Guide
What is CMM inspection, and when should a purchase order require a CMM report?
CMM inspection is the measurement of a part on a coordinate measuring machine, which touches the part with a probe at many points, builds the measured features in software, and compares their size, location and form with the drawing in the datum reference frame the drawing defines. A purchase order should require a CMM report when the drawing carries position, profile or other datum-related geometric tolerances that hand gauges cannot check, and it should say whether that report is wanted on the first article only or on every lot.
By Saola Manufacturing engineering team · Published · Updated
Most pages about CMM inspection are written for the metrologist who runs the machine. This one is written for the engineer who draws the part and the buyer who writes the purchase order. It covers what a CMM is good and bad at, how the machine's alignment relates to the datums on your print, how to read the report a supplier sends back, and how to word the requirement so you get the evidence you need without paying for measurements nobody will read. A CMM report, a dimensional inspection report and a first article inspection report overlap but are not the same document, and the difference matters when you write the PO.
On this page
What does a CMM actually measure, and how does it work?
A CMM measures the coordinates of points on the part surface and lets software fit geometric features to those points. Every result on a CMM report, whether a diameter, a hole position or a profile deviation, is calculated from those points rather than read directly from a scale.
The common machine is a bridge CMM: a granite table, a moving bridge, and a probe head that travels along three perpendicular axes. A touch-trigger probe registers a point each time its stylus tip deflects against the surface, while a scanning probe stays in contact and collects a dense line of points. The software then fits a plane, cylinder, circle, cone or free-form surface to each group of points and reports the result. For this kind of machine, the length-measurement acceptance and reverification test is ISO 10360-2:2009. Portable arms, laser trackers, optical scanners and CT systems also produce coordinate data, but they have their own test parts in the ISO 10360 series, and on small machined features their uncertainty is often larger than a bridge CMM's.
Two consequences follow for a buyer. First, the result depends on how many points were taken and where, so a hole checked with a handful of points tells you less about its form than a scanned one. Second, the result depends on the fitting algorithm. Many CMM programs default to a least-squares fit, which averages the points, while the size and datum definitions in ASME Y14.5-2018 are based on the actual mating envelope and on contact with the high points of the surface. On a part with good form the difference is small. On a bowed face or an out-of-round bore it can decide whether a feature passes.
Which drawing callouts need a CMM, and which are better checked another way?
A CMM is the right tool for position, profile, and orientation tolerances that refer to datums, and for any feature whose location must be related to other features in three dimensions. It is the wrong tool, or only a partial one, for thread fit, surface roughness, thin flexible walls, tiny edge breaks, and tight size tolerances on simple diameters.
The table below matches common callouts to the method that gives the most trustworthy answer. It is the list to check before you write "100 percent CMM" on a purchase order, because a CMM report that includes thread and roughness lines does not prove those features are good. Threads are accepted with the gauges named in the thread standards, roughness with a stylus profilometer reporting the profile parameters defined in ISO 21920-2:2021, and the CMM is reserved for the geometry that only it can relate back to the datums.
| Callout on the drawing | Best method | Why |
|---|---|---|
| Position of a hole pattern to datums | CMM, or a functional gauge at MMC for production | Needs the datum reference frame and true-position math |
| Profile of a surface or line | CMM with scanning, compared to CAD | Hundreds of points against a nominal model |
| Perpendicularity, parallelism, angularity to a datum | CMM, or surface plate and indicator | Either works; the CMM documents it in one setup |
| Flatness of a large face | CMM scan, or surface plate and indicator | Sparse touch points can miss a local high spot |
| Runout and total runout | Indicator with the part rotated on its datum axis | The definition is a rotating measurement; a CMM only approximates it |
| Tight size on a turned diameter or bore | Micrometer, bore gauge, air gauge | Two-point size with lower uncertainty than a fitted cylinder |
| Internal or external thread fit | GO and NO-GO thread gauges | Thread acceptance is functional; a CMM cannot assess flank fit |
| Position of a tapped hole | CMM probing a threaded locating plug | Locates the thread pitch axis, not the minor diameter |
| Thin or flexible walls | CMM with low probe force, or non-contact scanning | Probe force and fixturing can deflect the wall |
| Surface roughness Ra or Rz | Stylus profilometer | A CMM stylus tip is far too large to resolve roughness |
| Edge breaks, small radii, chamfers | Optical comparator, vision system, radius gauge | Features too small for reliable touch probing |
How does a CMM alignment tie back to the datums on the drawing?
