What Should a Document Scanning Pilot Prove?

A document scanning pilot should inform a decision you need to make about your records. A folder of attractive PDFs is less useful if nobody can explain which working task was tested, what needed correction or which parts of the archive remain unexamined.

Pexatech's document scanning offer includes a free pilot of up to 500 pages. Collection and return delivery costs are not included. Use that opportunity to agree a useful trial scope; a page allowance alone does not define the output, the checks or a delivery date.

For an organisation arranging scanning in Sydney, a good starting question is: what would we need to learn before deciding on the next batch? The following approach helps prepare that discussion without changing the existing offer or treating a trial as approval to process the remaining archive.

Choose the decision before choosing the pages

Write the question as a task someone will perform. For example: can an authorised member of staff find a complete job file by its reference and read the approval attached to it? Identify the person, the reference they start with and the result they need.

That is more useful than asking whether the scan is "good". It distinguishes finding the right file, locating text within it and obtaining a checked value for another system. Use the output selection guide if those requirements are still being mixed together.

State Records NSW's guidance on assessing the need for digitisation connects record selection with the programme's goals, intended users and access needs. That guidance addresses NSW public offices. The practical questions here adapt those planning ideas for a commercial buyer; they are not a universal government specification.

Choose a bounded decision rather than asking one trial to establish every future requirement. If the question concerns job-reference retrieval, say whether reading handwritten approvals is also in scope. Name anything that needs a separate decision.

Include complete records and explain what they represent

Select examples from the groups relevant to your question, including known difficulties. Keep the context needed to perform the task: a cover, continuation sheet or attachment may be essential even when it looks less interesting than the main page. Ask how complete records can fit within the agreed pilot count before separating them.

State Records NSW's quality-assurance guidance discusses coverage of the range and quality of records, including poor originals, and the sequence and completeness of multipage items. It also recommends testing quality checks during pilots. This does not establish a statistical sample size for your commercial trial.

Consider a fictional Sydney maintenance business whose task is to retrieve job files and read their approvals:

This is a selection example, not a Pexatech client result or a confirmed package. The useful output is a short explanation of coverage. Do not describe a deliberately chosen set of easy and difficult examples as a random sample or use it to promise an archive-wide accuracy percentage.

Confirm that the trial can actually be carried out

Give the provider a general description of the proposed records, their condition, your suburb, the required output and any deadline. Resolve how pages will be counted, including sides and blank-page handling, against the existing cap. For the broader inventory and packing discussion, use the document preparation guide.

Before committing to dates or handing over records, confirm the service and handover arrangements, who will perform and review the work, and whether the required equipment and processing are available for the agreed scope. A readiness question is not evidence that a supplier already has every capability you might need.

Agree access to originals during the trial and what happens to them afterwards. Identify who may authorise sharing and arrange an appropriate transfer method before sending confidential material. Acceptance of the digital result should not be treated as an instruction to destroy originals.

State Records NSW's planning guidance includes staffing, equipment, responsibilities and documented pilot or staged approaches. For your enquiry, keep each unresolved dependency visible instead of turning an intended start date into a confirmed promise.

Make the result observable, including corrections

Before processing, record the expected answer to the task using the source records. Identify the job reference, the correct file and the approval page the reviewer should find. Agree which records and details will be checked and what would prevent acceptance. Use the scanning quality checklist for the detailed delivery checks.

Run the agreed task on the returned files in the application where staff intend to work. Note the output version, the item checked and the result. Keep "not tested" distinct from "passed": an omitted logbook or an untried import is still an open question.

If OCR is included, compare the important recognised text with its source. Adobe's Acrobat guidance describes reviewing uncertain words against the scanned image and correcting them. This illustrates a review step, not a statement that Pexatech uses that product or configuration.

Distinguish the first returned output from a corrected version. If a reference becomes searchable only after manual correction, record both that change and the review it required. A corrected result can be useful while revealing work that must be considered for later batches. It should not be presented as the untouched OCR result.

Decide what can proceed and what needs another check

Review the findings against the original question, rather than declaring the whole archive ready because some files opened successfully. A practical decision can be:

• Proceed within the tested scope: the agreed checks support the intended task for the represented records, with the next batch's scope and arrangements still to be agreed.
• Revise and retest: a file boundary, image setting, recognition step or review method needs a change before that task works reliably enough for the agreed purpose.
• Leave a group unresolved: material outside the trial or a remaining handling requirement needs separate assessment.

Record who made the decision and which files support it. Preserve the distinction between accepting the pilot output and committing to a larger job. Do not extrapolate a small, selected trial into a promised production speed, completion date or saving.

Bring a clear question to the free-pilot discussion

Start with the decision you want to make, the record groups that could inform it and the person who can assess the result. Include a realistic description of known difficulties and anything you propose to leave outside the trial. Unknowns can be discussed; there is no need to invent a precise page count or an output requirement you have not decided.

Confirm the sample, counting method, output, review and delivery arrangements before records move. The existing free pilot remains capped at 500 pages, with collection and return delivery costs excluded.

If anyone proposes using trial documents or results in a public example, agree that separately with the record owner. Participation in a pilot should not be treated as permission to publish names, documents or findings.