Point Cloud to BIM — Existing Building

A design team needed an accurate BIM model of an existing building with no current digital documentation, ahead of a planned renovation.

The building had been constructed decades earlier and, like most real buildings of that age, was not perfectly square, level or plumb anywhere — meaning a decision was needed upfront about how much of that as-built irregularity the new model should actually capture versus idealise for design purposes.

The supplied point cloud data also had a few areas of incomplete coverage where scan access had been restricted, which needed to be handled honestly rather than silently filled in with assumed geometry.

Part of the building was still occupied and in active use during the scanning and modelling phase, meaning some areas could only be captured under time constraints that limited how thoroughly they could be scanned compared to vacant areas.

The client's renovation design team included consultants who had not worked from a scan-derived model before, meaning the handover needed to include enough context about the model's specific conventions and confidence levels for them to use it effectively rather than treating it identically to a model built from scratch.

The building also had a partial mezzanine level added at some point after original construction that wasn't reflected in any surviving documentation at all, meaning the scan itself was the only way to establish this level's geometry and structural support arrangement.

Several rooms were also still fitted with furniture and stored equipment at the time of scanning, meaning parts of the scan data captured temporary contents mixed in with the building fabric itself, which needed to be distinguished carefully during modelling to avoid modelling a filing cabinet as if it were a permanent architectural feature.

Supplied point cloud data was used to model the existing building directly in Revit, checked against the scan for dimensional accuracy before drawings were derived.

Scan registration accuracy was checked first, since a model built from poorly registered scan data would carry that error forward invisibly into every downstream design decision.

A decision was agreed with the design team upfront on how much as-built irregularity to capture — enough to be honest about real wall alignment and level changes that would affect the renovation design, without modelling every minor surface imperfection that wouldn't actually matter to the project.

Areas with incomplete scan coverage were modelled at a lower confidence level and clearly flagged, rather than filled in with assumed geometry presented as if it were scan-confirmed.

For the occupied areas captured under tighter time constraints, we noted the reduced capture confidence in those specific zones separately from the more thoroughly scanned vacant areas, rather than presenting uniform confidence across the whole model.

A short handover briefing was prepared for the design team explaining the model's conventions — how confidence levels were flagged, what LOD had been applied, and how to interpret the areas of lower certainty — specifically for consultants who hadn't worked from a scan-derived model before.

The undocumented mezzanine's structural support was modelled carefully from the scan data alone, with a note flagging that its structural adequacy had not been independently engineered-assessed and should be reviewed by a structural engineer before the renovation design relied on it for any new load.

Furniture and stored equipment visible in the scan data were reviewed and excluded from the architectural model wherever they could be reliably identified as temporary contents rather than building fabric, with a small number of ambiguous items flagged for the design team's own visual confirmation rather than guessed at either way.

The design team received a working as-built model to design the renovation from, in place of the missing original documentation.

Flagging the areas of lower-confidence scan coverage meant the design team knew exactly where to commission additional site verification before finalising decisions in those specific areas, rather than treating the whole model as uniformly reliable.

Separating capture confidence for the time-constrained occupied areas gave the design team a realistic picture of where the model was strongest and where extra caution was warranted, rather than an artificially uniform impression of accuracy.

The handover briefing meant consultants unfamiliar with scan-derived models could use this one effectively from the start, rather than misinterpreting its conventions or over-trusting areas that were flagged as lower confidence.

Flagging the undocumented mezzanine's structural adequacy as unassessed meant the renovation design team knew to commission a proper structural review before relying on that level for any new load, rather than assuming a modelled structure was automatically an engineered one.

Excluding identifiable temporary contents from the model kept the as-built architectural model clean and genuinely representative of the building fabric, while flagging the ambiguous items rather than guessing avoided either cluttering the model with furniture or accidentally omitting an item that turned out to be a fixed architectural feature.