Document Digitisation

Document Scanning Quality Checks for Australian Buyers

By Pexatech8 min readUpdated

Before accepting a delivered scanning batch, check that the files correspond to the records supplied, that the images and requested text or data are usable, and that corrections can be traced to the delivery version your team reviewed. A folder that opens successfully is a starting point, not a completed acceptance review.

This document scanning quality checklist is a suggested buyer review method to agree with your supplier. It is not a legal standard or a statement of Pexatech's operating procedures. Use it alongside your written scope, delivery requirements and organisation's records policies.

1. Identify exactly what arrived

Give the delivery a reference and record its receipt date and version. Keep the supplied package identifiable while reviewing it; avoid mixing replacement files into an unlabelled folder that everyone already uses.

Compare the delivery register with your handover inventory. If no register was agreed or supplied, request enough information to map the delivered files back to the boxes, folders or documents supplied. Useful fields to discuss include the source identifier, delivered filename, document boundary, image count and reported exceptions.

Separate three lists: items expected, items received and items still unexplained. An item might be outstanding, intentionally excluded, combined with another file or named differently. Establish which explanation applies rather than treating every mismatch as a missing scan.

If the original inventory only counted boxes, it cannot establish an exact expected image count. Record that limitation. The document preparation guide explains how to establish an inventory before handover; at delivery, the task is to reconcile against the evidence actually available.

2. Check completeness beyond the total page count

A matching count can conceal an omission. In a fictional six-page folder, repeating page three while omitting page four still produces six images. The total balances, but the record is incomplete.

Use the agreed document boundaries to inspect sequence and continuity. Does the next image follow the previous one? Is an attachment with its parent document? Has a divider become a separate file when it should identify the following group? Check whether a large sheet scanned separately has a clear reference from its original position.

Do not assume a gap in printed page numbers proves a scanning error: the source may already have been incomplete. Compare with the original or a reliable handover record, and log an unresolved discrepancy when that comparison is unavailable.

Ask how blank sides, covers and duplicate originals were treated under the agreed scope. Two visually similar pages may be intentional copies with different annotations. A duplicate-removal decision needs that context, not just a matching filename or count.

3. Separate source defects from capture defects

Open representative files at a readable size and inspect the information your team needs. Compare unclear areas with the source where possible. A faint mark already missing from a photocopy is a different problem from a clear mark lost during scanning.

Describe the observation precisely: “right edge cuts through the final characters” is more useful than “bad quality”. Record the file and image position, what is affected, and whether the same feature is visible on the original. If access to the original is pending, say so instead of assigning a cause.

For NSW public offices, State Records NSW's digitisation specifications connect capture choices to intended use, legibility and essential detail, including small characters and punctuation. This supports comparing the details that matter with the source; it does not establish one universal scanning resolution for every Australian business.

Agree what happens next: request a new image for a capture defect, investigate a possible source limitation, or ask the records owner to assess an unresolved exception. Do not quietly replace ambiguous text with a guess.

4. Test the image and searchable text separately

A page can look readable while its recognised text contains an error. Adobe's Acrobat guidance on correcting scanned text describes reviewing uncertain OCR words against their displayed image. That is a product example of the distinction, not a requirement to buy Acrobat or a statement about Pexatech's software.

For searchable PDFs, choose references visible on reviewed pages and try finding them. Copy a short passage or important reference into a plain-text field and compare it with the image. Note whether a failure concerns an image, its text layer or the application's search behaviour.

A successful search for one familiar word does not establish that every field is correct. If structured data was requested, trace selected values to their source pages and check them after import into the intended system. Keep the review tied to the required task rather than a general “OCR passed” label.

The output selection guide explains the differences between searchable PDFs, an index and extracted fields. For a format demonstration, the fictional PDF and CSV sample was authored directly, not produced by production OCR; it is not a client result or a service-quality benchmark.

If PDF/A is specified, use the PDF versus PDF/A guide to request format-validation evidence for the delivered files while keeping image, completeness and OCR checks separate.

5. Agree who reviews which material

Assign responsibilities before the review becomes a collection of unconnected comments. A records coordinator might reconcile deliveries, a person familiar with the documents might assess ambiguous content, and the destination-system owner might test imports. Name the person who can resolve disagreements and approve the batch.

Choose review coverage according to the material and consequences of an error. Discuss which checks need every relevant item examined, which can use sampling, and what findings would trigger a wider review. Include ordinary material as well as known difficult document types; neither an easy-only selection nor a difficult-only selection describes the whole batch.

State Records NSW's quality-assurance guidance, written for NSW public offices, discusses documented benchmarks, fault reporting, corrections and the limits of sampling. Its framework is useful context; this buyer checklist does not adopt a fixed sampling rate or claim that a successful sample proves an entire delivery is error-free.

