Resources
What to check on a first article inspection report
Published 2026-08-13
A first article inspection report is the last moment before volume when either side can still change its mind. Signing it locks in the arrangement that produced that one part — machine, fixturing, toolpath, gauges — and the next few thousand get made the same way. Here is the order to read one in, what to look at, and when to send it back.
A first article proves the process repeats, not that one part passed
Plenty of buyers treat the FAI report as a pass certificate: every number falls inside tolerance, sign and release. That reads it too small. The question a first article actually answers is whether the arrangement that made this part can be repeated across the next few thousand. Passing is the necessary condition; repeatable is the point.
So beyond the numbers, the report has to show where the numbers came from: which datums the measurement was set up on, what gauges were used, who measured, and on what date. A report with values but no provenance quietly moves the process risk onto the buyer.
Put differently: run the same drawing again with a different shift on a different machine, and the values should broadly reproduce. Anything that can't clear that bar isn't a first article — it's a sample. A sample proves the part can be made; a first article proves it can keep being made.
Dimensions: datums first, then the criticals, general tolerances last
Don't read the report top to bottom. Start with the datums: did the inspection set up its coordinate system on the datum features the drawing specifies? If the datums don't match, every hole position below was measured in a different frame — it can look beautiful on paper and still land off in assembly. If this line fails, the next forty rows aren't worth reading yet.
Second, the handful of dimensions you individually toleranced on the drawing. That is the precision you paid for, so check every one of them; sampling them defeats the purpose. A value sitting hard against the limit isn't a failure, but it earns a question: is the process centred off to one side, or was this part unlucky? Dimensions whose process centre sits near a limit are always the first to walk out of tolerance in volume.
Third, and only third, the bulk riding on the general tolerance class (ISO 2768 m or f is the usual choice for machined features). Spot-check these, but look for a whole group leaning the same way — that usually means a systematic offset in tool compensation or work-offset setup rather than random spread.
One more family doesn't belong to the CNC at all: the wall thicknesses, groove widths and outer envelope that come from the extrusion die. Those are judged against the profile standard (precision profiles typically to EN 12020, on the order of ±0.1 mm). Holding the section to machining tolerances is the most common misjudgement in a first article review — it rejects conforming parts, and it distracts from the features that genuinely need attention.
Material and temper: the full call-out, traceable to a batch
'Aluminium' in the material field isn't enough, and neither is '6061' on its own. Alloy and temper travel together: 6061-T6, 6063-T5. Temper sets mechanical properties and cutting behaviour, so leaving it off leaves half the specification unstated. The working set here is 6060, 6061, 6063, 6005A and 6082 in T5 or T6 — whichever the drawing names is what the report should show, and 'equivalent substitute' is not an answer.
The mill certificate has to tie to the metal in front of you: same alloy and temper as the report, and traceable to the specific batch the first article was cut from. A certificate that can't be traced to a batch is barely better than no certificate.
If the part will later be welded, anodized or carry load, this field stops being paperwork. Within the 6000 series, alloys differ in how consistently they take anodizing colour and how they behave when welded. This is a line worth pushing back on rather than waving through to save a round of email.
Cosmetics and finish: burrs, clamping marks, masking boundaries
Appearance is the field most likely to be closed out with 'looks good'. Ask for something specific instead: how far deburring goes, what kind of witness marks are acceptable where the part was clamped, which faces are cosmetic and which are not. If the drawing never said, the first article is the last chance to write it down — put it in the approval and the next few thousand parts have something to be judged against.
On finished parts, the report should carry critical dimensions before and after treatment, or at minimum state which condition the listed values belong to. Coating consumes clearance, so close-fitting bores and shafts cannot be left ambiguous here.
Masking is the other high-rework field: grounding faces that must stay conductive, threads that must still accept a go gauge. The first article should show that they were in fact masked and that the boundary falls where you can live with it. For boundaries, one photograph beats ten sentences of description — if the report has no photos, ask for one.
Threads and mating features: if you can trial fit, don't settle for numbers
'M6 × 12 deep, pass' is not enough on its own. Ask how it was verified — go/no-go gauges are the normal practice, and recording the gauge designation and result says more than a single measured value. Cut tapping, form tapping and threaded inserts are verified differently, so whatever the drawing calls for is what the first article should be checked against, not whichever method was quicker on the day.
Whenever the mating part is available, the highest-value line in the whole report is a trial fit: actually assemble the first article against its counterpart, or run it through a buyer-supplied check fixture. The classic failure — every part conforming, the assembly refusing to go together — is caught here or not at all.
If the trial fit refuses, don't jump straight to a rejection. Separate a single out-of-tolerance feature from an accumulated tolerance stack: the first is a process fix, the second means going back to the drawing. Those two conclusions mean very different things for the next few thousand parts, and confusing them costs another round.
Walk this list before you sign
This is the list we walk ourselves before a first article leaves the plant. When the buyer reads the report in the same order, the places where the two sides disagree usually surface in one pass rather than three.
| Check | What skipping it costs |
|---|---|
| Measurement set up on the drawing's datums | A different frame makes good hole numbers meaningless |
| Every individually toleranced critical checked, not sampled | This is the precision you paid for — sampling it wastes it |
| Noted whether values sit hard against a limit | An off-centre process is the first to drift in volume |
| Section dimensions judged against the profile standard | The common misjudgement — it rejects conforming parts |
| Alloy with temper, certificate traceable to a batch | No temper is half a spec; no batch is barely a certificate |
| Stated whether values are before or after coating | Coating clearance decides whether close fits assemble |
| Thread verification method and masking boundaries recorded | Threads and masking drive the most rework |
| Trial fit performed wherever the mating part exists | All-parts-conform, assembly-won't-go shows up here |
Once signed, the report becomes the reference for every batch after it
An approved first article stops being an inspection record and becomes the benchmark every later shipment is judged against — when a dispute comes up, both sides come back to it. That is why three extra questions before signing beat discovering, when the third shipment lands, that some requirement was never defined in the first place.
If a line is unclear, don't tick it through on your own. Our practice is to send the measurement record, the gauges used, and photos of masking and trial fit together, then walk through whatever doesn't read clearly. Whether the answer is a process change or a drawing change, it gets settled before volume starts.
Keep reading