PDF Tables in Reinsurance Submissions: Make the Totals Tie
On this page
- Key Takeaways
- Why Do PDF Tables in Reinsurance Submissions Break Across Pages?
- What Is the Reassembly Control for a Multi-Page PDF Loss Table?
- How Do Subtotals and Carried-Forward Lines Get Double Counted?
- How Do You Detect Repeated Headers and Split Rows in Extracted PDF Tables?
- What Do PDF Extraction Tools Do With Treaty Tables That Span Pages?
- How Should Underwriting Teams Extract PDF Tables in Reinsurance Submissions?
- How Does Insurnest Read PDF Tables in Submission Emails?
- Key Terms
- Sources
- Frequently Asked Questions
Last reviewed: November 2026
The cedent's five-year loss table ran to three PDF pages. Extracted into a spreadsheet, it showed EUR 32.85m incurred.
The PDF's own grand total said EUR 15.35m. The difference was two "Brought forward" lines counted as claims.
PDF tables in reinsurance submissions are loss, exposure and premium tables inside PDF files, and they break wherever the page breaks.
Unchecked, one reassembly error can double the loss ratio a treaty is quoted on.
This guide gives you a reassembly control formula, a worked loss table with right and wrong totals, a sensitivity table and detection formulas.
Key Takeaways
- A PDF has no table object. Extraction rebuilds the table from text positions.
- Page lines are not claims. Subtotals, carried-forward and brought-forward lines must go.
- Tie to printed totals. Every page subtotal and the grand total must match exactly.
- Count rows too. Split rows leave sums unchanged but add phantom claims.
Why Do PDF Tables in Reinsurance Submissions Break Across Pages?
PDF tables in reinsurance submissions break across pages because a PDF stores text at positions on a page, not rows and columns.
Report writers then add page furniture at every break. Extraction tools see that furniture as more rows.
Which page artefacts get mistaken for data rows in extracted PDF tables?
The page artefacts mistaken for data rows in extracted PDF tables are the same few lines on almost every cedent report.
- Repeated header. Column names reprinted at the top of each page.
- Page subtotal. The sum of rows on that page only.
- Carried forward (c/f). The running total at the foot of a page.
- Brought forward (b/f). The same running total, repeated at the top of the next page.
- Split row. One claim whose description wraps onto the next page.
- Rotated page. A landscape page that extracts as scrambled columns, or not at all.
What Is the Reassembly Control for a Multi-Page PDF Loss Table?
The reassembly control for a multi-page PDF loss table proves that data rows alone produce every printed total.
REASSEMBLY CONTROL (per amount column)
1. Page test : sum(data rows on page p) = printed subtotal p
2. Carry test : b/f on page p = c/f on page p-1
= subtotal 1 + ... + subtotal (p-1)
3. Grand test : sum(all data rows) = printed grand total
4. Count test : data rows = claims listed (or last row number)
5. Tolerance : 0.00, unless the cedent prints rounded units
Run it on paid, outstanding and incurred separately. A pass on incurred alone can hide offsetting errors.
How Do Subtotals and Carried-Forward Lines Get Double Counted?
Subtotals and carried-forward lines get double counted when a cleaning filter removes lines labelled "total" but keeps the rest.
In the illustrative table below, c/f lines read "Total c/f", so the filter removed them. The "Brought forward" lines stayed in.
| Page | Data rows (EUR) | Printed subtotal | c/f printed | b/f printed |
|---|---|---|---|---|
| 1 | 6,200,000 | 6,200,000 | 6,200,000 | n/a |
| 2 | 5,100,000 | 5,100,000 | 11,300,000 | 6,200,000 |
| 3 | 4,050,000 | 4,050,000 | n/a | 11,300,000 |
| Grand total | 15,350,000 | 15,350,000 |
The right method sums data rows only: 6.2m + 5.1m + 4.05m = EUR 15.35m. It ties to the printed grand total.
The wrong method also sums the b/f lines: 15.35m + 6.2m + 11.3m = EUR 32.85m. Every claim before page 3 is counted twice or more.
How much can a kept brought-forward line move a treaty loss ratio?
A kept brought-forward line can move a treaty loss ratio by more than any rate change under discussion.
Assume, for illustration, five-year subject premium of EUR 60m.
| Extraction error | Incurred used (EUR m) | Loss ratio | Change |
|---|---|---|---|
| None (data rows only) | 15.35 | 25.6% | Base |
| b/f lines kept | 32.85 | 54.8% | +29.2 points |
| Page subtotals kept | 26.65 | 44.4% | +18.8 points |
| Rotated page 3 dropped | 11.30 | 18.8% | -6.8 points |
| Split row on page 1 | 15.35 | 25.6% | Sum unchanged; row count +1 |
The last line matters for frequency. A phantom claim with no amount still changes claim counts and averages.
Clean figures then flow into loss triangle data preparation and your large loss register.
