Scan-to-BIM Workflow Guide: Step by Step (2026)

This guide walks through the five phases, the LOD level you should target, the accuracy you can expect, and how ATIS.cloud fits into the workflow on the diffusion side (sharing the cloud with the modeler, the architect, the client) so the scan stops being a 200 GB file stuck on someone's desktop.

Quick definition: what scan-to-BIM actually does

Scan-to-BIM is the workflow that captures an existing building with a 3D scanner, produces a point cloud, then converts that cloud into a parametric BIM model. The output is what the industry calls an "as-built" model: a digital twin that reflects the building as it actually stands, including the gap between the original drawings and decades of unrecorded modifications.

The comparison between the scan and the BIM model is itself a workflow, often referred to as "scan vs BIM" (called "as-built"). The interesting part is not the scanner. The interesting part is the chain that goes from raw points to a model someone trusts enough to base a renovation tender on.

The 5 phases of a scan-to-BIM project

The consensus across industry guides (Matterport, NavVis, Autodesk) is a five-phase pipeline. Skipping or rushing any phase shows up later as rework, missed clashes, or a model your structural engineer cannot trust.

Phase 1. Planning and scope

Before any scanner leaves the office, you negotiate the deliverable. The two key parameters: Level of Accuracy (LOA, how much real-world tolerance the points have) and Level of Development (LOD, how detailed the BIM model needs to be). The USIBD publishes an LOA standard; the BIM Forum / AIA publish the LOD standard.

• Renovation: target LOD 200 to LOD 300, LOA 20 to LOA 30 (5 to 15 mm).
• Facility management twin: can stop at LOD 200.
• Fabrication-ready MEP twin: needs LOD 400.
• Get this wrong: you either over-deliver (and lose money) or under-deliver (and get the call from the architect three weeks later).

Planning also covers the survey control: do you tie scans to a project grid, a national geodetic reference, or a local arbitrary origin? It covers logistics: access times, occupied zones, dust, vibrations, scaffolding. It covers the scanner choice (static for accuracy, mobile or handheld for speed, drone for facades and roofs). And it covers the registration strategy: targets versus targetless cloud-to-cloud registration.

Phase 2. Field scanning

On site, the operator positions the scanner at predefined stations. For static terrestrial scanning, expect 30 to 50 percent overlap between stations so software can later align them. Each station produces a panoramic scan in a few minutes (static terrestrial scanners from FARO, Leica, Trimble and Riegl are all in this family).

Mobile mapping with NavVis or FARO Orbis trades a bit of accuracy for an order-of-magnitude speed gain. Matterport sits at the bottom of the accuracy ladder but covers a small house in an afternoon. All the scanner brands that ATIS.cloud handles natively (FARO, Leica, NavVis, Riegl, Trimble, Viametris, Matterport, etc.) cover the bulk of professional scanning hardware.

Phase 3. Data processing and registration

Back at the office, individual scan stations are aligned ("registered") into a single coordinate system. Industry guides report a target of under 2 mm registration error on a well-planned static project. Manufacturer-side processing software from FARO, Leica, Trimble and NavVis does most of the work, with cloud-to-cloud registration assisted by AI or targets.

• Noise removal: filter out ghost points caused by moving objects or atmospheric interference.
• Vegetation filtering: drop tree foliage and grass that hide built surfaces.
• Classification: separate ground from building from clutter.
• Export: usually E57 for portability, RCP/RCS for downstream Revit work, LAS/LAZ for survey workflows.

Phase 4. Quality assurance

QA is the phase most teams skimp on, and the phase that distinguishes a clean deliverable from a costly mess. Check three things.

1. Registration accuracy: target under 2 mm on static projects, look at the per-station error report.
2. Coverage gaps: overlay the cloud on a floor plan and look for blind zones.
3. Control point alignment: verify that the cloud lines up with surveyed reference points.

Phase 5. Deliverables and BIM modeling

Two deliverables: the cleaned point cloud, and the BIM model derived from it. The point cloud usually goes out in E57 (the open ASTM standard) plus a Revit-friendly RCP for the modeler. The BIM model is built in Autodesk Revit (the dominant platform), sometimes in ArchiCAD or AllPlan, by tracing walls, slabs, columns, beams, doors and windows over the cloud.

