Point Cloud to BIM: A Step-by-Step Workflow Guide

For anyone who has worked through a Scan to BIM project before, the workflow makes intuitive sense once you've seen it end to end. For teams evaluating it for the first time, the process can feel like a black box — you hand over site access, and eventually a Revit model arrives. What happens in between is worth understanding, both because it helps you set realistic expectations and because it makes you a better client when you're scoping and reviewing the work.

Here's a straightforward breakdown of how a point cloud to BIM workflow actually runs, from the first day on site to final model delivery.

Why the Source Data Quality Determines Everything

Before getting into the steps, it's worth establishing one principle that runs through the entire workflow: the quality of the BIM model is constrained by the quality of the point cloud. A poorly registered scan, a dataset with significant coverage gaps, or a point cloud captured without proper ground control will produce a model with corresponding limitations — and no amount of skilled modeling can compensate for source data that wasn't captured correctly.

This is why point cloud capture and registration is not a commodity step. The decisions made on site — scanner placement, ground control strategy, coverage planning for above-ceiling spaces — directly determine what's possible in the modeling phase. When something goes wrong late in a Scan to BIM project, it almost always traces back to a decision made during capture.

Step 1: Pre-Mobilization Planning

Before a scanner arrives on site, there's a scoping conversation that should define several things: the project coordinate system, the required Level of Development by discipline, the software platform for delivery, access constraints, and what areas of the building are in scope. This conversation shapes every subsequent decision.

LOD specification is particularly important to establish upfront. LOD 200 and LOD 300 models look significantly different from each other in terms of geometric precision and element detail, and they require proportionally different amounts of modeling time. A scope that changes from LOD 300 to LOD 350 mid-production isn't a minor adjustment — it's a material change to the deliverable.

Access constraints are equally worth mapping before mobilization. Above-ceiling plenum spaces, mechanical rooms, and locked areas all affect what the point cloud can capture and therefore what the model can contain. Identifying those limitations before the field crew arrives allows for a realistic scope, rather than discovering gaps after the fact.

Step 2: Laser Scan Capture

On site, the scanner is positioned at multiple locations throughout the building to ensure complete coverage of all surfaces required by the project scope. Each position is called a scan station, and the number of stations required depends on building geometry, room sizes, and the complexity of MEP systems in scope.

For a typical commercial building, scan stations are set at intervals of roughly 5 to 10 metres, with additional stations added in complex areas like mechanical rooms, stairwells, and anywhere line-of-sight between stations is limited. The goal is to ensure every surface that will be modeled is captured from at least two stations, reducing shadow zones and providing redundancy for registration.

Targets or spheres are placed throughout the space to aid registration. Each station captures the same targets from different positions, giving the registration software reference geometry to align stations into a coherent dataset.

Step 3: Point Cloud Registration

Registration is the process of aligning all individual scan stations into a single unified point cloud. Software compares the shared geometry between overlapping stations — typically the targets placed during capture, supplemented by cloud-to-cloud matching across overlapping areas — and computes the transformation needed to align each station to a common coordinate system.

A well-registered point cloud has consistent positional accuracy throughout, with no visible misalignment between stations at surface boundaries. Registration quality is assessed by reviewing residual errors at control targets and visually inspecting alignment at wall corners, column edges, and other sharp geometric features where misalignment is immediately visible.

The registered cloud is then cleaned — removing scan artifacts, noise, and non-permanent objects like furniture and equipment that should not appear in the as-built model — and exported in the format required for BIM production.

Step 4: BIM Modeling from the Point Cloud

With a clean, registered point cloud linked into Revit, the modeling team begins building the BIM model by tracing and snapping elements to the point cloud geometry. Walls are placed at their scanned face positions. Column centerlines are derived from the scanned column geometry. Duct centerlines follow the point cloud surface through the plenum.

The key discipline here is modeling to what is actually in the point cloud, not to what the existing drawings suggest should be there. Where the scan shows a wall that is 80mm off from the drawing, the model reflects the scanned position. Deviations from the original design intent are flagged in the model scope report, not silently corrected.

Modeling progresses discipline by discipline — typically architectural first, then structural, then MEP — with periodic coordination reviews between disciplines to catch conflicts before they compound.

Step 5: Internal QA Review

Before the model reaches the client, it goes through a structured QA process. Every modeled element is reviewed against the source point cloud for positional accuracy. LOD compliance is checked category by category against the agreed specification. File health is validated — purging unused families, resolving warnings, confirming coordinate alignment — so the model arrives ready to use rather than requiring cleanup before it can be linked into the coordination environment.

This step is where shortcuts taken earlier in the process surface. A model built carelessly from a mediocre point cloud will fail QA repeatedly; a model built carefully from a well-captured dataset typically passes in a single review cycle.

Step 6: Delivery and Handoff

Final delivery includes the BIM model in the agreed format and version, the source point cloud in a format suitable for your software environment, and a model scope report documenting scan coverage, LOD achieved by element category, coordinate system, software version, and any areas where physical access limitations affected model completeness.

The scope report is not a formality — it's the document that tells your design and coordination team exactly what the model contains and where its limitations are. A Revit model without that context is harder to use confidently, particularly for renovation and retrofit projects where the model's accuracy directly informs procurement and construction decisions.

If your team is using the model for MEP coordination, a brief onboarding review with the production team is worth scheduling. Walking through the dataset together before coordination begins surfaces questions early, when they're easy to answer, rather than mid-coordination, when they slow everything down.

What Makes the Difference Between a Good Scan to BIM and a Great One

The technical workflow is the same from project to project. What separates high-quality Scan to BIM deliverables from average ones is the rigor applied at each step: how well the capture was planned, how cleanly the point cloud was registered, how faithfully the model reflects the scan rather than the drawings, and how thoroughly the QA process was run before delivery.

For teams evaluating Scan to BIM vendors, those are the right questions to ask. Not just what software they use or what LOD they can deliver, but how they handle coverage gaps, how they document deviations, and what their QA process actually looks like in practice.