What does comparing site photos against the BIM model actually catch?

What it catches, and when it pays

Deviations between the design and the build are not rare and they are not usually anybody’s fault — a service gets rerouted around an obstruction, a bracket goes where it can actually be fixed, a penetration moves 300mm. Each one is reasonable on the day.

The cost is entirely a function of when it is found. A duct at the wrong height, found before the ceiling grid goes in, is a conversation. Found after, it is a demolition — and found at handover, it is a piece of as-built fiction that the facilities team will pay for over the next decade.

So the value is concentrated in a narrow window: after the work is done and before it is concealed. That is the same window pre-cover photography lives in, and it is the reason both disciplines share a cadence.

Side by side, not overlaid

It is worth being precise about what this comparison is. The capture and the model view for the same location are shown next to each other, synchronised by location. That is enough to see a difference, and it is what a human needs in order to judge whether the difference matters.

It is not a geometric overlay of the image onto the model, and it does not do clash detection. Zepth views models and CAD in the browser and works alongside design tools rather than replacing them — the comparison puts two pictures of the same place side by side, and the judgement stays with the engineer.

Where it goes next

A deviation that matters is not a photograph, it is a quality event, and it should escalate as one. A local, one-off difference is an observation and a conversation. A systemic one — the same deviation appearing floor after floor — is not an observation any more, and it escalates through the normal path from observation to non-conformance.

The point of holding the capture in the same system as the quality record is that this escalation is a routing decision rather than a re-typing exercise.