What a site engineer should record, and what they should not

The site engineer is the only witness to most of what happens on a project. The head office, the client, the contractor's management, the auditor who arrives two years later — all of them see a report of it, never the thing itself.

That is not a compliment. It is a description of the job. What the engineer wrote down happened. What the engineer did not write down is available to be argued about, later, by people who were not there.

So what to record is not a clerical question. It decides what will still be true when everyone's memory has gone soft.

What must be recorded, because it will be asked about#

Start with the work itself.

• What work was done, and where exactly. Not "slab work". Which element, which grid, which level, which block. A location that only makes sense to the person who wrote it is not a location.
• Quantities measured, and by whom. A quantity has an author. Two people measuring the same pour will differ, and knowing who held the tape is half of settling that difference later.
• Material received, and its condition. What arrived, how much, and what state it was in at unloading. Damage noticed after the lorry has left the gate is your damage.
• Labour present, by agency. Not a single headcount. So many from this contractor, so many from that one, in these trades.

Then the entries that decide disputes.

• Weather, or any event that stopped work. The hours lost and the reason. Delay claims are built out of these entries or they are not built at all.
• Instructions given and received, with who gave them. An instruction with no author is not an instruction. It is a rumour that cost money.
• Anything covered up before it could be inspected. Reinforcement concreted over, a service buried, a joint plastered. Record it before it disappears, because afterwards your note is the only evidence there is.
• Any deviation from drawing, with who approved it. Deviation is normal on every site. The unapproved deviation is the one that becomes expensive.

What should not be recorded#

This half matters as much, and almost nobody is told it.

• Opinions about people. That a supervisor is careless, that a gang is lazy, that a vendor cannot be trusted. These sentences are read out in rooms you will not be in, years after the person has changed. They damage the writer as much as the subject, and they prove nothing.
• A guess written as a measurement. If the pour was not measured, the entry is that it was not measured. An estimate recorded in the same column as a measurement contaminates every figure around it.
• A quantity copied off a docket and presented as measured. The whole point of an independent measurement is that it was taken independently, and the difference between the two is the subject of measured and docketed quantity. A copied number destroys the only check anybody had.
• A photograph with no note of where and when. An unlabelled photograph of reinforcement is a photograph of reinforcement somewhere. It answers nothing and it can be answered back with another photograph.

An observation is a record, a conclusion is not#

This is the distinction underneath everything above.

The joint was cast at 3 p.m. and the previous pour was completed the previous evening is an observation. The cold joint is the contractor's fault is a conclusion. The first cannot be attacked. The second invites somebody to attack it, and once they do, the observation buried inside it gets attacked along with it.

The engineer's advantage is that they were there. That advantage is spent the moment they start writing findings instead of facts. A conclusion can be reached later by anyone holding good observations. An observation cannot be reconstructed later by anyone at all.

Where a conclusion genuinely has to be recorded — a decision was taken, work was stopped — write it as what it is, with its author, and keep it in a different sentence from the facts that led to it.

Record it at the moment, not in the evening#

The evening write-up is where accuracy dies. By six o'clock the sequence has collapsed, the two loads have merged into one, and the time an instruction was given has become "in the afternoon".

Record at the moment. Whatever the medium, the entry is worth more the sooner it is made, and a rough entry made at the time beats a tidy one made at eight. If the practical thing at that moment is to speak rather than write, speak — the argument for that is in the voice note is the report. The tidy version can be assembled afterwards into the daily progress report, and it will be assembled out of something real.

A photograph with a note beats a note. The note gives the photograph its where and when; the photograph gives the note something that cannot be rewritten. Neither is worth much alone, which is the argument in a site photograph is a record.

The record he already makes#

There is a version of all this that engineers do without being asked, and most software never sees it. He sends a voice note — hands dusty, screen in the sun, one hand on a ladder. Speech is the only input available, and it usually carries more than the form would have collected.

We treated those as a labelled blob nobody opened for far too long. Transcribing them was the easy half. The half that mattered was deciding that the words afterwards take the same path a typed message takes, rather than a parallel one, because a parallel path is a second implementation of every rule and second implementations drift. That account is the voice note is the report.

If an office wants an engineer to record more, the honest move is to accept the medium he already uses and do the structuring at the desk, where somebody has a keyboard and a moment.

Your own copy is not the office's copy#

The office copy exists to run the project. It is edited, summarised, merged into other documents, and eventually archived by somebody with different priorities.

The engineer's copy exists for one purpose: to say what this person saw, on this day, in their own hand. It should be complete, dated, and kept where a change of employer, a change of software or a change of mood at head office cannot reach it. Nobody is being accused of anything by that. Records get lost innocently far more often than they get lost deliberately, and the reasoning is set out in keep your own copy.

When the two copies disagree, the one written at the site on the day is the one that persuades people.

The short version#

Record what happened, where, in what quantity, and on whose instruction. Record what was covered up and what departed from the drawing, with names against both.

Do not record opinions, guesses dressed as measurements, docketed figures dressed as measured ones, or photographs with no place and no date.

An observation is a record. A conclusion is not. Keep them apart, write at the moment, and keep your own copy.