Is this ordinary OCR?

It uses character recognition, but goes further. Plain OCR returns a wall of text; here the page is also interpreted, so a quantity lands in the quantity column rather than in a paragraph.

Will free OCR tools do the same job?

They will read the characters. What they generally won't do is rebuild the line-item table, separate it from the letterhead and totals, or keep numbers as numbers.

Does it handle carbon copies and thermal paper?

Usually, yes — faint duplicate copies are common in the trade. Very washed-out print can come back with gaps, so check those against the original.

What about a docket photographed at an angle?

Moderate skew and uneven lighting are tolerated. Filling the frame with the page and keeping your shadow off the table gives noticeably cleaner results.

What file types are accepted?

PDFs and common image formats up to 15MB per file, including multi-page PDFs, which come back as one continuous table.

OCR

Delivery docket OCR

Dockets are about the worst input an OCR engine can be handed: printed on cheap paper, signed over, folded into a pocket and photographed in a van. Reading the characters is the easy half.

Try docket OCR free

Encrypted files
Protected in transit and at rest
Private to your account
Never used to train AI models
Payments by Stripe
We never store your card number
Test before you buy
Check accuracy on your own docket

Security, privacy & accuracy

Where plain OCR stops

Character recognition gives you the words on the page in roughly the order they appear. That is fine for a letter and close to useless for a table — you get product codes, descriptions and quantities interleaved as one stream, with the letterhead and the terms mixed in.

No sense of which text belongs to which column
Quantities and prices returned as text, so totals fail
Signature boxes and small print land in the middle of your data
A second page starts over as if it were a new document

Reading the page, not just the characters

The page is interpreted as a document: this band is the header, this is the line-item table, this row is a total, this scribble is a signature. Because the structure is worked out from meaning rather than from ruled lines, a crooked scan with faded table borders still produces clean columns.

Columns identified by what the values are, not by pixel position
Wrapped descriptions kept with their own line item
Multi-page dockets joined into one continuous table
Numbers typed as numbers so the sheet calculates

Capture tips that improve any OCR

Lay the docket flat and fill the frame with the page. Shoot straight down rather than at an angle, and keep your own shadow off the table area. For very faint carbon copies, a flatbed scan at higher contrast beats a phone photo every time.

What comes back

An Excel workbook with a Data sheet of line items and a Summary sheet naming the supplier, docket number and date — emailed to you and kept in your dashboard, so the original scan no longer needs to be findable.

Frequently asked questions

Is this ordinary OCR?: It uses character recognition, but goes further. Plain OCR returns a wall of text; here the page is also interpreted, so a quantity lands in the quantity column rather than in a paragraph.
Will free OCR tools do the same job?: They will read the characters. What they generally won't do is rebuild the line-item table, separate it from the letterhead and totals, or keep numbers as numbers.
Does it handle carbon copies and thermal paper?: Usually, yes — faint duplicate copies are common in the trade. Very washed-out print can come back with gaps, so check those against the original.
What about a docket photographed at an angle?: Moderate skew and uneven lighting are tolerated. Filling the frame with the page and keeping your shadow off the table gives noticeably cleaner results.
What file types are accepted?: PDFs and common image formats up to 15MB per file, including multi-page PDFs, which come back as one continuous table.

Stop re-typing paperwork

Check accuracy on your own test document free within 7 days. No card needed, no automatic charge.

Get started