Document scanning: how many dpi you need and when OCR matters

It is the most frequent question and the answer is not "as high as possible". A contract scanned at 1,200 dpi is no more legible than at 200: it weighs twenty times more, opens slowly and will not fit in an email.

The scale we use at the counter:

• 200 dpi — text documents: contracts, invoices, course packs, correspondence. The standard for reading on screen, archiving and attaching to an email. A hundred-page PDF at 200 dpi in black and white weighs a few megabytes.
• 300 dpi — documents with photographs or charts, and anything that might be reprinted. If the file will be used to make a paper copy again, 300 dpi is the minimum for the difference not to show.
• 600 dpi — photographs, negatives, vintage prints, valuable originals to preserve. Here the extra detail genuinely shows, and the file size is a price worth paying.
• Above 600 — only for art reproduction or when a detail must be enlarged considerably. Outside those cases it is waste.

Colour, greyscale or black and white. Black text on white paper in bitonal weighs very little and reads perfectly; greyscale matters when photographs or stamps would otherwise be lost; colour when colour carries information — a blue signature that must be distinguishable from a photocopy, a chart, a highlighter. Choosing black and white "to save space" on a document with a blue ink signature is the mistake that forces a redo.

A plain scan is a photograph of the page: you can see it, but the computer does not know there is text on it. You cannot search a word, you cannot copy a sentence.

OCR (optical character recognition) adds the recognised text beneath the image: the page looks identical to the original but becomes searchable and copyable. Across a three-hundred-page archive that is the difference between finding a name in two seconds and leafing through everything.

When it genuinely matters: archives to be consulted, contracts where you need to find a clause, course packs to quote from, documents heading into a management system, and anything that must be accessible to screen-reader users.

Where OCR fails, and it is worth knowing in advance:

• Handwriting — it does not read it, barring lucky exceptions. A form filled in by hand stays an image.
• Stamps and signatures over text — they confuse recognition exactly on the lines that usually matter most.
• Skewed or faded scans — OCR is sensitive to angle: an original placed straight beats any later correction.
• Complex tables — it captures the text but not the structure: columns can come out interleaved.
• Old typefaces or other alphabets — to be checked case by case.

Rule of thumb: OCR does not replace checking. On a document that matters — a deed, an expert report — the recognised text should be reread, because one wrong character inside a number is invisible and changes everything.

A plain PDF is fine for exchanging a document, attaching it to an email, sending it to a colleague.

PDF/A is a PDF with stricter rules, designed for long-term archiving: it embeds everything needed to reproduce the document — fonts included — and forbids anything that might stop working over time, such as dynamic content or links to external resources. A PDF/A opened in twenty years still looks as it does today.

When PDF/A is required: filings with public bodies, planning applications, deeds, compliant archiving, and theses at many universities. In those cases it is not a preference but a requirement, and a plain PDF is rejected by the receiving system.

If you are unsure whether your PDF really complies, you can check it free on our site — there is a dedicated tool — or bring it in and we will check together. It is a thirty-second check that saves discovering the problem at the deadline.

One note on size. A two-hundred-page colour PDF at 300 dpi can exceed 100 MB and will not go through email. If the file is only for reading it can be reduced; if it is for reprinting, it cannot. Tell us when you order: that choice is made before scanning, not after.

Not everything goes into an automatic feeder, and forcing it is the quickest way to ruin an original.

Bound volumes are scanned open on the flatbed. The consequence to accept: the area near the gutter comes out slightly curved and darker — that is physics, not a machine defect. If the book can be sacrificed, taking the binding apart gives a perfect and much faster scan; if it cannot, we work page by page and it takes time.

Drawings and floor plans up to A0 are scanned on the plotter: cost is calculated as for printing, € 5 per metre. The same care as technical printing applies: if the drawing will be measured, the scale must be stated and verified.

Fragile or old documents — thinned paper, sheets with aged tape, curling photographs — should be looked at together first. Sometimes the answer is not the feeder but the flatbed, with more time and more care.

Stapled or spiral-bound originals must be taken apart to use the feeder, and that is your decision. We do not remove staples or spirals without asking, because an original cannot always be put back as it was.

On confidentiality: documents stay in the shop only for the duration of the job, the file goes to you and we do not keep it longer than necessary. If there is sensitive data — medical records, deeds, identity documents — say so, and we will handle it accordingly.