What good evidence looks like

Good evidence is contemporaneous, retrievable and consistent.

• Contemporaneous: created at the time, not reconstructed afterwards.
• Retrievable: findable in minutes by somebody who was not involved.
• Consistent: it agrees with what the record says.

Most evidence failures in further education are failures of the second property rather than the first. The evidence exists. It was seen, checked and stored by somebody who has since left, in a place nobody thinks to search.

Retrievability is not administrative tidiness. At audit, evidence that cannot be found within a reasonable time does not exist.

Test retrievability rather than existence when sampling: ask somebody uninvolved to produce the evidence for a named learner within a set time.

Where evidence is created retrospectively because it was genuinely missed, mark it as retrospective and record why. An undated reconstruction presented as contemporaneous is a far worse finding than an honest one.

Evidence answers five questions about every funded learner.

• Who are they, and are they who we say?
• Were they eligible for this funding?
• What were they enrolled on, and did it start when we say?
• Did they receive what we claimed for?
• How and when did it end?

Organising evidence by question rather than by document type makes gaps obvious. Most colleges can answer the first and third questions easily and struggle with the second and fourth.

Where a stream has specific requirements, such as off-the-job training records for apprenticeships or the learning support chain, those sit under the fourth question.

Build the evidence expectation per stream from the funding rules for the funding year, and record it in a single document that the enrolment and delivery processes are designed against.

That document is the thing you hand a new compliance officer, and the thing you test your sampling against.

When data that has already been reported is changed, record why it was changed and who authorised it.

An unexplained change between returns is a standard audit question, and reconstructing the reason months later is much harder than writing it down at the time.

This is the cheapest control in the entire assurance framework and the one most often missing. A free-text reason recorded at the point of change answers a question that would otherwise take a day.

Where the student record system supports a change reason, require it for funding-critical fields. Where it does not, maintain a change log for bulk changes at minimum.