Scan-to-BIM, Explained: Turning Point Clouds into Usable Models

A laser scan captures reality in millions of points. But a point cloud isn't a model — you can't tag it, schedule it, or run a clash test against it. Scan-to-BIM is the work of turning that raw cloud into clean, intelligent geometry your team can actually build from.

If you're renovating, retrofitting, or documenting existing conditions, here's what scan-to-BIM is, when it pays off, and how to brief it so you get something usable.

What scan-to-BIM actually is

The process has three stages:

1. Capture — a laser scanner (or photogrammetry) records the space as a point cloud, often hundreds of millions of points.
2. Registration & cleanup — overlapping scans are aligned into one coordinate system, and noise, clutter, and stray points are removed.
3. Modeling — a modeler builds accurate BIM geometry (walls, floors, structure, MEP, equipment) to an agreed level of detail, referenced to the cloud.

The deliverable is a model — typically Revit — that reflects the building as it really is, not as the old drawings claim it was.

When it's worth it

Scan-to-BIM earns its keep when existing conditions matter and the drawings can't be trusted:

• Renovations and tenant fit-outs in older buildings
• Retrofits where new systems must thread through existing structure and MEP
• Clash-sensitive coordination in occupied or congested spaces
• As-built documentation for handover, facilities, or future projects

If you're building greenfield, you don't need it. If you're touching something that already exists, accurate as-builts prevent the expensive surprises that show up during construction.

The key decision: level of detail (LOD)

The single biggest driver of cost and usefulness is how much detail you ask for. Modeling every conduit and bracket to a high LOD takes far longer than capturing primary architecture and structure. Match the LOD to the decision the model needs to support:

• Lower LOD — overall geometry, room layouts, primary structure. Great for planning and space studies.
• Higher LOD — accurate MEP routing, equipment, and connections. Needed for tight coordination and fabrication.

A good modeler will help you pick the right LOD per system instead of over-modeling everything — you rarely need the same detail across the whole project.

How to get a clean result

• Share the scan registered, if possible, with a known coordinate system and units.
• State the tolerance you need (how closely the model should track the cloud).
• Prioritize systems. Tell the modeler what matters most — structure, architecture, a specific MEP run — so effort goes where it counts.
• Agree the deliverable format up front: native .rvt plus exports (.ifc, .dwg, .pdf) as needed.
• Build in a review pass so you can confirm accuracy before it's finalized.

What you should get back

A usable scan-to-BIM deliverable is clean (noise removed), accurate (modeled to the agreed tolerance and LOD), coordinate-correct, and handed over in native files you own. From there it slots straight into your design and coordination workflow.

The bottom line

Scan-to-BIM converts "we think the wall is here" into "the wall is here." Brief it with the right LOD, the right tolerance, and clear priorities, and you get an as-built model that pays for itself the first time it catches a clash before the field does.