Record the items actually examined and the checks performed. “Sample reviewed” without identifiers leaves the next reviewer unable to tell whether a reported fault was inside or outside that sample.

6. Try the real destination and a real retrieval task

Use the delivered version in the application where staff will work, with an appropriate test arrangement agreed by its owner. A preview in the supplier's delivery portal is not the same as a successful import or retrieval in your system.

Ask a user to find a record using information they would normally know, open it and locate the relevant page or value. Observe where the task breaks: the filename, index mapping, search, permissions or file itself. If files work before import but fail afterwards, involve the destination-system owner before assuming rescanning is the answer.

Record the tested application and import settings with the result. Check that source references still lead to the correct file after any renaming or transfer. Use the approved handling arrangements for the review; do not send sensitive test files to an unapproved service for convenience.

7. Keep one actionable issue log

Use stable issue references and precise locations so corrections can be discussed without circulating unnecessary document contents. Agree a suitable channel for any sensitive supporting evidence.

The following log is entirely fictional. Its folder IDs, issues and suggested next steps were authored to illustrate a review process. They are not client records, measured results or completed Pexatech checks.

Fictional location Issue to investigate Requested action Recheck status
FOLDER-A / image 4 Image appears to repeat image 3; expected sequence unclear Compare with the original folder and explain or replace the affected image Awaiting source comparison
FOLDER-B / image 2 Right edge cuts through a reference visible on the original Supply a corrected image and identify the replacement file version Awaiting replacement
FOLDER-C / image 7 Reference is legible in the image but cannot be found by text search Check recognised text, then repeat the search in the intended viewer Recheck not yet performed

For your project, add the person responsible, date raised, supplier response and decision evidence. Keep “replacement received” separate from “rechecked”: receiving a new file does not show that it resolves the issue.

8. Reconcile corrections before closing the batch

Ask for a correction list identifying what changed and which files or index rows were replaced. Agree whether the next delivery is a complete replacement batch or only corrected items. Without that distinction, staff can accidentally retain conflicting versions or overlook an unchanged file that still needs review.

Recheck the reported issue, then check its immediate context. Replacing a page could affect order, attachments, searchable text or a linked index entry. If the correction changes an identifier, update the mapping and repeat the relevant retrieval or import check.

Close each issue with a reason and the version reviewed. If an exception remains, record its effect on the intended use and who reviewed it. Escalate recurring faults under the agreed review plan rather than closing them individually while assuming the rest of the batch is unaffected.

A delivery acceptance record you can copy

Adapt this record to your agreement and approval process. These labels describe review decisions, not contractual wording or a certification of the archive.

Batch and delivery version: [reference, receipt date, replacement version if applicable]
Scope assessed against: [quote/specification version and agreed output]
Reconciliation: [inventory/register references; outstanding discrepancies]
Review coverage: [reviewers, item identifiers, checks and limitations]
Destination test: [application, task, settings and result]
Corrections: [issue-log reference, replacements received and recheck evidence]
Decision: [accepted for the stated scope / awaiting correction / reviewed exceptions requiring a named decision]
Remaining actions: [owner, action and agreed next review]
Approval: [authorised reviewer and date]
Originals: [separate retention, return or disposal decision; responsible person]

Do not merge digital-output acceptance with permission to destroy originals. The State Records NSW disposal checklist addresses quality control alongside retention and authorised disposal under GA51 for NSW public offices. It is not a disposal permission for all Australian businesses. Confirm the requirements for your organisation and records separately.

If the review method was never specified, discuss it before making assumptions about included checks or corrections. The quote-comparison guide helps clarify that scope; the document scanning service overview provides a starting point for discussing your records and required delivery.

Frequently Asked Questions

Can we review a delivery without the original documents?

You can check whether files open, follow naming rules and work in your destination system. You cannot resolve every completeness or legibility question without a reliable source comparison. Record that limitation and arrange access to the relevant originals or agreed reference material before closing those issues.

Does correcting OCR repair a poor scan image?

Correcting recognised text and replacing a defective page image are different actions. If the image has lost information, changing its text layer does not restore the visual record. Identify which output is defective, agree the correction and recheck both the image and any associated text or data.

Who should approve a delivered scanning batch?

Agree an approver before delivery: someone authorised by your organisation to decide whether the scoped output is usable and to review unresolved exceptions. Technical or subject-matter reviewers can provide evidence for that decision. Approval should identify the delivery version and reviewed scope rather than imply every unexamined item was checked.

Discuss your scanning review requirements

Describe the document types, approximate volume, required output and checks your team needs. Send general project details only; do not include confidential documents in the initial enquiry.

Discuss Your Requirements

Related Articles

How to compare document scanning quotes in Australia

Document Digitisation · 8 min read

Searchable PDF vs data extraction: what should you request?

Document Digitisation · 9 min read