TL;DR
TL;DR: ±5mm (pencil-width) should mean registered accuracy — the assembled dataset verified against control — not the sensor spec. It comfortably serves floor plans, LOD 200–350 BIM, and MEP coordination; buy 1–2mm tripod zones only where fabrication demands it. Always ask: "registered, and verified how?"
Every scanning proposal quotes an accuracy number, and ±5mm is the one you'll see most from professional mobile-LiDAR providers, including us. But the number alone tells you little: accuracy claims differ in *what they cover*, *how they're verified*, and *whether your deliverable can even use them*. This guide translates the spec into decisions — what ±5mm buys, when it's plenty, and when to pay for more.
What is ±5mm accuracy in physical terms?
About the width of a standard pencil, or the thickness of three credit cards. If a scan says a wall opening is 914mm wide, the true value is between 909 and 919mm. Compare the alternatives: careful tape measurement of a building is typically trusted to about ±1/2 inch (±13mm) — and degrades badly over long runs, around obstructions, and above ladders — while photo-based capture commonly lands in the ±20mm class. Survey-grade scanning is not slightly better than manual as-builts; it's a different category of evidence.
What does the accuracy number actually cover?
Three different things get called "accuracy," and good providers are explicit about which they mean:
- Sensor accuracy — what the scanner measures at the pulse level. Marketing numbers live here.
- Registered accuracy — the accuracy of the assembled dataset after individual scans are registered into one coordinate system. This is the number that matters, because it's the accuracy of the data you'll use. Our ±5mm is a registered figure, checked against control.
- Deliverable tolerance — how faithfully the model or drawing follows the cloud. At LOD 300, our Revit model faces hold ±10mm to the registered cloud, documented in the accuracy report.
The chain matters: a great sensor sloppily registered produces bad data with a good spec sheet. Ask any provider the question that separates the tiers: *"Is that registered accuracy, and how is it verified?"* — more on this in the provider buyer's checklist.
How is scan accuracy verified?
Against control: independently measured reference points distributed through the building. After registration, the cloud is compared to those references and the residuals documented in a registration report that travels with the delivery. This is why scan data holds up in disputes and claims — its provenance is checkable, not asserted. On projects requiring state plane coordinates or tie-in to site survey, a control network does the same job at higher rigor.
When is ±5mm enough — and when isn't it?
±5mm comfortably serves the overwhelming majority of building documentation:
- 2D floor plans and BOMA calculations — drawn at 1/8" or 1/4" scale; the drawing can't even display ±5mm.
- Revit models at LOD 200–350 — modeling generalization absorbs it easily.
- MEP coordination — clash tolerances typically sit in the 10–25mm range, so ±5mm capture leaves the whole budget for the model.
- Renovation design, as-builts, verification questions, digital twins.
Pay for more when the deliverable is fabrication: steel connections, prefabricated assemblies cut to the scan, and monitoring work justify 1–2mm-class terrestrial scanning in the zones that need it — the platform-choice guide covers that hybrid. The efficient pattern is mobile capture at ±5mm building-wide, tripod zones where tolerance demands.
Does higher accuracy always mean a better project?
No — it means a more expensive one. Accuracy beyond what the deliverable can consume is pure cost: a floor plan drawn from 1mm data looks identical to one drawn from 5mm data. The professional move is matching the accuracy chain to the decision — capture, registration, and deliverable tolerance all specified in a written scope, then verified in the report. That's the difference between buying a number and buying evidence. Our scope template shows exactly where the accuracy lines go.