Scan to BIM: A Practical Guide for AEC Professionals

Scan to BIM sounds simple on paper: point a laser scanner at a building, get a 3D model out the other end. Anyone who's actually run a project knows it's messier than that. Between point cloud noise, modeling decisions, LOD arguments with clients, and software that chokes on 500 million points, there's a lot of room for things to go wrong.

This guide walks through what scan to BIM actually involves, where it works well, where it doesn't, and how to plan a project that doesn't blow your budget or timeline.

What Scan to BIM Actually Means

Scan to BIM is the process of capturing an existing physical space with 3D laser scanning (or photogrammetry) and converting that raw spatial data into an intelligent BIM model — typically in Revit, ArchiCAD, or a similar authoring tool.

The workflow has two distinct phases that get lumped together but require different skill sets:

  1. Scanning (data capture) — surveying the physical space with terrestrial laser scanners, handheld/mobile scanners, or drones, producing a point cloud.
  2. Modeling (BIM authoring) — a technician or modeler traces that point cloud into parametric BIM objects: walls, doors, ducts, structural members, MEP runs, and so on.

A point cloud by itself isn't BIM. It's just millions of XYZ coordinates with color/intensity values. BIM only exists once someone builds actual model geometry with object data referencing that cloud.

Why Teams Use Scan to BIM

  • Renovation and retrofit projects where as-built drawings are missing, outdated, or wrong (which is most of the time on buildings over 15-20 years old).
  • Facility management — owners want a model tied to real-world dimensions for space planning, maintenance, and asset tracking.
  • Clash detection before construction — confirming existing conditions before routing new MEP or structural elements through occupied or historic buildings.
  • Historic preservation — documenting geometry that can't be easily re-measured or that needs to be archived precisely.
  • Litigation and insurance documentation — capturing precise as-built conditions as evidence.

If your existing drawings are reasonably accurate and the renovation is minor, scan to BIM might be overkill. It earns its cost on complex, high-risk, or heavily altered buildings.

The Core Workflow

1. Define Scope and LOD Before You Scan

This is the step teams skip, and it's the one that causes the most rework. You need to agree, in writing, on:

  • Which building systems get modeled (architectural only? structural? full MEP?)
  • Level of Development (LOD) for each system — LOD 200, 300, 350, etc.
  • Level of Accuracy (LOA) — how close the model geometry needs to be to reality (per USIBD guidelines, this is often specified separately from LOD)
  • Which areas are in scope (whole building vs. specific floors/rooms)
  • Deliverable format (Revit model, IFC, point cloud only, hybrid)

Skipping this conversation is how you end up with a client expecting duct insulation thickness modeled from a scan that only captured architectural shell.

2. Site Scanning

Common capture methods:

| Method | Best For | Limitations | |---|---|---| | Terrestrial laser scanner (TLS) | High-accuracy interior/exterior capture | Slower, needs multiple setups, line-of-sight only | | Mobile/handheld SLAM scanner | Fast walkthrough capture, large areas | Generally lower accuracy than TLS | | Drone photogrammetry/LiDAR | Roofs, facades, site topography | Weather-dependent, less useful indoors | | Photogrammetry (ground-based) | Budget-constrained, texture-heavy documentation | Accuracy varies with lighting, camera, processing |

Scan registration (aligning individual scan setups into one coordinate system) happens either through targets (checkerboards, spheres) or cloud-to-cloud registration using overlapping geometry. Target-based registration is generally more reliable for high-precision work; verify current recommendations from your scanner manufacturer's documentation since workflows and software vary by hardware generation.

3. Point Cloud Processing

Raw scans need cleanup before modeling:

  • Removing noise, outliers, and moving objects (people, equipment that isn't permanent)
  • Registering/aligning multiple scan positions
  • Decimating or optimizing the cloud so it doesn't crush your modeling software
  • Converting to a format your BIM software can consume (RCP/RCS for Revit via ReCap, E57, PTS, etc.)

This stage is where a lot of project time disappears, especially on large or complex sites. Don't underestimate it when scoping schedules.

4. Modeling in BIM Software

The modeler traces the point cloud into actual BIM elements. This is manual, skilled work — not automated, despite what some marketing suggests. A human decides where a wall centerline actually is when the cloud shows a wall that's out of plumb by 3 inches over its height.

Key modeling decisions include:

  • Snapping tolerance — how strictly to follow the scan data versus regularizing geometry (e.g., do you model the wall exactly as it leans, or as a straight wall at the average plane?)
  • Element families — using existing family libraries versus building custom families for unusual conditions
  • Splitting elements — a single scanned wall may need to be split into multiple wall types if materials or thicknesses change
  • Handling occlusions — scanners can't see behind furniture, inside cavities, or above ceiling tiles that weren't removed. Modelers either interpolate, mark it as assumed, or flag it as unmodeled.

