What accuracy do I actually need?
Less than you think, usually. And the honest answer is that it depends on the part, the environment, and what decision the measurement is going to drive.
Accuracy is a budget, not a target. Every step tighter costs equipment, time, controlled conditions and patience. Buying more of it than the job needs is one of the commonest ways people waste money on 3D scanning.
Start from the tolerance, not the scanner. Metrology’s long-standing rule of thumb is that your measuring system should be roughly ten times finer than the tolerance you are checking. On a ±0.5 mm tolerance that puts you around 0.05 mm, which handheld scanning does comfortably. On ±0.02 mm it does not, and you are into a CMM. Plenty of shops work to 4:1 rather than 10:1, so treat it as a starting point for the conversation, not a law.
Then look at where the part will be measured. This is the one people forget. Every dimension on every technical drawing applies at 20 °C — that is ISO 1, and it has been the reference temperature since 1931. Steel moves about 11.7 microns per metre per degree. Aluminium moves about twice that. An aluminium housing 400 mm long, measured on a shop floor at 28 °C rather than 20 °C, is roughly 74 microns longer than the drawing says it should be — which has already blown a ±0.05 mm tolerance before anyone has touched it with a scanner.
So chasing sub-0.1 mm accuracy in an unheated shop in February, or in a foundry, or outdoors in the sun, is chasing a number the environment will not let you keep. That is not a reason to skip the scan. It is a reason to be honest about what the scan can be used to claim.
And look at the condition of the part. A worn, corroded, cast or hand-finished part is not its drawing any more. Scanning a worn bushing to 0.02 mm captures the wear beautifully — you have measured erosion, not design intent. Valuable when wear is what you are studying; close to useless when you were hoping to recover the original dimensions. Same with sand-cast surfaces, weld spatter, filler, paint build-up and anything that has been in service for thirty years: the surface variation is already larger than the accuracy you were arguing about.
The question we will ask you. Not “how accurate do you want it?” — everybody answers “very” to that. The question is what you are going to do with the number. Machine a mating part, quote a repair, prove a warranty claim, print a display model, check for deflection, or archive a shape before it is gone? Each has a different right answer, and some of them do not need a scanner at all.
Accuracy, resolution and repeatability are three different things
They get used interchangeably in sales material and they should not be.
Accuracy is how close the measurement is to the truth. Resolution is the smallest feature the scanner can distinguish — how fine the detail is, not how correct it is. Repeatability is whether you get the same answer twice.
A scanner can have beautiful resolution and be consistently wrong. It can be repeatable and inaccurate, which is the dangerous combination, because consistency reads as confidence.
And on any handheld scan there is a further figure that matters more than the headline one: accuracy over distance, the error that accumulates as you track across a large object. A machine quoted at 0.1 mm point accuracy might add 0.3 mm for every metre you travel. On a bracket that is irrelevant. On a twelve-metre hull it is the only number that matters. Compare the figure that applies to the size of thing you actually scan.
Can you scan a worn part and give me the original dimensions?
Not from the scan alone, no. A scan records the part as it is today, including thirty years of wear, corrosion, denting and whatever the last person to repair it did.
What can be done is reconstruct intent: symmetry, nominal radii, whole-number dimensions, standard thread and bearing sizes, and the unworn areas of the part that tell you what the worn areas used to be. That is a reverse engineering job done by a person who knows what they are looking at, and the scan is the raw material for it rather than the answer.
If an original drawing exists, even a bad one, bring it. If a second, less worn example of the part exists, that is often worth more than any amount of extra scanning accuracy.
