Accuracy, precision and repeatability

You type 10.00 mm into a dimension and you feel like you've asked for something exact. The machine doesn't hear it that way. It hears a target it will aim at, hit slightly off, and hit slightly off again differently on the next part. Three separate things decide how close the printed part lands to your number, and people mush them into one word — "resolution" — which measures none of them. Pull them apart and you can finally answer the question that actually matters: how tight a clearance can I put in the model and still trust it?

Three different things, one target

The classic picture is a shooter aiming at a target.

• Accuracy is hitting the true value — the shots cluster around the bullseye. An accurate printer makes a 10.00 mm feature that measures 10.00 mm.
• Precision is repeating tightly — the shots land close together, whether or not that's on the bullseye. A precise printer makes ten 10.00 mm features that all measure the same, say 10.12 mm every time.
• Repeatability is getting the same result run to run — you print the part today, next week, after a filament swap, and it comes out the same. It's precision stretched across time and conditions rather than across one plate.

These come apart in every combination, and the combinations matter more than any single one.

Why the machine's precision caps your design

Here's the practical payoff. The tightest clearance you can reliably put in a model isn't set by the machine's advertised resolution — the smallest step its motors can command, often a few microns (sub-micron in Z), a number that means almost nothing for real parts. It's set by the machine's precision and repeatability: the spread of what actually comes off the bed. If two nominally identical holes vary by ±0.1 mm part to part, then a fit that needs to be right to ±0.05 mm is a coin toss, no matter how fine the advertised resolution. The honest tolerance you can design to is the scatter, not the step.

And the two failures are not equally curable, which is the whole reason to tell them apart:

• Precise but not accurate is the easy problem. The machine repeats tightly but lands consistently off — every hole comes out 0.15 mm undersize. That's a constant offset, and a constant offset you can calibrate out: measure it once on a test coupon, bake the correction into your dimensions, done. A repeatable error is a fixable error.
• Imprecise is the hard problem. If the machine scatters — this hole 9.9, the next 10.1, unpredictably — there's no single correction to apply, because the error is different every time. You can't calibrate away randomness. Your only defence is to open up the tolerance: design the fit with enough slack that even the worst part in the scatter still works. An imprecise machine forces loose clearances on you.

Find out where you stand

You don't guess at these numbers — you measure them. Print a test coupon with known features (a row of pins, a set of holes, a 20 mm cube) two or three times, measure with calipers, and read the three properties straight off: how far the average sits from nominal is your accuracy, how tightly the repeats cluster is your precision, and whether a fresh run reproduces it is your repeatability. That's the empirical clearance your design must live within, and it's exactly what Real printed clearances shows you how to turn into fits you can trust.

Machine behaviour isn't the only thing that decides what you can attempt, though — sometimes it's the air around the print. Next, Enclosures and multi-material.