How Accurate Are Polycam Scans?
Polycam has a standard accuracy of +/-½ inch on standard interior captures. This article explains the main factors that affect accuracy, so you know what to expect and how to get the best results from your scans.
Accuracy is determined by three interacting variables: what you're measuring, how the space was captured, and which capture mode you used. Each variable is covered in its own section below. Throughout this article, we report deviations as absolute distances in millimeters or inches against a known reference, because that's the figure you can actually act on for your specific use case.
This article focuses on Space Mode and Floorplan Mode captures. Object Mode models are made from photos and are not dimensionally accurate unless you use the Rescale tool to match them to a real-world measurement.
How Mobile LiDAR Measures Space
Polycam uses the LiDAR sensor built into iPhone Pro and iPad Pro devices. The sensor emits pulses of infrared light and measures the time each pulse takes to return. That return time converts into a distance, and thousands of these distances per second accumulate into a point cloud: a spatial map of the surfaces in the room. The point cloud is then processed into a 3D mesh and, in Floorplan Mode, a 2D plan.
Two consequences of this pipeline drive everything that follows. First, measurements are only as good as the captured geometry, so anything that degrades the point cloud (too little signal, too few points, conflicting returns) degrades the measurement. Second, the reconstruction engine fills gaps by estimating, which helps produce a continuous-looking model, but means under-captured areas are interpolated rather than measured.
Methodology
All Polycam measurements were compared against a professional laser distometer, treated as ground truth. Reported deviations are the absolute difference between the Polycam value and the corresponding laser distometer reading for the same dimension. Rooms were scanned across combinations of lighting, scanning speed, and scanning distance to isolate the contribution of each variable.
What Are You Measuring?
What you get from a scan depends on your goal. Floorplan Mode is designed for documenting interiors and provides a 2D floor plan and a 3D model, so most people use it for wall-to-wall measurements, ceiling heights, and room areas. Space Mode is more flexible and can be used for many purposes, from interior as-builts to capturing objects, so the accuracy you need will depend on your project. The next sections explain how different measurements and surfaces behave, so you can set the right expectations.
Room Dimensions and Ceiling Height
Wall-to-wall measurements and ceiling heights are where Polycam performs most consistently, because walls and ceilings are large, flat surfaces that give the sensor an abundant signal. Ceiling height is the strongest case: it captures a single large, unobstructed horizontal plane above the scan path. For typical interior rooms, wall-to-wall measurements generally land within roughly half an inch to two inches of the laser reference, depending on room size and conditions. Ceiling height typically matches the reference exactly or within a few millimeters.
Room Size and Accumulated Error
Smaller rooms are more forgiving than larger ones. In a room of roughly 100 ft² or less, a single session captures every surface with sufficient density and overlap, keeping errors small and consistent regardless of conditions. As the room size grows, small per-surface errors have more room to accumulate across the sweep of the capture.
For larger spaces, particularly those over 5,000 ft² such as open-plan offices, warehouses, or retail floors, users should expect a wider margin of variation in overall dimensional accuracy. The type of space matters too. An open, uncluttered floor plan gives the sensor clear sightlines and consistent surfaces to work with. A cluttered space with lots of furniture, equipment, or obstacles limits coverage and forces the reconstruction engine to estimate more frequently. Spaces with shiny or reflective floors and walls, such as polished concrete, glazed tile, or glass partitions, pose extra challenges because reflective surfaces scatter the sensor's infrared signal rather than returning it cleanly.
For large spaces where dimensional accuracy is critical, use good lighting, a slow and deliberate capture pace, and thorough coverage of all walls and corners to achieve the best results.
Small and Thin Objects
Space Mode and Floorplan Mode are optimized for capturing rooms and architectural spaces, so the sensor prioritizes large surfaces like walls, floors, and ceilings. Smaller or thinner objects, such as chair legs, curtain rods, lamp stems, railings, and narrow pipes, may not appear with full detail in the final model. This is a level-of-detail consideration rather than an accuracy issue, because the room dimensions, wall measurements, and floor areas will still be accurate even if the fine details of individual objects aren't fully captured.