The CMM builds a part coordinate system from the datum features in the order the feature control frame lists them, so for a typical prismatic part the primary datum establishes a plane, the secondary orients the part within that plane, and the tertiary removes the last remaining freedom to fix the origin. If the programmer aligns to different features, or in a different order, every position and profile result on the report is measured in the wrong frame.
On a gauge or fixture, datum feature simulators are physical: a granite plate contacts the high points of datum A, a pin or rail locates datum B, and a stop locates datum C, which is how ASME Y14.5-2018 establishes the datum reference frame. A CMM has no plate. It probes points on each datum surface and constructs the datums mathematically, so the result depends on whether the software simulates the high-point contact the standard describes or simply fits a least-squares plane through a bowed surface. On a flat, machined datum face the two agree closely. On a cast or warped datum face they can disagree enough to change a position result, which is why datum choice on cast parts is worth a separate look.
Three questions catch most alignment problems when you review a report. Does the report header name the datum features and the order used for the alignment? Were datum features of size simulated at the material condition the drawing calls for, at MMB, LMB or RMB? And where a datum feature of size is referenced at MMB, did the report apply the datum shift the drawing allows, or did it ignore it and reject parts a functional gauge would accept? A supplier who answers these in the report header saves a dispute later.
How do you read a CMM report?
Read each line as nominal, tolerance, actual and deviation, then check the out-of-tolerance flag, and only trust the line if it maps to a ballooned characteristic on your drawing. A report that lists features by the software's internal names, with no balloon numbers, cannot be audited against the print.
A typical CMM printout gives, for each characteristic, the nominal value from the drawing or CAD, the upper and lower tolerance, the measured actual, the deviation from nominal, and a flag or bar graph that shows how much of the tolerance was used. For geometric tolerances the "actual" is the tolerance value the feature consumed, such as the diameter of the position zone, and the report should also show any bonus tolerance applied. The header should state the part number and revision, the program name, the alignment, the date, the operator, and the machine with its calibration status, so the measurement can be traced per the approach on NIST's metrological traceability page.
| Balloon and characteristic | Nominal and tolerance | Actual / deviation | How to read it |
|---|---|---|---|
| 4: bore diameter | 12.000 mm +0.018 / 0 | 12.011 / +0.011 | In tolerance; uses about 60 percent of the band |
| 7: position, 4x holes to A|B|C | ⌀0.10 mm at MMC | ⌀0.12; allowed 0.14 with bonus 0.04 | Passes only because bonus tolerance applies; check the hole size |
| 9: profile of surface to A|B | 0.20 mm | 0.23 | Out of tolerance; flag should be raised and a deviation report opened |
| 12: flatness of datum A | 0.05 mm | 0.048 | In tolerance but at the limit; ask how many points were taken |
Two habits make reports easier to use. Ask for the deviation and the percentage of tolerance used, not just pass or fail, because a part family that keeps using most of a band is a process drifting toward rejects. And ask that any result close to a limit is judged with the measurement uncertainty in mind, which is what ISO 14253-1:2017 addresses with its decision rules: measurement uncertainty counts against whoever is proving conformity, so a supplier's reading just inside the limit, with an uncertainty band that crosses it, does not prove a pass.
Is a CMM accurate enough for your tolerance?
Usually yes for general machining tolerances, but not automatically for tight ones. Check the machine's stated maximum permissible error from its ISO 10360 acceptance test against your tightest tolerance, and remember that the task-specific uncertainty on your part is larger than that figure.
CMM makers state length measurement error as a maximum permissible error, written E0 or E150 in ISO 10360-2:2009 and in its US counterpart ASME B89.4.10360.2-2008, usually as a constant plus a term that grows with the measured length. Probing error is tested separately under ISO 10360-5:2020. Those figures describe the machine on calibrated artifacts in a controlled room. Your part adds its own error sources: temperature, fixturing, stylus choice, number of points, form error on the feature, and the fitting algorithm.
The working rule many quality engineers use is that the measuring process should consume only a small fraction of the tolerance, the same idea behind the 10:1 rule for gauges. When a tolerance is so tight that the CMM's uncertainty becomes a large share of it, a dedicated gauge, an air gauge or a micrometer on a temperature-stable part often gives a more defensible answer. For a gauge repeatability study on a CMM program, the same acceptance logic used for gage R&R applies.
When should a purchase order require a CMM report: first article only, or every lot?
