Three dates, three meanings

Expiry is the date after which the document no longer has effect. Review is the date by which someone should look at it again — usually well before expiry, so that a renewal can be arranged without a gap. Supersession is when a newer document replaces it, which can happen at any time and often without notice to you.

Test reports typically have no expiry at all. Certificates usually do. Assessments often carry a validity period set by the assessing body. Treating them all as "documents with a date on" is what produces the classic gap: an expired certificate still cited on a live datasheet.

The effect of expiry is on the claim, not the file

When a certificate lapses, nothing happens to the PDF. What changes is the support behind every claim that relied on it. If the claims are linked to the evidence, that change is visible immediately and you can list the affected products in seconds. If they are not, nothing surfaces at all.

The right output is a status a person can act on — the claim needs technical review — not a verdict. Software is not in a position to say whether a product complies with anything, and a record that pretends otherwise is worse than one that says nothing.

Scope narrowing is the quiet one

Certificates are renewed with amended scope more often than they are withdrawn. A construction is dropped, a size limit is reduced, a hardware option is removed. The certificate number is unchanged, the expiry date has moved forward, and everything looks healthier than it was.

Comparing scope between issues is tedious and exactly the kind of thing that gets skipped. Recording scope as data — rather than as an attachment — is what makes the comparison possible at all.

A workable routine

Give every piece of evidence an expiry date where it has one and a review date where it does not. Set review dates far enough ahead to act. Link evidence to the claims it supports so that a change propagates. Record supersession explicitly, with a pointer from the old document to the new one, so historical records stay interpretable.

Then let the system raise the work: evidence expiring, evidence expired, claims relying on expired evidence, claims with no evidence at all. Those four questions catch most of what goes wrong.