From Dots to Decisions: Defining a Fit-for-Purpose Image Standard for UK Capture
Photo: document scanner calibration test chart office quality control, via images.template.net
In many UK capture teams, 300 DPI has become folklore. It appears in supplier quotations, operator instructions and project specifications, often without anyone being able to say precisely why that number suits the records in front of them. It is a useful baseline for clean A4 text, but it is not a policy. A defensible image standard begins with the business purpose of the record, not with a scanner setting that has been copied from one project to the next.
Why the default persists, and where it fails
The 300 DPI convention has practical origins. It is generally sufficient for optical character recognition on ordinary typed business documents, and it produces manageable file sizes for routine correspondence. For a high-volume invoice or remittance process, it can be a sensible starting point. The difficulty is that it is applied indiscriminately to records with very different evidential demands.
A faint handwritten annotation, a coloured engineering stamp, a thermal receipt or a folded plan will not necessarily become fit for purpose simply because the scanner is set to 300 DPI. Colour mode, contrast, compression and operator handling may matter as much as resolution. Conversely, a clean printed letter may be perfectly legible at a lower setting if the chosen standard has been tested and documented. Treating one number as universal can either waste storage on low-risk material or leave high-value records with images that cannot support the decision they are meant to evidence.
Define the standard by record class
A workable approach is to create a small number of record classes and set an image standard for each. The classes should reflect business risk and evidential need, not departmental convenience. An illustrative matrix might include routine transactional records, signed agreements and HR evidence, high-detail technical material, and low-risk reference copies. The settings below are examples of the reasoning, not a universal prescription.
| Record class | Typical examples | Image consideration |
|---|---|---|
| Routine transactional | Invoices, remittance advices, standard correspondence | Legible text and reliable OCR for key fields; storage economy matters |
| Evidential records | Signed contracts, consent forms, HR letters, annotated approvals | Colour and tone may carry meaning; signatures and stamps must remain clear |
| High-detail material | Plans, drawings, dense forms, small print, technical schedules | Higher resolution and careful colour handling may be needed |
| Low-risk reference | Duplicates, published leaflets, convenience copies | Lowest defensible setting, provided legibility is proven |
For each class, the standard should state resolution, colour mode, compression approach, OCR expectation and any exception route. Colour should be used where colour carries meaning, such as highlighted amendments, red stamps or coloured signatures. Grayscale may be appropriate where tone matters but colour does not. Bi-tonal scanning can be efficient for high-contrast text, but it should be tested rather than assumed. Compression deserves equal attention: aggressive lossy compression can undermine fine detail, while lossless masters may be justified for evidential records. A common pattern is to keep a suitable master image and generate lighter access copies only where the workflow supports it.
Build an exception route, not an exception culture
Exceptions are where image standards often fail in practice. Damaged paper, bound volumes, thermal receipts, mixed batches and oversized documents do not fit a simple default. The standard should define when an operator may accept a lower-quality image, when a rescan is required, and when the record must be escalated for a different capture route. Each exception should be recorded with a reason. This prevents a temporary workaround from becoming an undocumented norm.
A testing routine that proves fitness
An image standard is only credible if it has been tested against the records it is meant to govern. Build a small sample set for each record class, including ordinary examples and difficult ones. Scan the same sample at candidate settings, then assess four things: human legibility, OCR accuracy against known critical fields, evidential fidelity for colour, stamps and signatures, and file economics per thousand pages. A control sheet containing small print, faint text, a signature and a colour mark is useful for repeat tests.
Record the results and keep the test images with the standard. Repeat the test after material changes to scanners, drivers, firmware or capture software. In 2026, with more capture workflows relying on AI-assisted recognition, image quality is not only a human reading issue; it affects how confidently downstream tools can classify and extract data. A supplier default should not silently overwrite a tested standard.
Document the standard so it survives change
A defensible standard should be short enough to use and detailed enough to audit. Include its scope, record classes, settings, exception rules, test evidence, owner and review date. Place it with the capture procedure rather than in a project folder that disappears after go-live. Reference it in supplier contracts and handover packs, so a new operator or outsourced provider inherits the reasoning as well as the settings.
Staff changes are a practical test of documentation quality. A new team member should not have to infer the standard from a colleague's habits. Supplier handovers should include the tested settings and the sample evidence. Audit responses should be able to show why a record class was scanned in a particular way, how exceptions were handled and when the standard was last reviewed. The strongest position is not that an organisation always scans at a particular number. It is that the organisation can demonstrate that its image standard suits the record, has been tested, and is applied consistently.