Image Attachments in Reinsurance Submissions: Read or Rekey?
On this page
- Key Takeaways
- What Counts as an Image Attachment in a Reinsurance Submission?
- Should Reinsurers Rekey, OCR or Ask for the Original Image File?
- What Resolution Does an Image Need Before OCR Can Read Treaty Figures?
- Which Image Errors Corrupt Treaty Figures After Capture?
- How Do Control Totals Prove an Image Was Captured Correctly?
- How Should Reinsurers Handle Image Attachments in Submissions?
- How Does Insurnest Treat Image Attachments in a Submission?
- Key Terms
- Sources
- Frequently Asked Questions
Last reviewed: November 2026
The five-year loss list arrived as four phone photos of a printed page. The bottom rows curved away from the lens and the last column was cut off.
An analyst rekeyed 42 claims by hand that evening. Nobody compared the typed total with the printed one.
Image attachments in reinsurance submissions are photos, scans, screenshots or faxes of schedules and loss lists, sent as pictures.
One misread digit in a loss list can move the burning cost a treaty is priced on.
This guide gives you a method matrix by image type, an image error list with tests, resolution rules and a control-total check.
Key Takeaways
- Choose the method per image. A screenshot and a phone photo need different capture routes.
- Resolution decides OCR quality. Aim for about 300 dpi before trusting any OCR output.
- Totals catch what eyes miss. Reconcile every captured column to the printed total.
- Ask for the source file early. Long lists are cheaper to re-request than to rekey.
What Counts as an Image Attachment in a Reinsurance Submission?
An image attachment in a reinsurance submission is any schedule, loss list or slip page sent as a picture file instead of text or cells.
Common formats are JPEG, PNG, TIFF and phone formats such as HEIF. Scanned PDFs behave the same way, because each page is only a picture.
Why do cedents still send photographed schedules to reinsurers?
Cedents still send photographed schedules to reinsurers when the data sits in a printed report, a legacy system screen or a signed paper slip.
Small cedents and branch offices often photograph pages rather than export them. Brokers then forward the photos unchanged.
Should Reinsurers Rekey, OCR or Ask for the Original Image File?
Reinsurers should choose rekey, OCR or a re-request by image type, row count and how the figures will be used.
Use this matrix at intake.
| Image type | Typical problems | OCR fit | Rekey fit | Best route |
|---|---|---|---|---|
| Flatbed scan of a printed table | Faint print, slight skew | Good if about 300 dpi | Slow for long tables | OCR, then control totals |
| Screenshot of a system or spreadsheet | Columns cropped outside the window | Good, text is sharp | Fine for short lists | OCR; ask for an export if columns are missing |
| Phone photo of a printed page | Curved lines, shadow, perspective | Weak | Fine up to about 30 rows | Rekey with double entry, or re-request |
| Fax or re-faxed copy | Black-and-white, broken characters | Weak | Fine for short pages | Re-request the source file |
| Photo of handwritten amendments | Ink over print, initials | Poor | Only route | Rekey and confirm with the broker in writing |
What decision rule picks the capture method for a submission image?
The decision rule for a submission image starts with use: figures that feed pricing need the strictest route.
- Pricing data over about 30 rows. Ask for the original spreadsheet or PDF first.
- Clean scan or screenshot. Run OCR, then reconcile control totals.
- Short photo or fax. Rekey twice, by two people, and compare.
- Handwriting anywhere. Rekey, then confirm every amended figure in writing.
The 30-row line is a working threshold, not a standard. Set your own from the time your team spends rekeying.
What Resolution Does an Image Need Before OCR Can Read Treaty Figures?
An image needs about 300 dpi before OCR can read treaty figures reliably.
Tesseract's documentation says it works best on images of at least 300 dpi. An A4 page at 300 dpi is about 2,480 by 3,508 pixels.
Microsoft's Document Intelligence layout model accepts images from 50 by 50 to 10,000 by 10,000 pixels.
It needs text at least 12 pixels high on a 1024 by 768 image, about 8-point text at 150 dpi.
Small print in a photo of a whole page often fails that test.
Tesseract's guide also advises rotating pages so text lines are horizontal and keeping a border of about 10 pixels around the text.
Which Image Errors Corrupt Treaty Figures After Capture?
The image errors that corrupt treaty figures after capture are predictable, so test for each one by name.
| Error | What you see after capture | Test |
|---|---|---|
| Skew or curved page | Values slip into the next column | Row count per column differs |
| Cropped columns | A whole column missing, or cut digits | Compare column headers with the cedent's template |
| 1 read as 7 | Amount too high by 6 in one digit place | Difference is 6 x 10, 100, 1,000 and so on |
| O read as 0, l read as 1 | Policy or claim references change | Reference fails your format check |
| Decimal comma | 1.250,00 read as 1.25 or 125,000 | Sum differs by a factor of 100 or 1,000 |
| Handwritten amendment | Printed figure kept, ink ignored | Visual review of every inked mark |
| Two digits swapped | 18,460 captured as 18,640 | Difference divisible by 9 |
Treat any amount that changes treaty attachment or limit use as a pricing item. Those rows get a second pair of eyes regardless of method.
