CMM Inspection Report Explained: What Each Section Tells You

• 1
  What a CMM Inspection Report Actually Is
• 2
  Header Data: Part, Program, and Machine Setup
• 3
  Datum Reference Frames and Alignment Reporting
• 4
  Nominal, Actual, Deviation, and Tolerance Columns
• 5
  Reading GD&T Results: Position, Profile, and Bonus Tolerance
• 6
  Out-of-Tolerance Flags, Probing, and Uncertainty
• 7
  Report Sections and What They Decide

A CMM inspection report is the output of a coordinate measuring machine program run against a specific drawing revision. Each inspected feature appears as a line: nominal value, measured actual, signed deviation, tolerance band, pass or fail. The report is a record of one measurement run, not a certificate of quality on its own.

A first-article report often runs 5 to 40 pages because it covers every dimensioned feature on the print. A production report is usually shorter and covers the features the drawing marks as critical, plus the features the process is known to drift on. Ask which kind you are looking at before you read a single value.

The machine matters less than the setup behind the numbers. A 1.8 μm machine with a bad alignment produces worse data than a 3.5 μm machine aligned to the drawing datum sequence. So the first thing to read is not the table. It is the header and the alignment block.

One more boundary: a CMM report describes geometry at the moment of probing, in a room at a stated temperature. It does not describe how the part behaves under load, after heat treatment, or after coating. Read it as one snapshot of size and form.

The header block decides whether the numbers mean anything. Look for part number and revision, drawing revision, program name and revision, machine serial, probe tip diameter and material, stylus length, and the temperature the part was measured at. Standard practice is 20 °C ± 2 °C for part and room, with a soak long enough for the part to reach that temperature.

Probe tip diameter is usually a 2 to 5 mm ruby sphere. That value changes the reported result, because the machine compensates for the tip radius in software. A report that states a 3 mm tip while the rack actually held a 2 mm tip is wrong by roughly 0.5 mm in radius on every touch point.

Check that the program revision matches the drawing revision. A program built to drawing C and run against drawing D will report the wrong nominals, and every deviation on that page inherits the error. This is the single most common reason two suppliers disagree on the same part.

Also look for the calibration date of the probe and the machine. A stylus that was crashed and re-qualified without a fresh calibration can shift results by a few micrometres. Small number, big argument on a ±0.005 mm feature.

The alignment section shows which datum reference frame was used and how it was built. A typical report states the primary datum plane, the secondary datum axis or plane, and the tertiary clocking feature, plus the residual error of each. Those residual errors tell you how well the physical part matched the theoretical datum.

Watch for two things. First, an alignment built from the CAD model instead of the drawing datums. Second, a best-fit alignment used where the drawing calls for a specific datum sequence. A best-fit alignment splits error across all features, so position results look better than they are.

If the drawing calls out datum A as a plane, datum B as a bore and datum C as a slot, the report should show that order. An alignment that uses the bore as primary when the drawing says plane primary will rotate the whole coordinate system and move every position value.

Residual errors are worth reading. A primary plane residual of 0.002 mm on a ground face is normal. The same residual on a rough cast face suggests the datum itself is unstable, and the position results downstream should be read with that in mind.

The core table is usually nominal, actual, deviation, upper tolerance, lower tolerance, and status. Deviations are signed, and the sign convention depends on feature type. For a bore diameter, a positive deviation means the hole is larger than nominal. For a position callout, the reported value is often a diametral true position, which is twice the radial distance from the true location.

Confirm which convention the report uses before you argue about a number. A 0.08 mm diametral position is a 0.04 mm radial offset. Mixing the two turns a passing part into a failing one on paper. Most CMM software prints the convention in a column header or a legend, and many reports print both radial and diametral values.

Tolerance columns should match the drawing exactly, including any material condition modifiers. If the drawing says Ø10.00 ± 0.05 with a 0.1 mm position at MMC, the report should show the same modifiers. A report that quietly drops MMC is conservative on size and misleading on position.

Read the status column last. It summarizes the comparison, but it cannot tell you whether the comparison was set up correctly. A green row on a misaligned program is still a wrong row.

GD&T results need careful reading. Position is commonly reported as a diametral zone value, and bonus tolerance from MMC or LMC modifiers is added to the stated tolerance. A 10.00 ± 0.05 mm hole at MMC with a 0.1 mm position tolerance and an actual size of 10.04 mm gains 0.04 mm of bonus, giving 0.14 mm total.

Profile results are usually reported as a total zone, not a plus/minus value. A 0.1 mm profile tolerance means the actual surface must sit inside a 0.1 mm band centered on the true profile, so the maximum allowed deviation in either direction is 0.05 mm. Reading profile as ±0.1 mm doubles the allowance and hides real error.

For a surface profile callout without a datum reference, the result is form only. With datums, it is location and form together. The same 0.06 mm number means very different things depending on which type of callout sits on the drawing.

Bonus tolerance is not free size. It is a trade: the further the feature sits from MMC, the more position error the drawing allows. A hole at LMC has the most bonus and the least remaining wall thickness, so check both columns, not just the position result.

A single flagged feature among hundreds may be a probing artifact: a burr on a hole edge, chips under the part, a loose clamp, or a stylus that touched a fillet instead of the intended face. Re-measure the flagged feature alone, ideally with a different probe angle or on a different machine, before scrapping the part.

The opposite pattern matters more. A feature that passes but sits at 80 to 90 percent of its tolerance on every part in a run is a process signal, not a good result. Trend those values across lots and you will see tool wear or thermal drift long before the first failure.

Measurement uncertainty is rarely printed but always present. A bridge CMM in a controlled room carries a maximum permissible error around 1.8 to 3.5 μm plus roughly 3 to 5 μm per metre of travel, depending on machine class. Add probing form error, stylus bending on long extensions and fixturing repeatability, and a realistic uncertainty on a 0.05 mm tolerance is often 10 to 20 percent of the band.

Probing strategy changes results more than most people expect. A hole measured with 4 points gives a diameter but no form information. Eight to 12 points on two levels reveals taper and ovality. A plane measured with 3 points can be tilted by one high spot, shifting every downstream position result.