Calculate flooring material from a laser scan (LiDAR) or 3D scan

How to capture a room so the scan is worth measuring from

A take-off is only as good as the scan under it, and most unusable scans fail for the same three reasons: the room was walked too fast, the light was too flat, and the loop was never closed. A laser-scanning app on a phone or tablet builds the room by tracking its own position frame by frame; every time it loses that track it guesses, and the guess shows up later as a wall that bends or a floor that slopes.

Walk the room once around the perimeter at roughly the pace of a slow stroll, holding the device at hip height and angled slightly down so the floor–wall junction stays in view. That junction is the line the boundary will be traced along, so it is the one surface that has to be captured cleanly. Come back to your starting point and overlap the first few feet again: that closes the loop and lets the app correct accumulated drift instead of leaving it in the mesh.

• Turn the lights on and open the blinds — the laser scanner handles darkness, but the camera track that positions it does not.
• Keep at least three feet of clear floor along the walls; a scan taken over stacked boxes records the boxes, not the floor.
• Move furniture you can move, and scan the gap behind what you cannot — the floor still has to be laid under it.
• Scan glass, mirrors and gloss doors from an angle; head-on they return the reflection instead of the surface.
• For rooms longer than about 25 ft, walk the perimeter twice rather than once — drift grows with distance, not with time.

Check the scale before you trust the area

Scan apps report dimensions with a confidence they have not earned. A scan that drifted by two percent over a 20 ft room is off by nearly 5 in on the long wall, which is a plank and a half — enough to change the pack count and to make the cut list wrong at the far end. The check that catches it takes thirty seconds and a tape measure.

Before you leave the room, measure one long wall by hand and write the number down. After you import the scan into Autolay, trace that same wall and compare. If the two agree within a half inch or so, the whole scan is trustworthy, because scale error is global: it does not affect one wall and spare the rest. If they disagree, rescan rather than scale the model — a stretched scan hides its error in the corners, where the cuts are.

The second check is squareness. Trace the outline and look at the corners the room is supposed to have at 90°. A scan that closed its loop badly leaves them at 87° or 93°, and every course of planks then walks out of parallel with the wall as it crosses the room.

Which export format to pick: GLB, OBJ or PLY

Most scanning apps offer several exports and none of them explain the trade-off. Autolay reads glTF/GLB, OBJ and PLY point clouds, so the choice is about what survives the trip, not about compatibility.

GLB is the default worth using. It packs geometry and textures into one binary file, loads fastest in the browser viewer, and carries the metric scale the scanner recorded. OBJ is the safe fallback when an older app writes malformed glTF, but it splits textures into side files that are easy to lose and it stores coordinates as text, so the same room can be five times the file size. PLY is the right pick when you want the raw point cloud rather than a reconstructed mesh — useful for rough or cluttered rooms where the app's meshing invents a floor that is not there.

• GLB — one file, textures included, fastest to load. Use this unless something breaks.
• OBJ — widest app support, but keep the .mtl and texture files next to it.
• PLY — the unreconstructed points. Best when the mesh looks smoothed over furniture or debris.
• Decimate to roughly 1–2 million points before export if your app offers it; beyond that the viewer gets slower without the boundary getting more accurate.

Tracing the floor boundary: what counts as floor

The boundary you trace is the area that gets material, so the decisions at the edge are the ones that move the number. The rule that holds in almost every case: trace to the finished wall surface, then let the expansion gap and the baseboard cover the tolerance. Tracing to the plaster line behind a baseboard over-orders; tracing to the front face of the baseboard under-orders, because the plank has to run under it.

Fitted kitchens are the exception worth deciding deliberately. A kitchen run on a plinth usually does not get flooring underneath, so the boundary stops at the plinth line — but if the units may ever be replaced, laying through gives you a floor that survives the next kitchen. Islands, by contrast, almost always sit on the finished floor, so they are traced as an inclusion, not a hole.

Doorways are the third edge case. Flooring normally runs to the middle of the door leaf, where the threshold profile sits, so extend the boundary half a leaf thickness into the opening rather than stopping at the frame. Across a whole apartment that adds up: eight doors at 16 × 36 in is more than 30 sq ft that a boundary stopped at the frames would have left unordered.

From boundary to order: what the waste percentage actually covers

Once the boundary is traced, Autolay lays the material course by course inside it and counts what it had to cut. The waste percentage that comes out is that count expressed against the floor area — the offcuts the plan could not place anywhere else. It is not a safety margin, and this is the distinction that a flat 10% rule of thumb blurs.

Two things still belong on top of it. Breakage and mislaid boards on site are a handling allowance, not a geometry number, and a pack or so of spare stock for later repairs is a decision about the future rather than about this room. Both are worth adding consciously; neither is what the planner is measuring.

The plan's real payoff is the cut list. It names each cut piece, its length, and which offcut it comes from, so the fitter can work down the list instead of re-deciding at every course — and the pieces that were meant to be reused at the far wall actually get reused, which is where the computed waste percentage stops being a number on a screen and starts being material that was not bought.