Scanning gives you images. Digitization gives you information.

Two quotes can arrive for what looks like the same job and differ substantially, and the reason is almost never that one supplier is greedy. It is that one priced images and the other priced information. Read side by side, the difference is obvious. Read in isolation, it is invisible.

Why the cheaper option is often the expensive one

Suppose a finance team scans eight years of supplier invoices. The images are clean and the files are named by box. Six months later somebody needs every invoice from one supplier for a dispute. Nobody can produce it without opening files one at a time, so a temp is hired for three weeks. That cost was not saved. It was deferred and inflated.

Retro-fitting index data to an existing image archive is harder than indexing at the point of capture, because the person doing it can no longer see the physical file, the folder it sat in or the order it came in. Context that was free during preparation has to be reconstructed from the image alone.

Where a file starts and ends is the quietest decision in the project

Splitting is the part of digitization nobody raises in a sales meeting and everybody argues about after delivery. A run of images has to be cut into documents somewhere, and the cut point decides what a search result looks like. Cut per box and every search returns a single enormous PDF that someone scrolls through. Cut per sheet and a twelve-page tenancy contract arrives as twelve unrelated results with no relationship between them. Cut per document and the archive behaves the way people expect it to.

The rule has to come from you, because it depends on what a document means in your business. An invoice with a delivery note and a signed approval behind it might be one document or three, and both answers can be defended. What cannot be defended is leaving it unstated and meeting the supplier's default at handover. Re-splitting an archive after the fact means reprocessing it, and the index data has to be reattached along the way.

Scanning alone is genuinely enough when

• The driver is floor space and the records are effectively dormant.
• You need a disaster-recovery copy of documents you already retrieve by physical reference.
• A retention clock is running out and the records will be destroyed within a short, known window.
• The documents already carry a printed unique reference that can be read into the file name automatically.

You need full digitization when

• More than one team needs the same document, or people request files by attribute rather than by location.
• The records feed a process: approvals, claims, onboarding, tenancy renewals, payroll queries.
• You will be asked to produce a complete, evidenced set of documents for an audit or an inspection.
• The archive contains mixed Arabic and English records that staff need to search across.
• The content will feed a downstream system, a report or any automated extraction.

How to tell which one you were quoted

Vendor proposals use the word digitization loosely. These five questions separate them quickly, and the answers should be in the proposal rather than in a phone call.

1. 1How is a document defined, and what rule splits one file from the next? If the answer is one file per box, you are buying scanning.
2. 2Which index fields are captured, and how many per document? Silence here means none.
3. 3Is OCR applied to every page, and is the text layer embedded in the delivered file or supplied separately?
4. 4How is index accuracy measured, on what sample, and what happens when a field fails? A process with no verification step has no accuracy.
5. 5What is the delivery format and does it include a manifest my system can import without manual mapping?

The middle option nobody mentions

You do not have to treat the whole archive the same way. Splitting it is usually the sharpest commercial decision available. Index the active, audited and frequently requested categories properly. Scan the rest to image only, with a light index that at least records box, category and date range. If a dormant category later turns out to matter, it can be re-indexed as a targeted piece of work rather than as a second full project.