OCR scanning: the Archivogo process
OCR scanning brings page images, recognised text and organisation into one agreed archive process. A useful archive begins with the way people need to retrieve information. This guide connects scanning, search and organisation questions so a project can be scoped around real records and agreed deliverables.
On this page
Ocr scanning: agree the archive boundary
A document scanning process begins with deciding what enters the project and what stays outside it. Identify document categories, intended outputs and handling constraints. This prevents assumptions about sensitive material, original papers or future access from becoming part of the work without discussion. A clear boundary makes later quality checks more meaningful.
An OCR scanning pilot should have a clear decision attached to it. Decide whether the sample is intended to assess readability, recognised text, file grouping or several separate checks. Record the result of each check rather than using one overall impression. A representative example may expose an issue that does not appear on another document type. That difference is useful information for defining the process; it should not be hidden to make the pilot look uniformly successful.
- Define included categories.
- List exclusions and special handling.
- Agree the intended output.
Use a representative pilot
A pilot batch can reveal issues that a single clean page does not show. Choose a selection that represents the archive, including differences in condition and layout. Agree the checks that would make the pilot acceptable. The purpose is to refine the process before a larger handover, not to claim a universal result from a small sample.
- Include representative formats.
- Agree visual and text checks.
- Record exceptions before expanding the batch.
For example, a search by a known reference and a search by a phrase inside a letter can require different preparation. These are illustrative retrieval tasks, not descriptions of an existing client archive.
Capture, recognise and index
Scanning, OCR and organisation are distinct steps. Capture produces the page image, recognition supports text search and indexing helps a user locate the right record. Each step should contribute to an agreed retrieval need. Combining them without a defined output can produce files that are technically searchable but difficult to use in practice.
For OCR scanning, acceptance criteria can distinguish an image problem from a recognition problem. An upright, complete page can still contain uncertain recognised text, while a correct-looking text layer may not reveal a missing attachment. Choose the important checks around the intended use of the records. If a critical field needs review, define that field and its exception handling explicitly. The process should make it possible to understand what was checked without promising an exhaustive review that was never agreed.
- Check capture completeness.
- Review recognition limitations.
- Apply agreed identifiers and metadata.
Review against agreed criteria
Quality review should use the criteria established for the project. Completeness, orientation, readability and selected recognised fields may require separate checks. The level of review must be explicit rather than implied by the word digitisation. Record the issues that remain so recipients understand the archive they are receiving.
- Use the agreed checklist.
- Distinguish different quality checks.
- Make unresolved exceptions visible.
For the review stage, keep a record of the checks performed and the exceptions that need a decision. A file that opens successfully may still contain a missing page or an uncertain field. The quality record should explain those limits to the recipient rather than hide them.
Organise before handover
Folders and file names should make sense to the people who will retrieve the records. A consistent structure can reduce ambiguity, but complexity is not a sign of quality by itself. Agree how duplicates, versions and unusual documents are represented. Handover should include enough explanation for recipients to use the structure correctly.
An OCR scanning process becomes easier to manage when exceptions have an agreed route. Describe how an incomplete record, an unreadable area or an uncertain identifier is flagged and who decides what happens next. That decision may affect the output or the remaining scope. The record should preserve the distinction between a resolved issue and a known limitation. At handover, recipients should be able to locate those exceptions instead of discovering them only when a later search fails. For each exception, distinguish a technical correction from a decision about the source record. The process should make clear whether the output can be adjusted or whether missing information requires a choice by the responsible project party.
- Keep names consistent.
- Explain version handling.
- Provide a simple structure overview.
Agree recipients and transfer
The project should identify who is permitted to receive the deliverables and how the transfer will occur. Access, retention and treatment of original papers require explicit decisions. This page does not establish a certification or a blanket retention policy. Those details need to match the actual records and the agreed engagement.
An OCR scanning acceptance note should describe the actual checks performed. Keep an unresolved recognition issue distinguishable from a page-image correction.
- Identify authorised recipients.
- Agree an appropriate transfer route.
- Decide what happens to originals and retained copies.
Check the retrieval result
A useful final check is whether the proposed archive supports the questions it was designed to answer. Try the agreed search scenarios and confirm that the expected records can be located. Any limitations should remain part of the handover record. The process is complete only to the extent defined in its agreed scope.
- Test the agreed searches.
- Review any unresolved exceptions.
- Confirm the deliverables match the scope.
Frequently asked questions
Does OCR make every character accurate?
No. Text recognition can produce mistakes, especially with difficult originals. The scope should identify the fields and records that need review and how exceptions will be handled.
Should I email the entire archive first?
No. Begin with the document categories, approximate volume and desired retrieval use. Any transfer of records should follow an agreed scope and handling arrangement.
How are output and retention decided?
Formats, recipients, treatment of original papers and retention expectations are project decisions. They need to be agreed for the actual archive rather than assumed from a general service page.
Further reading
The National Archives digitisation resource page connects quality-management and record-handling guidance. Its U.S. federal scope is important context when using it to inform questions about an agreed scanning process.