Which Version Is Final? Ending Document Version Chaos

TL;DR: You know which version is final because someone designated it, not because of its filename. Once a final exists, superseded drafts have served their purpose and are usually transitory records that can be deleted. Version chaos is not a naming problem. It is the result of never deciding when a document is done.

Everyone has met the folder. Report_final.docx, Report_final_v2.docx, Report_FINAL_use_this.docx, and, most ominously, Report_final_v2_JS_edits_MERGED.docx modified at 11:47pm.

The usual diagnosis is that the team needs a better naming convention. That is not quite right. A naming convention tells you what a file is. It does not tell you which of six candidates governs, and adding version numbers to a bad situation produces precisely numbered confusion.

Syracuse University's records management guidance puts the actual principle in one line: once a final document or report is completed there is no longer a need for "Final_Report_version2" or other versions. (Syracuse University Libraries)

The problem is not that the versions are badly named. It is that they still exist.

How do you know which version of a document is final?

Because a person designated it, and that designation is recorded somewhere other than in a filename.

There are three reliable ways to record it, in descending order of robustness.

Version history in the platform. Google Docs, SharePoint and similar systems keep prior states inside one document. There is exactly one file, and the current state of it is current by definition. This removes the question rather than answering it.

A designated location. The final lives in a Final or records folder and drafts live elsewhere. The folder carries the status, not the filename.

Deletion of the superseded. Once a final exists, the drafts go. If only one version remains, identifying the final is trivial.

What does not work is encoding status in the filename, because filenames are written by whoever saved the file and are never updated when circumstances change. The word final in a filename is a claim made at a moment in the past by one person, and it is the single least reliable metadata on a shared drive.

This is why our file naming convention guide recommends keeping status out of filenames entirely. Names should say what a document is. Systems should say where it stands.

What should happen to old drafts?

Most of them should be deleted, and records guidance is more permissive here than people expect.

Princeton's university-wide recordkeeping guidance is direct: identify the final draft of a record and delete or securely shred previous versions no longer needed for reference. (Princeton University Records Management)

UBC's manual explains why this is allowed rather than merely tolerated. Drafts fall under transitory records, defined as "Records or data that are needed only for a limited time in order to complete a routine action or prepare a final record (e.g. drafts)." Its conclusion is explicit: such material "holds no evidential value and can be destroyed once the information has been read and is no longer required for operational use." (UBC Records Management Manual)

The University of California's cleanup handout turns this into a step you can perform today. Among its fifteen minute tips: "Review and delete drafts if a final version has been approved." (UCOP, Managing your electronic records)

Two caveats, and they are not small.

The first is that drafts documenting a decision process are sometimes themselves records. A draft policy annotated by a committee may evidence how a decision was reached. If your schedule treats that series that way, it governs.

The second is that under a legal hold, nothing gets deleted, drafts included. See legal holds and records destruction.

Why does version sprawl create real risk?

Because somebody eventually acts on a superseded document, and nothing about it looks wrong at the time.

The failure is quiet. A colleague sends a rate card to a client from a folder containing four versions, and picks the one modified most recently, which happens to be the one someone opened to read rather than the one that was approved. Nobody notices for a month.

Under audit the failure is louder. Producing five conflicting documents that all claim to be final invites a question you cannot answer well: how does this organization know what its own policies say? This connects directly to the question of which copy is authoritative, covered in official copy vs duplicate.

There is also a cost that never gets counted. Every person who opens that folder has to work out the answer again. They open two or three files, compare, and pick. Five minutes, every time, for every person, forever, because nothing about the folder improves as a result of the effort.

How do you stop version sprawl from starting?

Four habits, in order of impact.

Edit in place rather than saving new copies. Where the platform supports version history, one document with a hundred revisions is strictly better than a hundred documents. This single habit eliminates most sprawl.

Send links, not files. Carleton's guidance recommends "Sending links (not attachments) to invite review or collaboration on documents," because an attachment "will create a duplicate copy in the SharePoint library of the recipient." (Carleton University Records Storage Guide) Eight reviewers with a link produce eight sets of comments on one document. Eight reviewers with an attachment produce eight documents.

Mark completion as an event. The moment a document is approved, something has to happen: it moves to a final location, or the drafts get cleared. Most version chaos comes from documents that were never formally finished, only gradually stopped being worked on.

Sweep on a cadence. UCOP's framing is the right scale. Fifteen minutes, sorting by date, deleting drafts where a final was approved. A short recurring sweep beats an annual reorganization nobody schedules.

Can AI tell you which version is final?

It can group the candidates and show you how they differ. It cannot tell you which one was approved, because approval is an event that happened among people and frequently left no trace in the file.

This is a genuine limit rather than a temporary one. If a document was approved in a meeting and nobody recorded which file was on screen, that information does not exist in the drive. The best a system can do is narrow six candidates to two and show what changed between them, which is useful but is not the answer.

The Drive AI narrows it that far. It compares documents by content rather than by filename, so a version chain still groups together after someone renamed a copy on its way into another folder, and it can surface where every version of a document ended up across connected storage. On arriving files it applies your naming convention, which at least stops new files entering the drive as document(3).docx.

What it does not do is designate a final version, delete superseded drafts on a schedule, or decide that a draft has lost its evidential value. Those are calls against your retention schedule, and the output of the automation is a grouped list for somebody to resolve.

The realistic benefit is that the recurring five minute cost of working out which version governs becomes a one-time review, and new files stop arriving unnamed.

Frequently Asked Questions

How do I know which version of a document is final?

By a recorded designation rather than the filename: the platform's version history, a designated final location, or the fact that superseded drafts were deleted. A filename containing the word final is a claim made by one person at one moment and is not reliable evidence.

Should you delete old drafts of a document?

Usually yes, once a final exists. Princeton's guidance is to identify the final draft and delete or securely shred previous versions no longer needed for reference, and UBC classifies drafts as transitory records with no evidential value. Exceptions apply where drafts themselves document a decision process, and where a legal hold is active.

Is it better to use version numbers or version history?

Version history, wherever the platform supports it. It keeps one document with a recoverable past, which removes the question of which file is current rather than answering it. Version numbers in filenames reintroduce the same ambiguity in a tidier format.

Why does version chaos matter if we can just ask the author?

Because the author leaves, forgets, or is unavailable at the moment of need, and because the cost repeats for every person who opens the folder. Under audit it becomes a larger problem, since conflicting documents that all appear final suggest the organization cannot identify its own records.

Do drafts need to be kept for the retention period?

Generally no. Drafts are typically transitory records rather than records of the series, so they carry little or no retention. Your own schedule governs, and drafts that evidence how a decision was made can be an exception.

Where to start

Open the folder you already know is the worst one and do only this: find the approved version, move it somewhere obvious, and delete the rest if no hold applies. Do not rename anything. Fifteen minutes on one folder teaches more than a policy document about version control.

Then make the sweep recurring rather than heroic. What is ROT data covers the wider category this belongs to, and how to organize a shared drive covers the structure that stops the folder refilling.