How Accurate Is Laser Scanning? Understanding Tolerance and RMSE
Published 18 July 2026
"How accurate is your scan?" is one of the most common questions we get, and it deserves a more specific answer than "very." Accuracy isn't one number — it's measured at three distinct points in the workflow, and conflating them is the most common way a client ends up with the wrong expectation.
The three numbers that actually matter
| Measurement | What it tells you | Typical value |
|---|---|---|
| Point cloud ranging error | How precisely the scanner measures a single point | ±1 mm (FARO Focus S150 spec) |
| Registration RMSE | How well the whole point cloud fits the Total Station control network | ≤ 3 mm on our projects |
| Deliverable tolerance | How accurate the final CAD/Model/BIM output is, once modeling judgment is added | Varies by LOD tier — see each service page's accuracy table |
The first number — ranging error — is a hardware spec: how precisely the scanner itself measures the distance to a single point. It's the best-case accuracy of the raw measurement, before registration or modeling enter the picture. It's a real number worth knowing, but it's not the number that describes your finished drawing or model.
Why registration RMSE is the number that actually matters most
RMSE — root mean square error — measures how well the registered point cloud fits the control network it was tied to. This is where our hybrid Total Station + FARO workflow earns its keep: because every scan is registered against surveyed control points rather than aligned scan-to-scan alone, RMSE typically stays at or below 3 mm on our projects, even across a large multi-building site. A point cloud with excellent ranging-error specs but poor registration can still drift badly across a large site — RMSE is what actually tells you whether that happened.
Does higher LOD mean higher accuracy?
- Not exactly — LOD describes how much detail and information an element carries, not how precisely it was measured.
- A LOD 200 massing model and a LOD 400 fabrication model can both trace back to the same registration accuracy; the difference is how much of that accuracy is expressed as modeled geometry.
- Deliverable tolerance does typically tighten at higher LOD tiers, because fabrication-grade work has less margin for error than early massing does — see each service page's accuracy table for the specific numbers per tier.
What this means when you're evaluating a quote
If a vendor quotes you "accuracy" as a single number without saying whether it's ranging error, registration RMSE, or deliverable tolerance, ask which one — the three numbers can differ by an order of magnitude, and the one that actually determines whether your drawing or model is trustworthy is registration RMSE, not the scanner's spec sheet. It's a fair question to ask on any scoping call, ours included.
Questions about this article
Which of the three accuracy numbers should I ask for in a quote?
Can RMSE vary across a single project?
Does weather or lighting affect scan accuracy?
How is deliverable tolerance different from registration RMSE?
Should I be worried if a quote doesn't mention RMSE at all?
Scope your project in one call.
Tell us the site, the deliverable, and the deadline — we'll tell you the tier and turnaround that fits.