The drawing issue register: prove what you sent, and when

Ask a practice of four people where the latest revision of a particular sheet is, and you'll usually get a confident answer and a wrong one. Not through carelessness. The drawing exists, it was issued, somebody remembers issuing it, and the version in the folder looks right. It just isn't the one on site.

The register is the cheapest insurance in a studio. Ten minutes a week to maintain, and it answers without argument the only question that ever really matters after the fact: what did we send, to whom, and when.

Two documents often confused

A transmittal accompanies a single issue. It says: here are these eleven sheets, at these revisions, issued to you on this date, for this purpose. It travels with the drawings and it's addressed to somebody.

A register is cumulative and addressed to nobody. It's the studio's own running list of every drawing on the project and every time each one went out. One transmittal produces one row per drawing in the register.

Small practices tend to keep transmittals and skip the register, because the transmittal is the thing the client asks for. That's the wrong way round. A transmittal is useful for a week. The register stays useful for the life of the building, and for however long afterwards somebody might raise a question about it.

The columns that earn their place

A register with thirty columns will not get filled in. Keep it to the nine that answer real questions, and resist the urge to add a tenth.

Studios get that last point wrong more than any other. The register is append-only in spirit: you add a row, you don't edit the previous one. Editing destroys the history, and the history is the whole product. An audit log works on the same principle, for the same reason: corrections belong in new entries rather than in overwrites.

The columns to leave out

Every register that collapsed under its own weight did so because somebody added a column that sounded useful. Three that reliably do more harm than good: a file path, which is wrong the first time anybody reorganises a folder; a scale, which belongs on the title block and nowhere else; and a percentage complete, which is an opinion recorded as though it were a fact. If you want to know what a sheet looks like, open the sheet.

The one genuinely optional column worth considering is a reason for issue in free text — "revised after structural comments of 2 May" — which costs a sentence and saves an hour of reconstruction later. Add it only if you will actually write in it.

Status codes carry the meaning

The revision tells you the sequence. The status tells you what the drawing is for, and it's the status that keeps people out of trouble.

• For information. Here is where we have got to. Do not price it, do not build it.
• For approval. We need a decision on this before we go further.
• For comment. We want the consultants to mark it up. Distinct from approval, which asks for a yes or a no.
• For tender. This is the set contractors are pricing. Changes after this point cost money.
• For construction. Build this. The only status that should ever be on site.
• As built. What was actually constructed, recorded after the fact. Never an instruction.
• Superseded. This drawing has been replaced. Kept for the record, not for use.

A sheet can go out five times with the same drawing number and a different status each time, and each of those is a different instruction. A concept plan issued "for information" and the same plan issued "for construction" aren't two versions of one thing. They're two separate acts with very different consequences.

Numbering that survives a growing job

Every practice invents a numbering system on its first large project and regrets it on its second. The regret is always the same: the scheme had no room in it. Sheets called 01 through 40 are fine until the client adds a basement.

Use fields rather than a single running number. Discipline, level, type, sequence. Which letters you pick matters far less than keeping each field separate and leaving slack in each one.

For revisions, adopt letters before the set goes out for construction and numbers afterwards. Rev A, B, C through design and approval; Rev 1, 2, 3 once the drawing is a construction document. The change of character in the marker tells anyone glancing at a title block which side of that line the sheet sits on, which is information a letter or a number alone can't carry.

If your consultants work to ISO 19650

The formal version of the same two ideas, and worth knowing even on jobs that will never mention it by name. ISO 19650 splits what most studios call "status" into a revision code, which says how settled the container is, and a suitability code, which says what you may do with it. Small Indian practices meet it most often through a PMC or an international consultant on a larger job.

One sheet through a whole job

The abstract rules are easy to agree with and hard to picture. Here is a single drawing — the ground floor plan of a house — from first sketch to handover, exactly as it would appear in the register. Six rows, one drawing number, and every argument the project could later produce already answered.

Notice what the register makes visible that a folder never could. The gap between Rev B going out on 4 March and being approved on 11 March is the week the design was frozen. The jump from C to 1 is the moment the job became contractual. And the six weeks between Rev 1 and Rev 2 is where a variation lives, if the change came from the client — which is a question the variation log answers and the register merely dates.

The superseded set

How it actually goes wrong: a revision goes out on a Thursday. The site engineer is on leave and the email lands under twenty others. The foreman is working from a print taken off Rev B, pinned to the site office wall since February. Nobody is negligent, nobody is lying, and the slab goes in at the old dimension. The same goes for what is said on site: an instruction given during a visit belongs in a site visit report, issued like a drawing.

The question afterwards isn't really about liability. It's about evidence. If you can show the drawing was issued, to whom, on which date, at which revision, with the status "for construction", and that it was acknowledged, you're in a straightforward position. If your record is an email thread and a memory, you aren't.

Two habits shrink this to almost nothing. First, every superseded revision stays on the record rather than being overwritten, so the register can always show what was live on any given date. Versioned document sets do that by default: a replacement adds the new revision without discarding the old one.

Second, there's one place the current set lives and everybody knows it's the place. Not a folder that was shared once and has drifted since, but a workspace per client where the current revision is the one you see. The moment the answer to "which is the latest?" becomes "check your email", you've lost.

The third habit is the one nobody does: when you supersede a sheet, say so in the covering note, by drawing number and revision. "Rev 2 attached" leaves the reader to work out what died. "Rev 2 attached; Rev 1 of A-01-P-201 is withdrawn and must not be used" leaves them nothing to work out. The transmittal template has a line for it, because it is the line most often left blank.

Running it without it becoming a job

A register nobody updates is worse than no register, because it looks authoritative and isn't. Keep the ritual small enough to survive a busy month.

Who keeps it, at three studio sizes

On the smallest jobs a spreadsheet is enough, and anyone telling you otherwise is selling something. It stops being enough at roughly the point where more than one person in the studio issues drawings, because a shared spreadsheet has no way of noticing that two people edited the same row. If you're at that point, think about what the practice runs on rather than adding another tab.

When the register has already gone wrong

Most studios come to this article mid-project, with eight months of issues behind them and no register at all. Reconstructing one is dull but not difficult, and it is worth doing on any job still running.

1. Work backwards from sent mail, not from the drawings folder. The folder tells you what exists; only the sent items tell you what left the studio and when. Search by attachment type and by the client's domain.
2. Enter what you can prove, and mark the rest. A row with a date you are sure of and an acknowledgement you are not is still worth having. A row you invented to fill a gap is worse than the gap.
3. Draw a line at today and re-baseline. Issue the complete current set at current revisions, with a covering note listing what it supersedes. From that date the register is complete, and you only ever have to defend the period after it.
4. Do not backdate anything. Ever. A register whose credibility rests on a date somebody typed in later has no credibility at all, and this is the one failure that turns a document from an asset into a liability.

One last thing, slightly outside document control but bearing on it directly. Your register is also the record of what you licensed and when, which matters if a question ever arises about who owns the drawings. And a register showing a set was issued against a stage payment is quietly useful when a fee conversation turns awkward — which is the argument for attaching the invoice to the set rather than sending the two separately and hoping.