How Do You Detect Repeated Headers and Split Rows in Extracted PDF Tables?
Repeated headers and split rows in extracted PDF tables can be flagged with simple formulas before any sum is taken.
Assume claim reference in column A, description in B and incurred in F, with a page number in column H.
Repeated header : =A2="Claim ref"
Subtotal / c/f : =OR(ISNUMBER(SEARCH("total",B2)),
ISNUMBER(SEARCH("c/f",B2)))
Brought forward : =OR(ISNUMBER(SEARCH("b/f",B2)),
ISNUMBER(SEARCH("brought",B2)))
Split row : =AND(A2="",B2<>"",F2="")
Page test : =SUMIFS(F:F,H:H,1,I:I,"data")
Column I holds the line type you assign: data, header, subtotal, c/f, b/f or total. Only "data" rows feed any figure.
What should you do with rotated landscape pages in a PDF submission?
Rotated landscape pages in a PDF submission should be detected by page count before anyone trusts the output.
Compare pages in the PDF with pages that produced rows. Microsoft's Document Intelligence layout model reports an angle for each page.
What Do PDF Extraction Tools Do With Treaty Tables That Span Pages?
PDF extraction tools often merge treaty tables that span pages, but merging is not cleaning.
The Power Query PDF connector has a MultiPageTables option, which defaults to true.
It "controls whether similar tables on consecutive pages will be automatically combined into a single table".
The same page notes that multi-line rows may not be identified. It suggests Table.FillDown or Table.Group to repair them.
Microsoft's layout guidance suggests splitting multi-page tables by page, then post-processing them into one table.
Either way, removing subtotal and b/f lines remains your check.
How Should Underwriting Teams Extract PDF Tables in Reinsurance Submissions?
Underwriting teams should extract PDF tables in reinsurance submissions page by page, with the page number kept on every row.
- Ask for the source first. A spreadsheet beats any PDF; see Excel submission data extraction.
- Extract with page numbers. One page at a time, or tag each row's page.
- Tag line types. Data, header, subtotal, c/f, b/f, total, footnote.
- Merge split rows. Join continuation text to the row above.
- Run the control. Page, carry, grand and count tests on each amount column.
- Record the result. Note the tie-out and file it with the submission.
Tables typed into the email itself need a different routine; see email body table extraction for reinsurers.
How Does Insurnest Read PDF Tables in Submission Emails?
Reinsurance Email Intelligence reads the submission email and its attached PDFs, spreadsheets and other supporting files.
It extracts treaty structure, premium, limits, commissions and claims history into one structured submission view, and flags missing information.
It does not replace the reassembly control or the underwriter's judgement. Test it on your own multi-page cedent reports.
Key Terms
- Carried forward (c/f): The running total printed at the foot of a page.
- Brought forward (b/f): The same running total repeated at the head of the next page.
- Page subtotal: The sum of data rows on one page only.
Sources
- Microsoft Learn: Power Query PDF connector
- Microsoft Learn: Document Intelligence layout model
- Insurnest: Reinsurance Email Intelligence
Frequently Asked Questions
Why do PDF tables in reinsurance submissions break across pages?
PDF tables in reinsurance submissions break across pages because a PDF stores positioned text, not a table. Each page break adds headers, subtotals and carried-forward lines that extraction reads as rows.
Which page artefacts get mistaken for data rows in extracted PDF tables?
The page artefacts mistaken for data rows in extracted PDF tables are repeated headers, page subtotals, carried-forward and brought-forward lines, split rows, footnotes and rotated landscape pages.
What is the reassembly control for a multi-page PDF loss table?
The reassembly control for a multi-page PDF loss table says the sum of data rows only must equal each printed page subtotal and the printed grand total, with zero tolerance.
How do subtotals and carried-forward lines get double counted?
Subtotals and carried-forward lines get double counted when a filter removes rows labelled total but keeps brought-forward lines, which repeat earlier pages' sums inside the amount column.
How much can a kept brought-forward line move a treaty loss ratio?
A kept brought-forward line can move a treaty loss ratio a long way. In the illustrative three-page table, it lifts incurred from EUR 15.35m to 32.85m and the ratio from 25.6% to 54.8%.
What do PDF extraction tools do with treaty tables that span pages?
PDF extraction tools often merge treaty tables that span pages. Power Query's MultiPageTables option combines similar tables on consecutive pages by default, but subtotal lines still need removing.
How should underwriting teams extract PDF tables in reinsurance submissions?
Underwriting teams should extract PDF tables in reinsurance submissions page by page, tag each line type, keep data rows only, then tie them to page subtotals, carried-forward lines and the grand total.
How does Insurnest read PDF tables in submission emails?
Insurnest Reinsurance Email Intelligence reads PDFs attached to submission emails and extracts items such as claims history into one view. Teams still run the page-total reassembly control.

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 →