Insight · Choosing software

Document control, and the revision
everybody was working to.

Building to a superseded drawing is expensive, common, and almost never anybody being careless. What has to be true for the right revision to reach the person holding the tool.

Published ·Updated ·5 min read·Written by

A technical drawing being marked up at a desk

Building to a superseded drawing is almost never a filing failure. The register is usually right. What fails is distribution: the current revision reaches the office, and the revision in the van, on the wall and in the WhatsApp thread stays where it was. Control means knowing who holds what, not knowing where the master is.

How it actually happens

Nobody decides to build to an old drawing. The sequence is ordinary and it repeats.

Revision C is issued and goes out. Somebody prints it and pins it in the cabin. Somebody else saves it to a phone. A subcontractor is emailed it and forwards it to their own foreman. Three weeks later revision D is issued, because a service moved or a client changed their mind.

Revision D reaches the office and the register is updated correctly. It reaches the site manager's inbox, where it sits beneath forty other emails. The printout in the cabin still says C, and nothing about it changes on the day D is issued. The copy on the phone is still C. The subcontractor's foreman never saw D at all, because the email went to their office.

The work then gets built to C, correctly, carefully, by people doing their jobs properly.

The register was right the whole time. Being right in the office is not the same as being current in the hand holding the tool.

Why it is expensive out of proportion

Three reasons, and the third is the one that turns a mistake into an argument.

  • It is found late. Work built to a superseded revision is usually discovered when something adjacent will not fit, which is after the trade has left and often after it has been covered up. The rework is not the original task, it is the original task plus opening up, plus resequencing.
  • It moves other trades. A wall in the wrong place is a wall in the wrong place for the M&E first fix behind it and the fit-out after that. One revision failure becomes three programmes.
  • Liability turns on issue records. The commercial question is not what was built. It is who was issued with what, and when. Where there is a clean issue record showing D was transmitted and receipted, the position is one thing. Where the drawing arrived as an attachment in a group chat, it is another, and the argument is expensive whichever way it lands. That is one of the six failures in why WhatsApp fails as a site record.

What control actually requires

Four things. None of them is a document management system, and all four can be done badly with one.

  1. One place that is unambiguously current. Not a folder called Latest, which will contain three files. A single location where the act of issuing a new revision makes the old one visibly not current.
  2. Superseding that is visible at the point of use. The failure is not that the old drawing exists, it is that it does not look old. Whatever a person is holding has to tell them it has been replaced, or it will be used.
  3. A record of who was issued with what, and when. This is the half that carries the commercial weight. An issue register showing transmittal and receipt is what turns a rework argument from a discussion into a document.
  4. A route to the person, not to the company. Issuing revision D to a subcontractor's head office does not put it in front of the foreman. Control that stops at the company boundary is not control of the thing you care about.

The printed drawing problem

Worth its own section, because every honest answer here has to deal with it.

People print drawings. They print them because paper works in the rain, does not need a signal, does not run out of battery, can be marked up with a pencil and can be laid out on a bench next to another one. Any system that assumes site will stop printing is describing a site that does not exist.

So the practical answer is not to abolish paper. It is to make the printed copy traceable and short-lived. Two habits do most of the work: print with the revision and date in a large mark rather than a small title block, and make the cabin wall somebody's explicit job once a week rather than nobody's job continuously. A wall walked every Friday with the current register in hand is a low-technology control that works.

It is not only drawings

The same failure applies to anything issued in versions, and the others are easier to overlook because nobody thinks of them as controlled documents.

A RAMS revised because the crane position changed, where the gang signed the previous version, is exactly this problem with a safety consequence rather than a commercial one, and it is covered in RAMS that hold up. So are method statements, the construction phase plan, specifications, and the subcontract order that was superseded by a variation. If it has a revision, it can be built to the wrong one.

Where this touches the platform

Drawings, permits, RAMS and site records hang off one job record rather than a folder tree or a message thread, so what is current is current in one place and access is granted per person rather than per company. Documents carry their versions, and the field app puts the current one in front of the person on site rather than in an inbox at their head office, which is the fourth requirement above and the one most systems miss. Fifty-eight document types are generated from the data captured on site, on your letterhead, findable years later. The current revision is one tap away for the person standing in front of the work, and the issue record exists without anybody maintaining a spreadsheet of transmittals.

