The Construction Records Playbook: What to Keep, Why, and What It Wins

The principle: records are made before you know you need them

A record made contemporaneously carries weight precisely because it was made when nobody had a reason to shade it. There was no claim, no dispute, no incentive — just somebody writing down what happened because that was their job. That is what gives it evidential force.

A record assembled two years later, for the purpose of a claim, has exactly the opposite property. And tribunals can tell. Reconstructed records are too tidy, too consistent, too convenient — they agree with each other in ways that real records never do. A genuine site diary contradicts itself slightly, notes the weather when the weather did not matter, and records the morning somebody’s van was blocked in. That texture is the evidence of authenticity, and it cannot be manufactured in a way that survives cross-examination.

Which produces an uncomfortable conclusion: the quality of your records is fixed long before the dispute. By the time you need them, the decision has already been made, by whoever did or did not fill in the diary on an ordinary Tuesday.

The daily diary — and what makes one useless

The site diary is the backbone of every delay claim, and most site diaries are worthless. Not incomplete — worthless, which is different and worse, because an incomplete record has gaps while a bad record has confident, unhelpful entries every single day.

A useless diary says "progressed blockwork, 12 men, fine". It records that work happened. It does not record what was achieved, where, by whom, or what stopped anybody from achieving more. It cannot support a claim because it contains no facts a claim can be built from.

A useful diary records the things that will matter later, and the discipline is knowing what those are in advance:

• Quantities installed, by location — not "progressed blockwork" but "42 m² of blockwork, Level 3, Zone B". A quantity can be compared to a plan; a verb cannot.
• Manpower by trade and by area, so productivity is computable and so trade stacking is visible.
• What was NOT available: the front that was not handed over, the drawing that had not arrived, the material that was not on site. This is the half everyone omits, and it is the half a disruption claim is made of.
• Instructions received verbally, and from whom. This is the raw material of a constructive-variation claim, and it exists nowhere else.
• Weather, honestly — including the days it was fine, because a weather claim is only credible from a diary that records good weather too.

Photographs — and the shots that are worth more than all the others

Site photographs are the cheapest contemporaneous evidence a project produces. Thousands of them, for free, as a by-product of people doing their jobs. And most of that value is thrown away — not because the photographs are bad, but because they are unfindable and unattributable.

What makes a photograph evidence is not the image. It is the metadata and the link: when, where, by whom, and which record it belongs to. And that metadata is fragile in a specific way — it survives capture and dies in transit. Screenshot a photo and it is gone. Send it through a messaging app and it is stripped. Forward it twice and what arrives is an image of unknown provenance, which the other side’s expert will describe in exactly those words.

The highest-value photographs on any project are of work about to be concealed: waterproofing before the screed, reinforcement before the pour, services before the ceiling closes, fire-stopping before the boarding. Once that work is covered, the only way to see it again is to destroy something — so a thirty-second photograph turns a five-figure opening-up exercise, three years later, into a two-minute search.

Make it a gate, not a good habit. A required item on the cover-up inspection that refuses to be signed without it. Good habits do not survive a wet Friday with a concrete truck waiting on the gate — and that is precisely the pour that gets argued about later.

Correspondence — a contractual record, not a conversation

Project correspondence is where the contract is actually administered: the notice that starts a clock, the instruction that changes the work, the letter that takes a position somebody will be held to. And it is almost universally managed as though it were chat.

Two failures dominate. The first is the notice that was not a notice. A notice is valid only if served by the contract’s method, to the contract’s addressee, within the contract’s time — and under FIDIC 2017 it must also identify itself AS a notice and cite its clause. An email describing the problem perfectly, to the right person, inside the window, is not a notice if it does not say it is one. And actual knowledge does not cure a defective notice: the other side knowing exactly what happened is legally irrelevant.

The second is the personal inbox. The project engineer resigns; IT deactivates the mailbox on a ninety-day policy nobody on the project was consulted about; and three years of instructions leave with them. Not degraded — gone, on a date. Correspondence addressed to the PROJECT and held in a register survives resignation. Correspondence addressed to a person does not, and the absence of a document is invisible until somebody goes looking for it.

Quantities and hours — the two numbers a disruption claim is made of

Delay is about the completion date. Disruption is about efficiency — the same work taking more hours than it should have — and a project can finish on time while having been disrupted expensively throughout, by accelerating and throwing labour at it. That extra labour is the loss.

The strongest way to prove it is the measured mile: compare a period of your own unimpacted work against the same work while impacted, using the project as its own control. It is persuasive precisely because it does not depend on a theoretical productivity norm that the other side can attack. It depends on what your own crews actually achieved.

Which requires two records, kept weekly, from the start: installed quantities against location, and hours allocated to activities. Neither can be reconstructed. And the second one has a trap in it — unallocated or overstated hours inflate the denominator, making your crews look less productive than they were. That understates your claim and, worse, corrupts the norms you will price your next tender from.

Registers — the difference between having a document and being able to find it

Forty thousand unsorted photographs are barely better than none. Worse in one respect: they create the belief that the project is documented, which survives right up until somebody has to find something.

The test for any archive is retrieval, not size. Can you produce every photograph of a specific location, in a specific window, in under a minute? Can you produce the correspondence in which a position was taken, by number and date? If not, the archive is a liability rather than an asset — because in a dispute the other side will ask for exactly that, and the answer will be a fortnight of scrolling.

This is what registers are for. A document register with revision control, so nobody builds from a superseded drawing. A correspondence register with sequential numbering and action-by dates, so a position is findable and a response clock is visible. A risk register connected to the notice machinery, so a realised risk becomes an event rather than a status change.

Retention — the decision made once, or made for you

Records must be retained against the limitation period, not against the project. A project team disbands at handover; the drive gets archived; IT eventually reclaims the storage. All of this is reasonable, and all of it is planned around the project’s life.

But the claim is not on the project’s timeline. In the UAE, decennial liability for structural defects runs ten years from completion, and it cannot be excluded by contract. The average construction dispute takes over a year once it starts. The photographs, the diaries and the as-builts either still exist at that point, or they do not.

And that gets decided exactly once — deliberately, by somebody thinking about a dispute that has not happened — or it gets decided by a storage clear-down that nobody was consulted about, in a year nobody remembers. You will not find out which until the day you need the document.

What each record is actually worth

It is worth being blunt about the economics, because record-keeping is otherwise experienced as an overhead rather than as an asset.

• A pre-cover photograph: converts an opening-up exercise into a search. Thirty seconds of effort against a five-figure cost and a programme impact.
• A signed daywork sheet: converts an assertion into an entitlement. An unsigned one submitted at month-end is worth close to nothing.
• A CVI sent within days: converts a conversation into a document, using the other side’s silence. Sent three months later, when the work is built, it has lost the property that made it work.
• A weekly quantity record: makes a measured-mile analysis possible. Without it, the most persuasive disruption method in the industry is simply not available to you, however badly you were disrupted.
• A notice served correctly: preserves an entitlement that is otherwise extinguished — regardless of merit, and regardless of what the other side knew.