In favorable conditions, short nearby measurements can land within roughly a centimeter. A complete room or corridor model may differ from a reference scan by several centimeters, and a long, rushed, reflective, or poorly closed scan can drift much more.
That sounds less satisfying than “accurate to one inch,” but it is the useful answer. Phone LiDAR can save a contractor real time when the workflow respects what the sensor is good at. Trouble starts when one laboratory result gets stretched into a guarantee for every room, surface, app, and walking path.
What the phone actually produces
Apple’s RoomPlan framework combines the camera and LiDAR sensor to create a parametric room model: walls, doors, windows, openings, recognized objects, and dimensions. It is not simply pointing a laser at one wall. Software fits geometry to a stream of depth and camera observations while the user moves.
That creates three accuracy questions:
- How noisy is one nearby depth observation?
- How well does the software fit a wall or object to many observations?
- How much does the complete model drift as the phone moves through space?
A careful test can produce a very good answer to the first question without proving that a casual five-room walkthrough will be equally good. Contractors usually care most about the third.
What independent studies found
A controlled object study of the iPhone 12 Pro reported roughly ±1 centimeter accuracy for objects larger than 10 centimeters. Performance degraded for smaller objects, and point density fell with distance. That is strong evidence that the sensor can be precise nearby; it is not a warranty on fitted wall lengths or room area. See the open-access paper in Scientific Reports (retrieved July 29, 2026).
Whole-model tests tell a more practical story. In a peer-reviewed corridor study, iPhone 14 Pro meshes averaged about 3.0 to 5.2 centimeters from terrestrial-laser reference data, depending on the app and capture path. Some sequences were worse, and round trips occasionally suffered serious drift. The authors make clear that acquisition method matters. See the full study in Sensors (retrieved July 29, 2026).
In one indoor hall, moving-capture errors could exceed 25 centimeters on features longer than 4 meters. A deliberate set of overlapping static scans produced 1–2-centimeter linear differences, while area differences remained around one square meter. That gap between “careful static campaign” and “walk while talking” is exactly why a single headline number misleads. See the HBRC Journal study (retrieved July 29, 2026).
| Measurement situation | What the evidence supports | How to use it |
|---|---|---|
| Nearby object or wall patch | Centimeter-class performance is possible under controlled conditions. | Useful evidence, but verify the dimension if the tolerance matters. |
| Single residential room | Often useful for layout, openings, and planning quantities when capture is steady. | Build the Scope of Work and spot-check a known or longest dimension. |
| Long room, hallway, or multi-room loop | Model drift becomes more important and can exceed several centimeters. | Break the job into smaller scans and reconcile shared openings. |
| Glass, mirrors, dark or thin surfaces | Depth and object detection can degrade or fail. | Capture alternate angles and verify manually. |
| Fabrication or regulated work | No universal phone tolerance is published by Apple. | Use the required trade measurement, survey, template, or stamped document. |
Where errors come from
Distance and angle
Depth noise rises as the phone gets farther from a surface, and a shallow angle gives the sensor less reliable geometry. A railway-interior experiment saw noise rise notably beyond about 2.5 meters and found field-use accuracy around 4–5 centimeters in suitable scenarios. Abrupt motion and finite edges also hurt. See the open-access case study (retrieved July 29, 2026).
Surface material
Mirrors and full-height glass can confuse both depth sensing and room interpretation. Dark, glossy, transparent, repetitive, or textureless areas can weaken tracking. A clean white wall may fit beautifully; a glass shower enclosure beside a mirror may not.
Walking path and closure
The phone estimates its own motion as it scans. Small pose errors accumulate. Maintaining overlap within a short room scan gives the software recognizable geometry, but long round-trip paths can be especially prone to drift. Split larger jobs into zones instead of assuming that returning to the starting point will fix the model.
What the app does after capture
Two apps using the same phone can return different geometry because their reconstruction, wall fitting, filtering, and export pipelines differ. A raw point cloud, a mesh, and a RoomPlan-style parametric wall model should not be evaluated as if they were the same artifact.
Good uses on a renovation job
- Establishing rooms, wall runs, doors, windows, and major fixtures for a Scope of Work.
- Planning paint, drywall, flooring, trim, demolition, and protection allowances.