Where to start, on Monday

Go to your busiest site and look at what is pinned in the cabin. Compare each drawing against the register. Most firms doing this find at least one superseded sheet on the wall, and finding it on a Monday costs nothing.

Then ask the harder question of the subcontractor with the most work in front of them: which revision are your people working to. If they have to go and find out, the answer is that nobody knows, and that is the gap this article is about.

Asked most often

The follow-up questions.

The same failure with a safety consequence is in RAMS that hold up.

How do I make sure everyone on site has the current drawing?
Four things: one location that is unambiguously current rather than a folder called Latest, superseding that is visible at the point of use so an old sheet looks old, a record of who was issued with what and when, and a route that reaches the person doing the work rather than their head office. The failure is distribution rather than filing.
How do I control drawing revisions on site?
One source that everybody reads from, with the revision visible on the drawing itself and superseded versions removed rather than crossed out. Issue by transmittal so there is a record of who received which revision and when. The control that matters most is removal. A current drawing pinned next to an old one controls nothing, because the person who picks up the wrong sheet had no way to tell.
Who should own document control on a smaller job?
The site manager, with a routine rather than a system, and the routine written down. On a job too small for a document controller the risk does not disappear, it just has nobody watching it. A single shared location for current documents, a transmittal email for every issue, and a rule that superseded sheets come down is enough for most projects.
Why do teams build to superseded drawings?
Because the new revision reaches the office and the register, while the printed copy in the cabin, the file saved to a phone and the version forwarded to a subcontractor's foreman all stay as they were. Nothing about those copies changes on the day the new revision is issued, so they continue to look current to the people holding them.
Why is building to the wrong revision so expensive?
It is found late, usually when something adjacent will not fit and often after the work has been covered up, so the rework includes opening up and resequencing. It moves other trades, because one element in the wrong place affects everything that follows it. And liability turns on issue records rather than on what was built, so a weak record turns a mistake into an argument.
What happens if work is built to a superseded revision?
Establish what was issued and when before deciding anything, because that determines who carries the cost. Where the current revision was properly issued and the site built to an old one, it is your cost. Where the revision arrived after the work was under way, or was never transmitted, it is a variation or a claim. The transmittal record settles the question, and without one it becomes an argument about recollection.
Should we stop site printing drawings?
No. Paper works in the rain, needs no signal or battery, takes a pencil mark-up and can be laid beside another sheet. Any system assuming site will stop printing describes a site that does not exist. Make the printed copy traceable and short-lived instead: print the revision and date as a large mark rather than in a small title block, and make the cabin wall somebody's explicit weekly job.
Should drawings still be printed on site?
Printing is not the problem. Printing without a controlled reissue is. A printed sheet is a snapshot that stays on the wall after the drawing changes. So a print needs a date, a revision, and a routine that removes it when superseded. Where teams work from tablets on the current set, printing should be the exception and treated as a deliberate act rather than a habit.
Does revision control apply to anything other than drawings?
Yes, and the others are easier to overlook. A RAMS revised after a crane position changed where the gang signed the previous version is the same failure with a safety consequence rather than a commercial one. So are method statements, the construction phase plan, specifications, and a subcontract order superseded by a variation. Anything issued in versions can be worked to in the wrong one.
What else needs revision control besides drawings?
Specifications, schedules, RAMS, the construction phase plan, temporary works designs, technical submittals and any register somebody works from. The failure is identical in each case: a version in circulation that is no longer the version in force. Drawings get attention because the consequences are visible. A superseded specification produces the same rework with less to point at afterwards.
Next step

Put the current revision in the right hand.

Nobody builds to the wrong drawing on purpose. They build to the one they were sent, which was current when it was sent.

  • Thirty minutes, weekdays, from tomorrow.
  • Nothing to prepare. Bring a job number and we mock that job up.
  • You drive it. There is no slide deck.
  • You keep what you saw as a 14-day trial. No card.