Site records as Gateway 3 evidence: collation, not archaeology

Why do site records decide Gateway 3?

Strip the four BSR-named stall reasons to their roots and they are one failure repeated: the project cannot prove what it built. Gaps in fire and structural documentation are missing records. Weak change control is missing records of decisions. As-built discrepancies are drawings that no record reconciles to reality. Missing commissioning evidence is, literally, missing records. The completion assessment is an evidence review, and site records are the evidence.

The distinction that matters to an assessor is contemporaneity. A record made at the time, by someone who was there, dated and attributable, proves something. A reconstruction assembled at completion, however sincere, merely asserts it. Eighteen months after the cavity barriers were covered up, the choice is between a dated photograph and somebody's memory, and only one of those shortens a determination.

What does a Gateway 3-ready record set look like?

It maps to what the application must demonstrate. Inspection and test plan results, signed and dated, for the safety-critical work. Photographic records of concealed work taken before it was concealed, tied to locations. Product and delivery records connecting what was specified to what arrived to where it went, so the as-built drawings and the manufacturer literature agree. Commissioning records for every safety system, because integration evidence is a named stall reason. And the change control log that explains every departure from the Gateway 2 approval, with its category, justification and revised drawings.

Two qualities separate a usable set from a pile: structure and attributability. Structure means the records file against the golden thread from the start, not into a to-be-sorted folder. Attributability means each record says who, when and where without forensic work. The test is brutal and simple: could a stranger, two years from now, prove your wall build-up from what you kept?

Where do site communications fit?

Honestly: most of a project's real-time record already exists, scattered across WhatsApp threads, site diaries, snagging apps and email. The foreman's photo of the firestopping before the boards went on, the message agreeing the revised detail, the morning note that the mock-up passed. The problem is not that the evidence was never created; it is that it was created in tools that lose it: unattributed, undated in any useful sense, unsearchable, and gone with a phone upgrade.

That is why contractors preparing seriously for Gateway 3, the pattern across the firms winning approvals on our Gateway Projects Tracker, are formalising exactly this: inspection-and-test evidence, photographic records and change discipline captured as work proceeds, into systems that keep the date, the author and the location attached. The brief's phrase for the goal is the right one: Gateway 3 as a collation exercise, not an archaeology project.

Does the programme itself count as evidence?

It does, and it is the part most often left out of the pack.

Much of what the BSR needs to see is not just that a thing was done, but that it was done in the right order. Firestopping installed before the boards went on. The cavity barrier fixed before the cladding closed the cavity. A membrane inspected before it was covered. Once the work is concealed, the only thing standing between an assertion and a proof is a dated record made at the time, tied to the point in the sequence it belongs to.

The programme is where that sequence is written down, and it is usually the one document on the project that already carries the planned dates against which everything else can be read. A record set that sits entirely separate from it leaves the assessor to do the joining up, which is exactly the kind of work that turns a submission into a series of questions.

Two practical points follow. The first is that the planned dates matter as much as the actual ones: a baseline preserved through the job is what makes a variance legible later, and exchanging programmes in a format that drops baselines quietly throws that away. The second is that a date alone is weak evidence. A photograph timestamped the third proves work was happening on the third; it does not by itself prove an activity started then. Records that distinguish what was observed from what has been inferred survive scrutiny that records blurring the two do not.

None of this requires a different planning tool. It requires the record and the programme to be readable against each other, and the link between them to be made while the work is visible rather than reconstructed once it is behind plasterboard.

How do you set this up on a live project?

Start at Gateway 2, because the construction control plan and change control plan you submit there are promises about exactly this discipline. Assign ownership: one named role that audits the record weekly against the golden thread structure, trade by trade, while gaps are still photographable. Make the capture habit match how sites actually communicate, or it will not happen; a record system nobody uses is a stall reason with a licence fee. And treat the 8-week determination as the deadline it is: the evidence has to be complete when the application goes in, because that is the moment the archaeology bill, if you owe one, comes due.