Require a full CMM report on the first article and after any change to design, process, tooling or location, and require CMM data on every lot only for the characteristics that are critical to function or that a cheaper gauge cannot check. Writing "CMM report every lot" for every dimension adds programming and machine time to each shipment without adding much information once the process is proven.
The cost of a CMM requirement has two parts. Programming and proving the CMM routine is a one-time cost per part number and revision, and it is incurred anyway if a first article report is required. Running the program and writing the report is a per-part cost that repeats on every lot and grows with the number of characteristics and the sample size. That second part is what an every-lot, every-dimension requirement multiplies. The first article rules in SAE AS9102C already define when a full or partial re-inspection is due, which is a useful model even outside aerospace.
| Situation | What to require on the PO | Report scope |
|---|---|---|
| New part number or new supplier | Full CMM-based first article report against a ballooned drawing | Every characteristic, with gauge and profilometer results for the lines a CMM cannot check |
| Drawing revision, new tooling, process or site change | Partial (delta) first article report | Characteristics affected by the change |
| Repeat lots, stable process, datum-related critical features | CMM data on named critical characteristics, sampled per lot | Balloon numbers listed on the PO, sample size stated |
| Repeat lots, only size and simple features | Dimensional inspection record from hand gauges | No CMM report needed |
| Safety-critical or regulated part | CMM data on every part for the critical characteristics | Serialized results, uncertainty considered at the limits |
On the PO or the RFQ, say four things in words: which balloon numbers need CMM data, whether that is first article only or every lot, the sample size per lot, and the report format you want, such as the native CMM printout, a spreadsheet, or AS9102 forms. When we review a drawing for CNC milling or CNC turning, those four answers decide how the inspection plan is built and which features get a gauge instead of a probe. The quality page lists the first article and lot inspection documents available, and if a CMM report in a specific format is a hard requirement, put it on the RFQ so it is answered before the quote rather than at your dock. When the drawing is ready, send it for a quote.
Frequently asked
What does CMM stand for?
Coordinate measuring machine. The machine logs where its probe touches the part in three axes, then computes holes, faces and other features from those touch points. CMM inspection is the use of such a machine to check a part against its drawing or CAD model.
Is a CMM report the same as a first article inspection report?
No. A first article report accounts for every characteristic on the drawing, plus materials, processes and specifications, and it pulls in results from whatever instrument suits each feature. A CMM report covers only what the CMM measured. It is often the largest single attachment to a first article report, but not a substitute for it.
How do you check a CMM report you received from a supplier?
Match every report line to a balloon on your drawing, confirm the part revision and the datum alignment in the header, look for any flagged line, and look hard at results that use most of their tolerance. Then confirm that threads and roughness were checked by gauge and profilometer rather than reported from the CMM.
Can a CMM measure surface finish or threads?
Not in a way that proves conformance. Roughness needs a stylus profilometer, and thread fit needs GO and NO-GO thread gauges. A CMM can locate a threaded hole by probing a locating plug screwed into it, which is the right way to report the position of a tapped hole.
Does a CMM report need a calibration certificate behind it?
It should be traceable. Ask for the machine's last calibration or reverification date on the report header, and for the certificate on request. Reverification to ISO 10360-2 or ASME B89.4.10360.2 is how the machine's stated error is confirmed in service.
Sources
- ISO 10360-2:2009, Acceptance and reverification tests for coordinate measuring machines (CMM), Part 2: CMMs used for measuring linear dimensions, ISO, 2009
- ISO 10360-5:2020, Acceptance and reverification tests for coordinate measuring systems (CMS), Part 5: Coordinate measuring machines (CMMs) using single and multiple stylus contacting probing systems, ISO, 2020
- ASME B89.4.10360.2-2008, Acceptance Test and Reverification Test for Coordinate Measuring Machines (CMMs), Part 2, ASME, 2008
- ISO 14253-1:2017, Inspection by measurement of workpieces and measuring equipment, Part 1: Decision rules for verifying conformity or nonconformity with specifications, ISO, 2017
- ASME Y14.5-2018, Dimensioning and Tolerancing, ASME, 2018
- SAE AS9102C, Aerospace Series – First Article Inspection Requirements, SAE International, 2023
- ISO 21920-2:2021, Surface texture: Profile, Part 2: Terms, definitions and surface texture parameters, ISO, 2021
- NIST Policy on Metrological Traceability, NIST, accessed October 2026