5. QA/QC and Deviation Analysis

Before delivery, compare the finished model against the source point cloud to check that it falls within the agreed LOA. Some teams run deviation heat maps (color-coded distance between mesh/model surface and cloud) to visually confirm accuracy. This step should be built into your schedule and price — it's not optional if you're contractually promising an accuracy tolerance.

LOD vs. LOA: Don't Confuse Them

This trips up a lot of teams new to scan to BIM.

  • LOD (Level of Development) describes how much information and geometric detail an element contains — is it a generic mass, or does it have real dimensions, material, and manufacturer data?
  • LOA (Level of Accuracy) describes how closely the model's geometry matches physical reality, independent of how detailed the object is.

You can have a highly detailed model (LOD 400) that's inaccurate, and a very accurate model (tight LOA) that's still just simple boxes (LOD 200). Define both separately in your scope documents. The USIBD Level of Accuracy specification is the common reference framework for this — check the current published version rather than relying on secondhand summaries, since terminology gets revised.

Common Mistakes That Blow Budgets

  • Scanning everything at maximum resolution "just in case." More data isn't free — it costs processing time, storage, and modeling hours. Scan at the resolution the project actually needs.
  • No coordinate system plan. If scans aren't tied to a shared project coordinate system or known survey control, integrating with civil/site data later becomes painful.
  • Assuming scan to BIM eliminates field verification. Scanners miss concealed conditions. Someone still needs to physically verify anything behind walls, above ceilings, or underground.
  • Treating the point cloud as the deliverable. Clients sometimes think they're getting a "model" when what's been scoped is really just a registered point cloud. Clarify this upfront, contractually.
  • Underestimating modeling hours. Modeling time depends heavily on building complexity, MEP density, and how strictly the client wants geometry to match actual (sometimes chaotic) as-built conditions. Estimate conservatively and revisit assumptions as complexity is revealed mid-project.

Software and File Considerations

Typical stack:

  • Scanning/registration software: provided by your scanner hardware vendor (varies by manufacturer — check current compatibility with your BIM authoring software before committing to a scanner purchase).
  • Point cloud engine in Revit: Autodesk ReCap is commonly used to process and link point clouds into Revit. Verify current supported file formats and point cloud size limits in official Autodesk documentation, since these change between releases.
  • Modeling: Revit, ArchiCAD, or similar, with the point cloud linked as reference geometry.
  • Interoperability: IFC export is standard for sharing models outside a single software ecosystem. Confirm which IFC schema version your downstream software expects (IFC2x3 vs IFC4) before delivery.

Large point clouds are notorious for tanking performance. Common mitigations: splitting the cloud into regions, using lower-resolution clouds for reference and higher-resolution ones for spot-checking, and working on hardware with adequate GPU/RAM — check your BIM software's current hardware recommendations rather than relying on older specs.

Choosing In-House vs. Outsourced Scan to BIM

| Factor | In-House | Outsourced | |---|---|---| | Upfront cost | High (scanner hardware, training) | Low (pay per project) | | Control over quality | Direct | Depends on vendor SLAs | | Turnaround for one-off projects | Slower to ramp up | Often faster, scanner ready to deploy | | Best for | Firms with recurring scanning needs | Firms with occasional or specialized capture needs |

If you're doing occasional renovation work, outsourcing scanning (and possibly modeling) usually makes more financial sense than buying hardware that depreciates. If scan to BIM is becoming a regular service line, owning equipment and building internal modeling capacity pays off faster.

Practical Checklist Before Starting a Project

  • Confirm LOD and LOA requirements in writing, per building system
  • Identify areas with restricted access or safety concerns before scan day
  • Establish project coordinate system and any required survey control points
  • Clarify deliverable formats (native BIM file, IFC, point cloud, or combination)
  • Set expectations on how occluded/inaccessible areas will be handled
  • Build in time for QA/QC deviation analysis, not just modeling
  • Confirm software version compatibility across the whole team (scanning, processing, authoring)

Final Thoughts

Scan to BIM is a capture-and-translate discipline, not a magic button. The scanning hardware gets attention because it's tangible and expensive, but the real project risk lives in scope definition and the modeling decisions that turn noisy real-world geometry into clean parametric elements. Nail down LOD, LOA, and deliverable expectations before the scanner ever leaves the truck, and the rest of the project runs a lot smoother.

Software capabilities, file format support, and hardware requirements referenced here change across releases — always confirm current specifications in the official documentation for your scanning hardware and BIM authoring software before finalizing a project plan.