Smart objects (parametric families) replace dumb geometry. MEP systems are added if the LOD calls for it. The model is then exported in IFC (the open buildingSMART standard) for cross-platform exchange. Modeling time dwarfs scanning time: a typical commercial renovation with LOD 300 modeling runs 2 to 6 weeks of skilled BIM work, depending on building complexity. Estimates of 4 to 8 weeks end-to-end (scan plus model) for a mid-sized commercial project are common in industry guides.

LOD levels: choosing the right target

Level of Development (LOD) defines how complete a BIM model element is. The AIA defines LOD 100, 200, 300, 400, 500; the BIM Forum added LOD 350 for trade coordination. Here is what each level means, with the typical scan-to-BIM use case.

Accuracy: what to expect, what to demand

Static terrestrial laser scanners (FARO, Leica, Trimble, Riegl) deliver point accuracy in the millimeter range under good conditions, per manufacturer specifications and Wikipedia references on 3D scanning. Mobile mapping systems (NavVis, FARO Orbis) typically deliver centimeter-range accuracy. Photogrammetry from drones can match laser scanning on textured outdoor surfaces but degrades on smooth or reflective surfaces.

Two metrics matter on a scan-to-BIM project. Point accuracy: the deviation of each scanned point from the real surface. Registration accuracy: how cleanly all scan stations align. Industry guides target under 2 mm registration error on a careful static project. Real-world projects often see 3 to 5 mm. That ceiling cascades into the model: a BIM element cannot be more accurate than the cloud it traces.

File formats in the pipeline

• E57: open ASTM standard (E2807). The default exchange format. Use it whenever you ship a cloud across two organizations or two software stacks.
• LAS / LAZ: open ASPRS standards. The reference for outdoor and aerial LiDAR. LAZ is a lossless compression of LAS, typically yielding files several times smaller.
• RCS / RCP: Autodesk Recap native formats, the canonical input for Revit. ATIS.cloud reads and writes them without an Autodesk license, which is unusual.
• LGSx: proprietary Leica Hexagon format. The Hexagon ecosystem is migrating progressively from the "Leica" name to "Hexagon".
• IFC: open buildingSMART format for the BIM model output. The non-negotiable exchange standard for the model itself, not the cloud.

For a deeper look at file formats and the IFC standard, see our guides on IFC files and point clouds.

Tools the modeler actually uses

A scan-to-BIM stack mixes capture-side software (scanner vendors), processing software, modeling software, and diffusion software. Here is how the main tools split across the four roles.

Where ATIS.cloud fits in the workflow

ATIS.cloud is not a modeling tool. It does not produce BIM elements from a point cloud. What it does, very well: hold the cloud (up to 1 TB per file, up to 5 TB of total workspace), stream it to anyone in a browser, share it via permissioned links, and run a scan vs BIM (called "as-built") comparison once the model is back.

• Every scanner brand natively: FARO, Leica, NavVis, Riegl, Trimble, Viametris, Matterport, etc.
• Formats on every plan: E57, LAS, LAZ.
• Proprietary formats on all plans: RCS, RCP, LGSx.
• Closing the loop: the IFC from the BIM model can be loaded back into ATIS.cloud and compared against the cloud, where the as-built validation happens.

See our BIM managers use case and BIM features overview for more.

Six mistakes that ruin scan-to-BIM projects

1. Skipping the planning phase. No agreed LOD, no agreed LOA, no agreed coordinate system. Every downstream phase pays for it.
2. Under-scanning. Not enough stations, not enough overlap, no targets. Registration error blows up.
3. Over-modeling. Building a LOD 400 model when the client paid for LOD 200. You lose money and the client cannot tell the difference.
4. Ignoring QA. Shipping a cloud without checking registration error and coverage gaps. The architect calls back in week three.
5. Working in silos. The scanner team, the modeler, the client never see the cloud together. Issues surface late.
6. Stranding the cloud. The 80 GB scan sits on one disk, nobody downstream can open it. The asset is invisible to the project.

FAQ