What 3D scanner accuracy figures actually mean

Almost every 3D scanner on Amazon leads with a single number: 0.1 mm, 0.05 mm, 0.02 mm, sometimes 0.01 mm. The numbers get smaller as the price goes up, which makes them look like a straightforward ranking. They are not. They measure different things, under conditions you will not reproduce, and two scanners quoting the same figure can behave very differently on the same object.

This is desk research, not a lab test — we read specifications and listings rather than measuring scanners ourselves. But the specifications themselves tell you more than most buyers realise, once you know which words matter.

Point accuracy is the headline number

The figure in the title is usually single-frame point accuracy: how closely one captured point sits to the true surface, within a single frame, near the middle of the sensor’s working distance, on a matte test object under controlled light. It is the kindest possible measurement of a scanner, which is why it is the one that gets printed on the box.

So when a listing says 0.02 mm precision, read it as “under ideal conditions, on one frame, in the sweet spot of the sensor”. It is not a promise that a finished scan of your object will be dimensionally correct to 0.02 mm anywhere.

Volumetric accuracy is the number that matters

A scan is not one frame. It is hundreds or thousands of frames stitched together, and every join contributes a tiny error. Those errors accumulate across the object, so a scan that is locally beautiful can still be a millimetre out end to end. The specification that describes this is volumetric accuracy, and it is quoted as a figure plus a drift term — something like “0.02 mm + 0.08 mm/m”.

Read that as: 0.02 mm of local error, plus a further 0.08 mm for every metre of object size. On a 100 mm bracket the drift term adds almost nothing. On a car bumper or a motorbike fairing it dominates completely, and it is the only part of the figure you should care about.

You will notice that volumetric accuracy appears mostly on the more expensive machines — the Creality Raptor Pro and Sermoon range quote it, and the EinScan and Revopoint Metro lines do too. Budget scanners generally quote point accuracy alone. That absence is itself informative: a scanner sold without a volumetric figure is being sold for small objects, whatever the pictures of cars in the listing suggest.

Resolution is a different thing entirely

Resolution — also called point distance or point spacing — is how far apart captured points are, typically 0.05 mm to 0.5 mm. Accuracy is how right each point is. The two are independent, and confusing them is the single most common mistake in scanner buying.

• High resolution, poor accuracy gives a dense mesh of points that are all slightly in the wrong place. The scan looks detailed and measures badly.
• Lower resolution, good accuracy gives a sparser mesh that is dimensionally trustworthy. Fine for a part you need to fit something to; disappointing if you wanted surface texture.

Which you want depends entirely on the job. Reverse-engineering a mounting plate needs accuracy. Capturing a carved ornament for display needs resolution. Very few projects genuinely need both at once, and the scanners that deliver both are the ones at the top of the price list.

Why marketing figures are best-case by construction

Nothing dishonest is happening — the figures are real measurements. They are just measurements taken in the conditions where the scanner performs best, and several everyday realities degrade them:

• Distance from the sensor. Accuracy is specified at an optimal working distance. Hold the scanner too close or too far and it falls off, often sharply.
• Surface colour and finish. Matte mid-grey is the best case. Black, glossy, translucent or metallic surfaces return far less usable signal, and the fix is matting spray rather than a better specification.
• Light. Structured-light scanners lose accuracy in bright ambient light; blue-laser scanners cope better outdoors, which is why they cost more.
• Movement. Every frame assumes the object held still relative to the sensor. Scanning a person, or a part on a wobbly table, costs accuracy immediately.
• Alignment method. Feature-based alignment on a smooth, featureless panel drifts badly. Marker dots or a turntable with a fixed reference restore it. See handheld vs turntable scanning for how the two approaches differ.

How much accuracy do you actually need?

For scanning intended for 3D printing, less than people assume. A typical FDM printer draws lines 0.4 mm wide in layers 0.2 mm tall, so detail finer than about 0.2 mm cannot be reproduced at all — a 0.1 mm scanner is already finer than the printer that will make the part. Resin printing resolves more, and that is where a 0.02 mm scanner starts to earn its price.

Rough guidance:

• Fitting a printed part to an existing object — 0.1 mm point accuracy is usually plenty, but check the volumetric figure if the object is large.
• Reverse-engineering something mechanical — 0.05 mm or better, and measure the critical dimensions with calipers anyway.
• Figurines, ornaments and display models for resin printing — prioritise resolution and colour capture over accuracy.
• Bodies, faces and anything that moves — the quoted accuracy is largely irrelevant; capture speed and frame rate decide the result.

The practical conclusion

Treat the headline number as a rough class indicator rather than a specification you can rely on. Compare the volumetric figure where one is quoted, check the working distance and minimum object size, and assume real-world results one or two steps worse than the marketing figure on anything dark, shiny or large.

And if a dimension must be right, verify it with calipers. That is true of every scanner on the market at every price. Next, see what each price band actually buys in 3D scanner price bands in the UK, what the mesh needs afterwards in scanning for 3D printing, or browse every 3D scanner we track with live prices.