In 3D printing, accuracy is how close a printed dimension lands to its nominal CAD value. Precision is how closely repeated prints of that same dimension agree with each other. A process can be precise but inaccurate, with every part oversize by the same amount, or accurate on average but imprecise, with parts scattered on both sides of nominal. The two errors have different causes and different fixes: a steady offset can be compensated, while scatter has to be reduced in the process itself.

The definitions here follow the measurement vocabulary of ISO 5725, applied by analogy to a manufacturing process. The examples come from masked LCD resin printing, where washing and UV post-curing are part of the dimensional result, but the logic holds for any additive process.

Accuracy, trueness and precision: what the terms mean

Everyday engineering talk uses "accuracy" for closeness to the target. ISO 5725 is stricter. It calls closeness of the average result to the reference value trueness, calls agreement between repeated results precision, and uses accuracy as the umbrella term covering both. So "accuracy vs precision" in the common sense is really trueness vs precision.

This matters when you read a specification. A single "accuracy" figure could describe the average offset, the spread, or a mix of the two, and the document rarely says which.

Precision also comes in two strengths. Repeatability is agreement under the same conditions: same printer, resin lot, operator and settings, over a short period. Reproducibility is agreement when those conditions change, such as a different machine, shift or resin lot. A process that repeats well within one build can still reproduce poorly across a month.

Term | What it describes | How it shows up in printed parts | What it does not tell you
Trueness (everyday "accuracy") | Closeness of the average result to nominal | Every bore runs small by a similar amount | Whether individual parts agree with each other
Precision | Agreement between repeated results | The same bore varies part to part | Whether the average is on target
Repeatability | Precision under unchanged conditions | Copies in one build, or back-to-back builds | Behavior across machines, lots or operators
Reproducibility | Precision under changed conditions | Same file on another printer or resin lot | Short-term stability
Resolution (XY pixel size, layer height) | The smallest step the system addresses | Edge stepping, fine detail | Either trueness or precision
Tolerance | The deviation the drawing allows | Accept or reject limit | Whether the process can hold it

Resolution and tolerance are in the table because they get confused with the other four. Pixel size is an input to the process. Tolerance is a requirement placed on it. Neither is a measured result, a point covered in more detail in our guide to pixel size vs print accuracy.

What offset and scatter look like in measurements

Print ten copies of a part with a nominal 20.00 mm boss and measure them all the same way. Two outcomes are common.

In the first, all ten read slightly high and sit close together. Precision is good and trueness is off. In the second, the readings average close to 20.00 mm but individual parts fall well above and below it. Trueness is good and precision is poor.

Neither outcome is visible from one part. A single sample that measures on nominal says nothing about spread, and a single sample that measures off could be offset or just an unlucky draw. That is why one good demonstration print is weak evidence for a production decision.

<p style="text-align: center"> <img src="../upload/YYYYMM/IMAGE-ID.png" alt="Identical resin 3D printed test coupons labeled by build position and measured for accuracy and precision" data-width="1600" data-height="900"/> </p>

What pushes a printed part off nominal

Offset comes from effects that act the same way every time. In resin printing the usual ones are these.

Resin shrinks as it polymerizes and shrinks a little more during UV post-cure, so a part measured after final cure is smaller than the same part measured green.

Light cures slightly beyond the edge of the exposed image. Outer features grow and holes close up. This is why bores often print undersize on the same part where bosses print oversize, and why a single global scale factor cannot correct both.

The first layers are deliberately over-exposed to bond to the build plate, which thickens the base of the part.

Scaling between the image and the build plane can be slightly off in X or Y.

Because these effects are stable when the process is stable, they respond to compensation: scale factors, hole and shaft offsets, or exposure tuning.

What makes printed parts disagree with each other

Scatter comes from conditions that drift or differ between parts.

Position on the platform is the first place to look. If light intensity varies across the build area, a part in the corner cures differently from one in the center. Our article on resin 3D printer light uniformity covers how to check this.

Resin temperature changes over a shift, and with it viscosity and cure response. Pigmented resins also settle if they sit unstirred.

Post-processing adds its own variation. Wash time, how completely the part dries, and where it sits in the curing chamber all move the final dimension. A part measured while it still holds wash solvent will not read the same once it has dried.

Then there is the measurement. Caliper pressure on a thin wall, or on any elastomer part, can change the reading by more than the process varies. If the gauge is not repeatable, the scatter you see may not belong to the printer.

No offset value fixes any of this. Scatter is reduced by controlling conditions.

How to measure both on your own part

1. Check the measurement first. Measure one part several times, ideally with two people. If those readings disagree, fix the method or fixture before judging the printer.

2. Pick the features that matter: the datums, bores, mating faces and wall sections that carry a tolerance on the drawing.

3. Print the same part at defined platform positions, including corners and center, and label each position.

4. Repeat the build on different days, with the resin, settings, orientation and supports unchanged.

5. Run the full wash, dry and UV post-cure sequence, and measure in that final state.

6. For each feature, calculate the average deviation from nominal and the spread, as a range or standard deviation. The average is your trueness figure. The spread is your precision figure.

7. Compare both to the tolerance band. The spread has to fit inside the band with room to spare, and the average has to sit near its middle.

A few builds give an indication, not a capability figure. Formal capability statements need more parts and a longer period than a sample trial provides.

Reduce the scatter first, then correct the offset

The order matters. If you compensate while the process is still scattering, you are correcting to an average that will move on the next build. Stabilize platform position effects, temperature, post-processing and the measurement method first. Once the spread is tight and stays tight, apply the offset correction and verify it on the real geometry.

The exception is when the spread already fits the tolerance comfortably. Then the offset is the only problem, and compensation is a quick fix.

How to read an accuracy figure on a spec sheet

A figure such as "± X mm" is only useful if you know how it was produced. Ask the supplier:

• Does it describe average deviation, part-to-part spread, or both?

• On what feature size and geometry was it measured?

• With which resin, and after which post-processing?

• Across how many parts, platform positions and builds?

• With what instrument?

YIDIMU's own specification tables list XY pixel size and a stated print accuracy as separate lines, because they are separate quantities. The second is a manufacturer-stated figure under the manufacturer's test conditions. It is not a tolerance for your part and it is not a precision figure. The same reading applies to any brand's data sheet.

For repeated builds, precision usually deserves more weight than a headline accuracy number, because offset can be corrected and scatter sets the floor on what tolerance you can hold. That is the central question in resin 3D printing for low-volume manufacturing.

Where this breaks down

Offset is not always one constant. It often grows with feature size, differs between X, Y and Z, and changes with wall thickness and orientation. A correction verified on a 10 mm bore may not hold on a 60 mm one, so check compensation on each class of feature you care about.

Flexible parts are harder still. An elastomer lattice has no single unloaded dimension unless you define how it is supported and measured, and that definition should be written down before the first reading.

Next step

Take one real part with toleranced features, print it at several platform positions over a few builds, and measure after final cure. If you want this run on your file before choosing equipment, send your model and tolerance requirements to the YIDIMU application team.