- Attaching spatial context to walkthrough narration and photos.
- Finding a missed room or opening before the estimate leaves the driveway.
- Giving an estimator a structured starting point that can be reviewed and corrected.
A 2026 RoomPlan-based scan-to-model study reported strong layout overlap and much faster acquisition than a terrestrial scanner, while still noting wall-thickness error and positional drift. Its conclusion is a useful fit for contractors: mobile scanning is compelling for initial retrofit assessment where speed matters more than high precision. See the ISARC proceedings.
Measurements that still deserve another tool
- Countertop, shower glass, cabinetry, millwork, windows, or other tolerance-critical fabrication.
- Property boundaries, elevations, structural movement, or anything requiring a licensed survey or engineer.
- Stamped-plan dimensions and code-critical clearances.
- Long diagonals, high ceilings, reflective rooms, and dimensions that look wrong when the model closes.
- Final material orders when a small miss creates a large cost.
The right rule is not “LiDAR is accurate” or “LiDAR is inaccurate.” It is: the consequence of error determines the verification method.
A contractor-friendly capture protocol
- Scan one coherent space at a time. Split long paths into rooms or zones with overlap at doors.
- Move steadily. Avoid fast pans, sudden turns, and long shots from across the room.
- Show the corners and openings. Give the model multiple angles on doors, windows, and wall intersections.
- Mind the room. Add light where practical and recognize glass, mirrors, tall walls, clutter, and occlusion as warning signs.
- Maintain short-range overlap. Revisit recognizable geometry within each room, but split long round trips into zones rather than relying on one large loop.
- Spot-check one known dimension per room. Use the longest wall or the dimension with the greatest cost consequence.
- Rescan instead of rationalizing. If parallel walls visibly diverge or a room does not close, treat the artifact as suspect.
No LiDAR? The walkthrough still measures
Apple ships the LiDAR sensor only on Pro-model iPhones, but that is not the price of entry. On phones without the sensor, Biltum runs a Standard Video Scan: the same narrated walkthrough, with measurements produced from the video and photos instead of a depth sensor. LiDAR sharpens the geometry when it is available; the everything-else discipline in this article—spot-check the longest dimension, verify anything tolerance-critical—applies identically to both capture paths.
How Biltum uses the scan
Biltum treats the scan as one layer of evidence. The Scope of Work also uses the contractor’s narration, photos, and review. Scan measurements land as editable per-line values—each with a unit picker and stepper increments—so the scan is evidence, and the contractor stays the editor. The contractor remains the person deciding what work is included and which dimensions need verification before pricing or ordering.
That is the practical advantage of phone LiDAR: not a magic tape measure, but a fast spatial backbone for the information contractors already collect on a first visit. The same record earns its keep later, when a mid-job change needs to reference what the accepted scope actually said.
Frequently asked questions
How accurate is iPhone LiDAR for room measurements?
There is no single number. The controlled iPhone 12 Pro study above found roughly ±1 centimeter on nearby objects larger than 10 centimeters. Whole-room meshes from an iPhone 14 Pro averaged about 3.0 to 5.2 centimeters from terrestrial-laser reference data, and a rushed moving capture exceeded 25 centimeters of error on features longer than 4 meters. Careful, short, overlapping scans landed at 1–2 centimeters on linear dimensions.
Which iPhones have LiDAR?
Apple ships the LiDAR sensor on Pro-model iPhones, starting with the iPhone 12 Pro. The independent studies cited in this article tested the iPhone 12 Pro and iPhone 14 Pro.
Can I use it for takeoffs and ordering materials?
Yes for planning quantities: paint, drywall, flooring, trim, demolition, and protection allowances. For final material orders where a small miss creates a large cost, spot-check the controlling dimension with a tape first. The consequence of error determines the verification method.
What should I still tape-measure?
Anything tolerance-critical or expensive to get wrong: countertops, shower glass, cabinetry, millwork, and windows; code-critical clearances and stamped-plan dimensions; long diagonals and high ceilings; and the longest dimension in each room, as a spot-check on the scan.
Method note: This article distinguishes sensor tests from full-model tests and links to Apple documentation and primary research. Results from one device, app, room, or capture protocol should not be generalized into a guaranteed tolerance.
