1.20m |
| 2022-031 | 4,800 | USD 4.80m | Nil |
| 2023-007 | 9,750 | USD 9.75m | USD 4.75m |
| 2024-022 | 5,300 | USD 5.30m | USD 0.30m |
| Total | 26,050 | USD 26.05m | USD 6.25m |
Read correctly, the layer took USD 6.25m over five years, a burning cost of USD 1.25m a year.
Read without the header, the largest claim is USD 9,750. Nothing reaches the USD 5m retention, and the layer looks loss-free.
The email said the list covers claims above USD 2.5m.
A list where every value sits below that threshold fails the detection rule.
On a GNPI of USD 50m, USD 1.25m is a 2.5% burning cost rate. See burning cost reinsurance pricing.
Which Email Table Errors Should a Reinsurer Fix First?
The email table errors a reinsurer should fix first are scale errors, because units and currency move every figure at once.
- Units and currency. One missed header can shift a whole table by 1,000 times.
- Separators. These shift single values, often by 1,000 times as well.
- Totals. A kept subtotal doubles a block of losses.
- Structure. Merged and wrapped rows misassign years and claims.
- Footnotes and images. These change meaning or block extraction entirely.
Checked figures then belong in the large loss register.
How Can You Test Last Year's Broker Emails for Table Extraction Errors?
You can test last year's broker emails for table extraction errors by re-extracting a sample and comparing it with what was priced.
- Pick 20 renewal emails from last year that had tables in the body.
- Re-extract each table, applying the rules above.
- Compare each column total with the broker's stated total.
- Compare large losses with the figures in the pricing file.
- Log each difference by error type and broker.
- Send the brokers with repeat errors a request for spreadsheets.
Loss tables feeding development work need the same checks as loss triangle data preparation.
How Does Insurnest Help With Email Body Table Extraction for Reinsurers?
Insurnest helps with email body table extraction for reinsurers through Reinsurance Email Intelligence.
It reads the email body and attachments, and extracts claims history, loss ratios, exposure, premium and limits.
It standardises formats into one structured submission view and flags missing information. Fields depend on what the broker sent.
We do not claim it reads tables pasted as images. Checking that units and currency were applied stays with the underwriter.
See also multilingual reinsurance submissions for tables written in other languages.
Key Terms
- Header unit: A multiplier such as (000s) or m stated once above a column of figures.
- Locale: Regional settings that decide which characters separate thousands and decimals.
- Control total: A broker's stated total, kept to check that the extracted rows add up.
Sources
- WHATWG: HTML Living Standard, tables
- Microsoft Support: Change the character used to separate thousands or decimals
- Insurnest: Reinsurance Email Intelligence
Frequently Asked Questions
Why do tables pasted into treaty emails break when they are extracted?
Tables pasted into treaty emails break because the email keeps the look of the table, not its meaning. Merged cells, header units and local number formats get lost.
How do merged cells and wrapped rows distort a broker loss table?
Merged cells and wrapped rows distort a broker loss table by leaving blank cells under a shared label or splitting one record over two lines.
How do units and currency in a table header change every figure?
Units and currency in a table header change every figure because they sit once at the top. Miss (000s) and each value is 1,000 times too small.
How are thousands and decimal separators misread in pasted broker tables?
Thousands and decimal separators are misread when the sender's locale differs from yours. 1.234.567 in a German-style table is 1,234,567 in a UK one.
How does a loss table in 000s get misread by 1,000 times?
A loss table in 000s gets misread by 1,000 times when the header unit is dropped. A USD 9.75m loss becomes USD 9,750 and falls below the retention.
Which email table errors should a reinsurer fix first?
The email table errors a reinsurer should fix first are scale errors: units and currency. They move every figure. Separators, totals and structure come next.
How can you test last year's broker emails for table extraction errors?
You can test last year's broker emails for table extraction errors by re-extracting a sample and comparing totals with the figures the underwriter actually priced on.
How does Insurnest help with email body table extraction for reinsurers?
Insurnest helps with email body table extraction for reinsurers by reading the email body and attachments into one standardised submission view that flags missing information.

Hitul Mistry
CEO, Insurnest
An InsurTech leader with more than a decade of experience across insurance and technology, focused on solving business problems with the help of technology. Has worked with brokers, insurance carriers, and reinsurance firms across the India, UAE, and US markets.
View LinkedIn profile →