A coordinate measuring machine (CMM) measures a physical part by touching or scanning points on its surface, recording each as an X-Y-Z coordinate, and comparing those coordinates against the CAD model and its GD&T tolerances to judge whether the part conforms. It turns a metal part into numbers a computer can check against the drawing.
A caliper tells you the width of one feature. A CMM tells you where hundreds of features sit in three-dimensional space relative to each other and to the datums on the print, true position of a bolt circle, profile of a curved surface, perpendicularity of a bore. That is why it sits at the center of most serious dimensional inspection, and why understanding what it can and cannot do keeps an inspection plan honest.
What is a CMM and what does it do?
A CMM is a machine that finds the exact location of points on a part and reports them as coordinates. Most shop-floor CMMs are the bridge type: a rigid granite base gives a dead-flat reference plane, and a probe rides on three perpendicular axes, X, Y, and Z, so it can reach any point in the measuring volume. When the probe contacts the part, the machine reads the position of each axis and records that point as a coordinate. Collect enough points on a feature and software fits geometry to them: three points define a plane, a ring of points defines a circle's diameter and center, and so on.
From there the CMM does the part a caliper never could: it compares measured geometry to the nominal CAD model and the tolerances on the drawing, then reports the deviation for every characteristic. The output is not "1.502 inches." It is "true position of this hole is 0.08 mm, tolerance is 0.10, pass," across a whole feature list, in one report. That is what makes a CMM the backbone of first article inspection and complex-part verification.
How does a CMM actually measure a part?
By building a coordinate system from the part's own datums, then measuring features against it. A CMM does not care where the part sits on the table. The routine first probes the datum features named on the drawing, a face, a bore, a slot, and uses them to establish the part coordinate system, the same reference frame the designer used. Every measurement after that is reported in that frame, so results match the print no matter how the part was clamped.
The measuring itself is a loop: move the probe to a target, contact the surface, record the coordinate, repeat. The software already knows how many points each feature needs and where to take them, from a program written against the CAD model. One thing that trips people up: the probe ball has a real diameter, so the machine touches at the ball's surface but must report the part surface. Probe calibration, qualifying the stylus on a reference sphere, solves this by telling the software the ball's exact size and behavior, so it compensates for the offset. Skip or botch that qualification and every measurement carries a hidden error, which is why CMM work leans hard on the same measurement system analysis discipline as any other gauge.
Touch-trigger vs. scanning probes: which do you need?
Touch-trigger probes record one discrete coordinate at each point of contact; scanning probes drag along the surface and collect a dense, continuous stream of points. The choice follows what you are measuring, not which is "better."
- Touch-trigger is the workhorse for prismatic features, hole positions, bore diameters, plane-to-plane distances, angles. It takes a handful of well-placed points per feature and moves on, which is fast and more than enough when you need location and size.
- Scanning earns its keep on form and profile: the roundness of a bore, the profile of a turbine blade or a molded contour, the flatness of a large face. Capturing thousands of points along the surface reveals shape that a five-point touch would miss entirely.
Research comparing the two probing methods has found no statistically significant difference in measurement uncertainty between them for the features both can measure, so for a hole position, either works, and speed and programming effort decide. Reach for scanning when the characteristic is the shape of a surface, not the location of a feature.
How accurate is a CMM?
A well-maintained shop CMM commonly measures to a few microns, but the honest answer is a number, not a boast: accuracy is stated by a length-measurement error formula the manufacturer publishes and an acceptance test proves. Under ISO 10360-2 a CMM's permissible error for size measurement is expressed as a value plus a term that grows with length, error gets larger over longer distances, which is why a machine that nails a 25 mm gauge block can drift over a 600 mm span.
Several things eat into that accuracy in real life: temperature (steel grows with heat, so serious CMM rooms are climate-controlled to around 20 °C), probe qualification quality, part fixturing, the number and placement of points, and the operator's programming. A CMM is only as good as its calibration and its environment. Treat it like the precision instrument it is, controlled room, scheduled calibration, and probe qualification every session, and it earns its accuracy claim. Treat it like a bench tool and it will quietly hand you wrong numbers with impressive decimal places.
How do you run a CMM inspection from scratch?
A repeatable CMM inspection follows the same order every time, whether it is a first article or a routine audit.
- Start from the drawing and CAD. Identify the datums, the characteristics to measure, and their tolerances. The print, not the machine, defines the job.
- Qualify the probe. Measure the reference sphere to calibrate the stylus tip diameter and behavior for this session, so the software compensates correctly.
- Fixture the part and let it soak. Hold it without distorting it, and let it reach room temperature before measuring, a part straight off a warm machine will read out of tolerance for a reason that has nothing to do with the part.
- Establish the part coordinate system. Probe the datum features and build the reference frame the drawing calls out.
- Measure the features. Run the program, taking the planned points on each feature; the software fits geometry and evaluates each characteristic.
- Evaluate against GD&T and report. Compare measured values to nominal and tolerance, flag anything out, and generate the report, often the dimensional evidence attached to a first-article or incoming inspection record.
- Feed results back. Trend the numbers over time so a slow drift toward a limit is caught before it becomes scrap, the same way you would read a process on a control chart.
What is ISO 10360 and why calibrate a CMM?
ISO 10360 is the family of standards that defines how a CMM's performance is accepted and reverified, so an accuracy claim is a tested fact rather than a slogan. ISO 10360-2 covers acceptance and reverification for size measurement along the axes; ISO 10360-5 covers the probing system, including the errors that show up when you use multiple stylus tips. The acceptance test proves a new machine meets its spec; the reverification test lets you confirm it still does, periodically, over its life.
Calibration and periodic reverification matter because a CMM drifts like any instrument, bearings wear, temperature cycles, a crash knocks a probe out of true. Without a scheduled check, you cannot know whether today's "pass" is real. And the calibration only means something if it traces to a national standard through an accredited lab, which is the whole point of measurement traceability, and why the accreditation of that lab matters as much as the certification of your own plant. Keep the as-found and as-left records from each calibration, the same way you would for any gauge; our guide to as-found/as-left calibration covers why those two readings both matter.
The standards behind a trustworthy measurement
- ISO 10360-2:2009 specifies acceptance tests that verify a CMM's size-measurement performance against the manufacturer's stated error, plus reverification tests for periodic in-house checks.
- ISO 10360-5:2020 defines acceptance and reverification tests for contacting probing systems, including multiple-stylus errors evaluated on a reference sphere.
- The U.S. National Institute of Standards and Technology maintains the dimensional metrology references and calibration services that shop-floor traceability ultimately chains back to (NIST Dimensional Metrology).
CMM vs. hand tools: when is a CMM worth it?
Use a CMM when the characteristic is location, orientation, or profile in 3D, and hand tools when it is a single size on a simple feature at high volume. A caliper or micrometer is faster, cheaper, and lives at the machine, which is exactly right for the operator checking a bore diameter every tenth part. It cannot tell you the true position of that bore relative to two datums, or whether a molded surface follows its profile tolerance. That is CMM territory.
The practical rule: put the fast, few-feature, high-volume checks on gauges at the point of use, and reserve the CMM for first articles, complex GD&T, and periodic full layouts where measuring many features against a common datum frame is the job. Both live in the same inspection plan; the skill is matching each characteristic to the cheapest tool that can actually measure it, which is the same thinking behind choosing attribute versus variable inspection. Where Harmony helps is downstream of the measurement: capturing those inspection results and gauge checks digitally at the station so they become trends you can search and act on, not binders nobody reads, see how that plays out in our CLS case study.