If you need highly detailed captures of specific objects or smaller items, Object Mode is a better fit. You can then use the Rescale tool to align the model with a known real-world measurement and get an accurate, detailed result for that object.
Surface Type
Surfaces that absorb or scatter infrared produce less reliable geometry than surfaces that reflect it cleanly. Two categories cause most of the trouble:
- Highly reflective surfaces: polished concrete, shiny tile, metal, and glass scatter pulses in multiple directions instead of returning them cleanly. The sensor receives conflicting depth readings, leading to distorted or missing geometry. Reflective floors are a common source of floor-mesh distortion.
- Dark absorptive surfaces: black or very dark walls and ceilings absorb infrared rather than reflecting it. This is manageable under good lighting, but in dim rooms with dark surfaces, the sensor may receive too little signal to reconstruct those areas reliably.
How Was the Space Captured?
Capture technique was the single largest variable in the study, larger than the underlying hardware. Scanning the same room under different conditions showed that the difference between a rushed scan and a careful scan exceeded that between a careful scan and the laser reference. How you capture the space matters more than anything else, and it's also the factor you fully control.
Scanning Distance
The sweet spot for interior scanning is 3 to 6 ft (1 to 2 m) from surfaces. Getting too close blurs images and limits the sensor's angular coverage, producing drift artifacts at wall-floor corners. Getting too far reduces accuracy progressively, with the sensor's effective range for reliable geometry topping out at around 16 ft (5 m).
Capture Speed
Moving too quickly is the most common cause of diminished accuracy. The sensor needs time to accumulate enough points on each surface. A rushed scan results in sparse, uneven coverage, which harms both the visual output and the measurements. Both Space Mode and Floorplan Mode will prompt you to slow down when they detect excessive speed.
Surface Coverage
Gaps in the point cloud force the reconstruction engine to estimate, and the most consequential gaps are at the room's defining edges:
- The wall-floor junction. This corner defines where a wall ends and the floor begins. Miss it, and the engine estimates that boundary, which affects both wall geometry and room dimensions.
- All four walls. Capturing only the walls you care about leaves an open model that the system can't close into an accurate floor plan.
- Wall-to-wall corners. These are the room's spatial anchors. Pass them by too quickly, and there's not enough data to define the boundary accurately.
Walking over the same area more than once can affect both how your model looks and its accuracy. Scanning the same spot again can create conflicting data that the system has to sort out, which can lower model quality. For best results, plan your route ahead of time and move through each area once in a steady, careful pass.
Lighting Conditions
LiDAR emits its own infrared, so ambient light doesn't power the sensor directly, but it matters indirectly. The photographic textures captured alongside the depth data feed the reconstruction pipeline, and low light degrades both those textures and the algorithm's ability to align scan data. More importantly, dark surfaces absorb infrared. In a dim room with dark walls, low ambient light, and absorptive surfaces, reconstruction can fail outright. A test room with a black ceiling and dark blue walls produced near-complete mesh failure in unlit conditions, yet a functional model with all lights on.
The tool supplements ambient light for darker corners and surfaces. It's useful when scanning two dark surfaces that meet in a corner: it simultaneously identifies the area most likely to lack signal and the area most important for accurate room geometry.
Multi-Room and Extended Captures
For single rooms and smaller spaces, Polycam delivers consistent accuracy within the +/-½-inch tolerance described above. As the size and complexity of a capture grow, users should expect a wider margin of variation in overall dimensional accuracy. This is a characteristic of all spatial scanning systems, not something unique to Polycam: as the sensor travels farther and covers more surfaces, small errors across individual segments accumulate over the full extent of the capture.
When using Extend mode to capture several connected rooms in a single session, measurements entirely within a single segment will generally be more accurate than those spanning multiple segments. For large spaces where overall building dimensions are important, we recommend treating each room or zone as its own reference area and, where possible, verifying critical cross-room measurements against a known reference.