Every discipline that produces documents eventually invents version control, because every discipline eventually has the same accident: two copies of the same thing, different in some particular, and no way to tell which one governs. Engineering solved it with revision numbers. Law solved it with execution dates and revocation clauses. Software solved it comprehensively enough that the tooling is now free and universal.
Households have not solved it at all. They have the accident regularly.
The accident, specifically
A will is signed. Some years later circumstances change and a new will is signed. The old one is not destroyed — it is filed, because destroying a legal document feels reckless. A codicil is added at some point. A solicitor holds one version; the house holds another; an adult child was given a copy at some stage and still has it.
None of that is unusual and none of it is negligent. Each step was sensible on its own. The compound result is a family that, at the moment of death, holds several documents which all look authoritative, and must determine which is the last valid one — a determination that depends on execution formalities, dates, and revocation language, and which is exactly the kind of question that turns grief into litigation.
The same pattern runs through everything else, with lower stakes and higher frequency. Which version of the asset list is current. Whether the letter of wishes in the folder predates or postdates the one in the email. Whether the pension nomination on file reflects the second marriage.
What version control actually requires
It requires less than people expect. Three properties are enough, and none of them needs software.
First, every document carries a date on its face — not a filename date, not a folder date, a date written on the document itself. Metadata is lost in every copy operation; ink is not.
Second, superseded versions are marked as superseded, at the moment they are superseded. The discipline is to annotate the old copy the same hour the new one is signed, because that is the only hour when the relationship between them is unambiguous to everybody involved.
Third, one index records which version is current and where it lives. Not the documents — the index. A single line per document, kept in one place, updated when a document changes.
That is the whole system. It is dull, and it is the difference between an executor reading a clear chain and an executor building one from evidence.
The instinct to keep everything
The obstacle is rarely effort. It is the reluctance to destroy, and the reluctance is not irrational — a superseded will can occasionally matter, and legal advice on destruction is genuinely worth taking before acting. But there is a large difference between retaining a superseded document and retaining it unmarked, in a location that implies it is current, among documents that are.
Retention is a filing decision. Marking is a communication decision. Households conflate them, keep everything, mark nothing, and hand the interpretive work to whoever is least equipped to do it: a bereaved relative reading four documents for the first time.
The question is never whether the old version still exists. It is whether anyone can tell that it is the old version.
Drift is the normal state
It is worth being honest about the direction of travel. Records do not hold their accuracy on their own. Accounts close, providers merge, people move, relationships change, nominated beneficiaries stop being the right beneficiaries. A record written with great care in one year is quietly wrong in several particulars by the fourth, and nothing announces the change.
Version control does not prevent drift. It makes drift visible, which is the only thing that has ever worked. A dated document invites the question of whether it is still true. An undated one does not, and so it is never asked, and so the answer stays no.