iPhone LiDAR accuracy: measurement checks
Editorial update:
Check an iPhone LiDAR result by comparing selected dimensions with independently measured reference dimensions, using the same endpoints and units. Keep repeated raw readings, calculate the signed and absolute difference from the reference, and decide whether the result is suitable for the intended task. There is no universal centimetre or inch tolerance established for Spaceoverlay by this guide.
This Spaceoverlay editorial method is a practical comparison record, not a calibration procedure or measurement certification. The reference instrument and the way it is used also have uncertainty. A small observed difference does not establish the true error of either measurement. The completed worksheet below is entirely fictional and demonstrates arithmetic and decisions, not app performance.
Choose the decision before choosing a threshold
Write down what the dimension will be used for: a preliminary room discussion, a furniture clearance check, a formal survey or a construction decision are different tasks. Identify the spans that could change that decision. A doorway needed for access may deserve individual attention even when the overall room outline looks plausible. Do not approve every dimension from one convenient wall comparison.
Ask the person responsible for the task to define the acceptance rule before seeing the results. Record the allowable difference, unit, reason and escalation route, or record that no threshold has yet been agreed. Any such threshold is that assessor’s task-specific choice, not an Apple specification or a Spaceoverlay tolerance. Where a formal survey or specialist assessment is required, this informal worksheet does not replace it.
Define a reference you can explain
Mark or describe the two endpoints of each span and the height or plane at which you measure. “Room width” is ambiguous if one reading includes a recess and the other stops at the main wall face. Note whether you are comparing finished surfaces, structural boundaries or an opening. Use one consistent length unit; the metre, symbol m, is the SI unit of length.
Record the independent reference instrument, its identifier and its known calibration or check status. If that status is unknown, say so. Follow the instrument’s instructions and document the positioning method; an unsuitable reference procedure can undermine the comparison. Take repeat reference readings rather than treating one hurried observation as exact truth. Investigate disagreement between those readings before declaring the scan wrong.
Keep the scan context and raw observations
Record phone model, iOS version, app version, operator and date, together with the room identifier and scan file. Describe relevant conditions such as an obstructed corner, glass, a reflective surface, an open connection or a change in lighting. These are observations to preserve for investigation, not a claim that any one condition always creates a particular numerical error.
Repeat the capture or reading using the same defined endpoints and retain each result. State whether a repetition is a new scan, a new measurement of the same model or simply another view of the same displayed number. Those are different observations. Keep the displayed dimension and the exported artifact identifiable: a framework model, an app label and a printed drawing are different stages that should not be silently substituted for one another.
For a new Spaceoverlay capture, use a LiDAR-equipped iPhone with iOS 26 or later. A saved scan contains geometry and a model image, with a plan available after at least two walls are captured, not a room photograph. Keep independent property photography separate. If a required dimension is absent or the geometry is incomplete, write “not available” and arrange a further measurement; never enter zero to fill the cell.
Blank comparison record
Use the first table as a cover record for the observations. Copy it into your own file and complete the unknowns before making an acceptance decision. Include the reference readings themselves alongside the instrument record, not merely a statement that a tape or laser was used. Keep the original scan and the comparison sheet together so another reviewer can reconstruct what was compared.
| Record | Entry to complete | Owner or follow-up |
|---|---|---|
| Purpose and acceptance rule | Task; critical spans; threshold and unit, or not agreed; reason | Assessor and escalation contact |
| Device and software | Exact phone model; iOS; app version; scan file | Operator and date |
| Reference | Instrument and ID; calibration/check status; method; all raw readings | Reference operator; unresolved uncertainty |
| Room and endpoints | Room ID; endpoint descriptions; measurement plane; lighting and obstacles | Identify inaccessible or ambiguous spans |
| Repeat method | New scan or same-model rereading; observation IDs; units | Keep every observation, including failed checks |
Calculate differences without losing their direction
Signed difference equals the scan reading minus the reference reading. A positive value means the scan reading is larger; a negative value means it is smaller. Absolute difference is the magnitude of that signed value. For a reference of 4.000 m and a scan reading of 4.030 m, the signed difference is +0.030 m and the absolute difference is 0.030 m. This is a difference from the reference, not a claim of true measurement error.
