The Inspection Report Measures the Drawing, Not the Part

Most first-article inspection reports we review are technically correct and practically useless.

Sixty dimensions, all in tolerance, measured against the datums printed on the drawing. The part goes into the assembly and the housing will not close, or a connector lands a third of a millimeter off the board. Nothing is out of tolerance. What follows is two weeks of argument that ends with everyone agreeing the drawing was ambiguous, which is true and does not help.

The datums are the problem, not the dimensions

A datum scheme is usually chosen so the part can be fixtured and inspected. Assembly does something else. The drawing says measure from face A and edges B and C. The assembly locates on two clip features, a boss face and the screw column. Measure the same part from both sets of features and you get two different numbers, and both are correct.

The fix is cheap. Mark the dimensions that control fit as critical to function, and require the report to take those from the same features the assembly touches. Everything else can stay on the standard scheme. An hour on the drawing at the RFQ stage removes the most common cause of a rejected first article that was never actually out of tolerance.

Shrinkage is not finished when the part leaves the press

Semi-crystalline materials keep moving after ejection: POM, PA, PP, and filled grades of all three. Most of the movement happens in the first day at room temperature, with a smaller drift continuing for two to three days in thicker sections. Premise: unfilled or lightly filled grades, ambient storage, no annealing.

A report measured twenty minutes off the press therefore describes a part that will be a different size tomorrow. Ask for a stated rest period before measurement, 24 hours as a default and 48 where tolerances are tight, and ask for that condition printed on the report. If the part is annealed, measure after the anneal, because the anneal changes dimensions and that change is the result you actually care about.

Room temperature belongs in the same sentence. A part measured on a shop floor in July and one measured in a cold inspection room in January will not give the same number, and a report that omits the condition cannot be compared with the next one — including the one you will need in six months.

Match the instrument to the feature

Reports go wrong quietly when the instrument cannot reach the feature.

• Hole position and edge profile: CMM works, provided the stylus fits. A 2 mm ruby cannot enter a 1.2 mm slot; that feature needs vision or a contoured stylus, and a report that skips it without saying so is worse than one that flags it.

• Thin-wall sections: section a part and measure under a profile projector, or use ultrasonic. Calipers across a curved wall measure the chord over the outside, not the wall.

• Thread depth and fit: thread gauges. Calipers read crest to crest and flatter the result.

• Flatness on a large thin part: support it the way the assembly supports it. Clamp it flat in the fixture and the CMM is measuring the fixture.

• Texture and gloss: compare against a physical plaque at the same angle and light source, not a photograph.

Uncertainty against tolerance

If the tolerance is ±0.05 mm and the measurement repeats to 0.02, the report is largely describing the measurement system. The working rule most shops use is that the measurement system should consume no more than a fifth to a tenth of the tolerance band. When it does not, the answer is a decision rather than a tighter process: relax the tolerance if the feature is not functional, or invest in better measurement if it is. Arguing over a 0.03 mm difference the gauge cannot resolve is how a sound tool acquires a reputation it does not deserve.

Sampling that answers a question

A report should state which cavities, which shots, and when. Five parts from one cavity in the first hour says nothing about a sixteen-cavity tool eight hours into a run.

Minimum useful for a first article: every cavity, three consecutive shots each, at the start and near the end of the run, measured after the rest period. Capability needs more than that. Twenty-five to thirty parts per cavity is the point at which a Cpk figure starts to mean something; below it, report the spread and call it a spread.

What to send before T1

Before you approve the first article, send the drawing back once with two additions: the critical-to-function dimensions marked, and the datum features named — the ones the assembly actually locates on. Then ask that the report state rest time, room temperature, instrument and operator for each block of features.

Two pages of instruction, one revision cycle saved, and a first article that answers the question you asked.

An inspection report is an answer. If nobody wrote the question down, the answer will be "in tolerance," and you will still be having the argument.