A drawing says a bore is 12.00 mm ±0.02. A supplier says the parts are good. Both statements are worthless until someone answers a third question: measured how, with what, and how much of the reading is the measurement system rather than the part? Metrology turns "looks fine" into a number you can hold a factory to, and most quality disputes are really disagreements about measurement method.

The practical decisions are narrow: which instrument for which feature, how often to measure, and when to demand a formal report instead of trusting the process.

The Three Tiers of Measurement

Hand gauges

Calipers, micrometers, bore gauges, pin gauges, and go/no-go fixtures. A quality caliper resolves 0.01 mm and a micrometer 0.001 mm, both fast, cheap, and usable at the machine. Their limit is that they measure size between two points and nothing else: a caliper cannot tell you whether a hole is round, whether two faces are parallel, or where a feature sits relative to a datum scheme. It also introduces operator variation — how hard someone squeezes the jaws moves the reading by several microns.

Custom go/no-go gauges deserve more use than they get. A machined plug that either enters a bore or does not gives an unambiguous pass/fail in two seconds and requires no skill. On a line checking the same feature ten thousand times, a $600 gauge outperforms a $200,000 machine.

Coordinate measuring machines

A CMM moves a probe through a calibrated three-axis volume and records point coordinates, from which software fits planes, cylinders, and circles and evaluates them against a datum reference frame. A bridge CMM in a temperature-controlled room typically holds 1.5 to 3 microns plus a length-dependent term; shop-floor machines and portable arms are looser, often 20 to 60 microns, but usable where the parts are.

The CMM is the only instrument that properly evaluates geometric characteristics — true position, perpendicularity, profile of a surface — because those callouts are defined relative to datums, and establishing a datum frame is exactly what a CMM does. If your drawing uses the symbols in GD&T basics, you have implicitly specified CMM inspection whether you meant to or not.

Touch probing is slow, a few seconds per point. Scanning probes drag along a surface collecting thousands of points and are far better for profile evaluation of curved forms.

Optical and non-contact systems

Vision systems image a part and measure edges in software, ideal for flat parts, thin stampings, and anything a probe would deflect. Structured-light and laser scanners capture a point cloud in seconds and compare it to the CAD model, producing a color deviation map that shows warp, sink, and shrink across a whole molded part. The same capture pipeline is used for reverse engineering a part.

Industrial CT scanning measures internal geometry, wall thickness, voids, and assembled parts without cutting anything. Accuracy is typically 10 to 50 microns and cost is high — often $800 to $3,500 per session — but for a sealed assembly it answers questions nothing else can.

Gauge R&R: Is Your Data Real?

Before arguing about a part, prove the measurement can be trusted. A gauge repeatability and reproducibility study has several operators measure several parts several times and splits total variation into part variation, repeatability, and reproducibility.

The interpretation is simple: if the measurement system consumes under 10 percent of the tolerance band it is acceptable; 10 to 30 percent is marginal; above 30 percent you are measuring noise. A related rule is that instrument resolution should be about a tenth of the tolerance — checking a ±0.02 mm feature with a caliper reading to 0.01 mm is borderline before operator effects.

This is where many overseas quality disputes actually live: two parties measure the same part, get different answers, and each concludes the other is dishonest. The fix is to agree the method — instrument, datum setup, fixture, temperature, point count — and write it into the inspection plan. Aluminum grows roughly 23 microns per meter per degree Celsius, so a part measured on a hot factory floor and again in your lab genuinely reads differently.

When to Require a Formal Inspection Report

Not every shipment needs dimensional data. These milestones do:

  • First articles. Before mass production, on the first parts from production tooling, with every drawing dimension measured. The discipline is covered in first article inspection, and for molded parts it coincides with the first mold trial samples.
  • After any change. New tooling, a repaired mold, a different machine, a new sub-supplier, a material substitution — each resets the process.
  • Periodically during production. A reduced report on the six to twelve critical characteristics, monthly or per lot.
  • Whenever a field failure has a dimensional hypothesis. Measure retained samples from the suspect lot against good lots.

A usable report has a ballooned drawing, a table listing each balloon with nominal, tolerance, measured value, and pass/fail, the sample size and cavity each part came from, the instrument used per characteristic with its calibration date, and a signature. A report that says only "conforms" is not a report.

From Inspection to Process Control

Inspection tells you whether parts are good. It does not tell you whether the process is stable, and a process drifting inside tolerance today will be outside it next month. Once you have a repeatable measurement method, feed the numbers into control charts and capability indices rather than filing them — the mechanics of that transition are in statistical process control.

Two other things belong in the same plan. Decide which dimensions genuinely matter, because measuring all of them is expensive and dilutes attention; the analysis in tolerance stack-up analysis tells you which features drive assembly and which can float. And remember that dimensional data says nothing about surface condition, which needs its own instrument and its own specification, described in surface finish explained.

Build the Inspection Plan With the Design

The cheapest metrology decisions are made on the drawing: choosing datums a fixture can actually reference, tolerancing only what matters, and specifying a measurement method next to the characteristic. Projects House writes inspection plans and dimensional acceptance criteria for US clients and validates them with suppliers through a global manufacturing network. Send your drawings and current inspection data through our contact form.