Keep enough digits to reproduce the comparison, but do not infer instrument accuracy from the number of decimal places printed in a table. Compare repeat observations individually before considering a summary. An average can conceal a failed critical span or cancel positive and negative differences. If the access opening fails its declared criterion, an apparently good average across other walls does not clear that opening.
| Span / observation | Reference m | Scan m | Signed difference m | Absolute difference m | Decision / action |
|---|---|---|---|---|---|
| Defined endpoints; observation 1 | Enter raw value | Enter raw value or not available | Scan − reference | Magnitude of difference | Compare with agreed task rule |
| Same endpoints; observation 2 | Enter repeat reference | Enter repeat scan | Scan − reference | Magnitude of difference | Keep discrepancy and investigate |
| Critical opening | Enter raw value | Enter raw value or not available | Scan − reference | Magnitude of difference | Escalate unresolved clearance |
Fictional worked comparison: do not average away the doorway
The following numbers are invented. In this fictional exercise, assessor Alex selected an absolute-difference limit of 0.020 m solely for a preliminary comparison task. That choice is not advice for your task and is not a product tolerance. The fictional record identifies a LiDAR phone, iOS 26 and an app version field, but leaves the actual model, version and reference instrument unknown; it therefore cannot support a real acceptance decision.
Both wall observations exceed Alex’s illustrative limit, in opposite directions. The doorway differs by 0.050 m and is held for an independent recheck. Alex keeps all three rows, checks the endpoint definitions and reference procedure, and requests another observation. Changing the limit after seeing these results would change the decision rule rather than resolve the discrepancy.
| Span / observation | Reference m | Scan m | Signed difference m | Absolute difference m | Decision / action |
|---|---|---|---|---|---|
| Wall A / scan 1 | 4.000 | 4.030 | +0.030 | 0.030 | Over illustrative limit; recheck endpoints |
| Wall A / scan 2 | 4.000 | 3.975 | -0.025 | 0.025 | Over illustrative limit; retain both readings |
| Door opening / scan 1 | 0.900 | 0.850 | -0.050 | 0.050 | Hold clearance decision; independent recheck |
Read published studies within their actual setup
Apple’s RoomPlan overview explains camera-and-LiDAR room capture and model output; it does not supply a general Spaceoverlay measurement tolerance. A 2024 HBRC Journal paper compared an iPad Pro 2021 with a Trimble TX8 terrestrial scanner in a university hall, using static and moving acquisition approaches. Its hardware, scene, acquisition and comparison method belong to that study.
Those study results cannot establish accuracy for every iPhone, every app or your room. We do not transfer a published numerical result into an app guarantee. Use the paper to ask better questions about the reference, acquisition method and evaluation, then document your own observations. The separate scanning-mistakes guide covers capture preparation; this guide covers the independent comparison and the decision record after capture.
Close the check with a bounded decision
Finish with one of three clear outcomes: suitable for the stated preliminary purpose, repeat a named observation, or refer the decision to the appropriate measurement professional. Record who made that decision and which spans remain unknown. Do not turn “no discrepancy observed in these rows” into a statement that every wall, opening and exported label has been verified.
If an apparent failure disappears after correcting units or endpoints, preserve the original row and explain the correction. If it remains unresolved, keep the relevant dimension out of decisions that depend on it. A useful check leaves the next person with evidence, limits and an action, rather than a confident accuracy label unsupported by the room record.
Sources
- Apple Developer: RoomPlan overview — checked
- HBRC Journal (2024): study of iPad Pro and terrestrial scanner comparison — checked
- NIST: SI base units, including the metre — checked