Mixed files make this worse; see multi-attachment reinsurance submissions for sorting them first.
How Do Control Totals Prove an Image Was Captured Correctly?
Control totals prove an image was captured correctly when captured counts and sums match the figures printed on the image.
Check four numbers: rows, each amount column, the grand total and any subtotal per year.
What does the rule of nine tell you about a capture difference?
The rule of nine says a capture difference divisible by 9 often points to two swapped digits or a slipped decimal point.
It narrows the search; it does not prove the cause.
What does one illustrative photographed loss list look like after capture?
One illustrative photographed loss list shows how the tests locate errors without a full rekey.
| Check | Printed on image | Captured | Difference | Likely cause |
|---|---|---|---|---|
| Rows | 42 | 42 | 0 | None |
| Paid total (EUR) | 18,640,500 | 18,646,500 | +6,000 | 1 read as 7 in the thousands place |
| Outstanding total (EUR) | 4,212,800 | 4,212,080 | -720 | 720 / 9 = 80, so 800 likely keyed as 080 |
| Incurred total (EUR) | 22,853,300 | 22,858,580 | +5,280 | Sum of the two errors above |
The analyst fixed two rows, not 42. The rest of the list is then accepted against the printed totals.
Feed the corrected list into your loss triangle data preparation only after this check passes.
How Should Reinsurers Handle Image Attachments in Submissions?
Reinsurers should handle image attachments in submissions with one fixed routine, run before any figure reaches pricing.
- Class each image. Scan, screenshot, phone photo, fax or handwriting.
- Pick the route. Use the matrix and decision rule above.
- Check resolution. Below about 300 dpi, re-request rather than OCR.
- Capture. OCR or double-key, keeping the image beside the output.
- Reconcile. Rows, columns and totals must match the print.
- Log the method. Record who captured it, how and when.
Images that cannot be read go on your incomplete reinsurance submissions list with a dated request.
How Does Insurnest Treat Image Attachments in a Submission?
Reinsurance Email Intelligence starts from the submission email and reads its body plus attached PDFs, spreadsheets and other supporting files.
It pulls cedent, broker, treaty structure, premium, limits and claims history into one submission view and shows gaps.
Its product page does not claim OCR of photos or faxes. Test it on your own image attachments before relying on it.
It does not replace the control totals above or the underwriter's judgement.
For related email-bound data, see treaty terms trapped in email and PDF attachments.
Key Terms
- OCR: Optical character recognition, software that turns a picture of text into characters.
- DPI: Dots per inch, the scan density; 300 dpi is the usual floor for reliable OCR.
- Control total: A printed count or sum used to prove captured data is complete and accurate.
Sources
- Tesseract documentation: Improving the quality of the output
- Microsoft Learn: Document Intelligence layout model
- Insurnest: Reinsurance Email Intelligence
Frequently Asked Questions
What counts as an image attachment in a reinsurance submission?
An image attachment in a reinsurance submission is any schedule, loss list or slip page sent as a picture file, such as a phone photo, scan, screenshot or fax image.
Should reinsurers rekey, OCR or ask for the original image file?
Reinsurers should OCR clean scans and screenshots, rekey short photographed lists with double entry, and ask for the original file when a loss list or schedule exceeds about 30 rows.
What resolution does an image need before OCR can read treaty figures?
An image needs about 300 dpi before OCR can read treaty figures reliably. Tesseract's documentation says it works best at 300 dpi or more, which is 2,480 by 3,508 pixels for A4.
Which image errors corrupt treaty figures after capture?
The image errors that corrupt treaty figures after capture are skew, cropped columns, 1 read as 7, O read as 0, decimal commas and handwritten amendments that OCR ignores.
How do control totals prove an image was captured correctly?
Control totals prove an image was captured correctly when the captured row count, column sums and grand total match the figures printed on the image to the last digit.
What does the rule of nine tell you about a capture difference?
The rule of nine says a capture difference divisible by 9 often signals two swapped digits or a slipped decimal point, so check those rows first before rekeying everything.
How should reinsurers handle image attachments in submissions?
Reinsurers should handle image attachments in submissions by classing each image, choosing rekey, OCR or re-request, checking resolution, capturing, then reconciling control totals before pricing.
How does Insurnest treat image attachments in a submission?
Insurnest Reinsurance Email Intelligence reads submission emails with PDFs, spreadsheets and other supporting files. Its page makes no image OCR claim, so test it on your own image files.

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 →