A first-article inspection (FAI) is the cheapest insurance in overseas manufacturing: one part, measured completely, before the batch exists. Run it properly and most production disasters — wrong material, missed operations, drifting tolerances — get caught while they still cost one part instead of ten thousand. Here's what a real FAI looks like, how to review one, and where suppliers cut corners.
What a first article actually is
One part — not a prototype, not a showroom sample — made on the actual production setup: production fixtures, production programs, production material, and (for castings) the production tool. That last point is where cheap suppliers cheat. A first article machined from a hand-fed blank proves the CNC can hit the numbers; it proves nothing about the fixture your batch will run in. If the FAI didn't come off the setup that will make the order, it isn't a first article.
The deliverable is a pack, not a part: a ballooned drawing, a dimensional report, material certificates, and finish or treatment reports. The part itself matters, but the paper is what you're approving.
What a real FAI pack contains
- A ballooned drawing. Every dimension on your drawing gets a sequential number (a "balloon") — no exceptions, no "the rest per drawing". If a feature isn't ballooned, it wasn't checked.
- A dimensional report. For every balloon: nominal, tolerance, actual measured value, the instrument used, and pass/fail. Handwritten values on a spreadsheet are fine; a column that just says "OK" is not a measurement.
- Material certificates. Mill cert for the exact heat lot in your part, matching the grade you specified — 6061-T6 means 6061-T6, not "the aluminum we had". For critical parts, an independent spectrocheck is worth the small fee.
- Finish and treatment reports. Anodize thickness, coating thickness, salt-spray hours where specified — measured, with the number written down.
- Photos of the actual part from several angles, including the features you can't check remotely: seal faces, weld penetration locations, marking.
How to review the report — line by line
Sit down with the drawing and the report side by side, and check three things for each balloon:
- Is the value real? Actual values clustered suspiciously close to nominal ("20.00, 20.00, 19.99…") suggest the inspector filled the sheet from the drawing, not from the part. Real measurements scatter inside the tolerance band. This single pattern catches more fake reports than anything else.
- Is the tolerance the one you specified? Watch for quiet rewrites: your ±0.02 mm becomes "±0.1 mm per general tolerance". If the report's tolerance column doesn't match your drawing, the report is describing a different part.
- Is the measurement method credible? A ±0.01 mm bore measured with calipers (resolution 0.01 mm, accuracy worse) is a guess. Tight features need a CMM or bore gauge, and the report should say which instrument was used.
Then check completeness: count the balloons against your drawing's dimensions. Anything missing — especially a datum, a thread spec or a finish callout — gets re-reported before you sign.
The failures a first article catches
Done honestly, the FAI catches the expensive stuff while it's still cheap:
- Wrong or substituted material. The mill cert doesn't match the order, or the grade was "equivalent".
- Missed operations. Deburring skipped, threads not tapped, the anodize mask forgotten — all invisible in a photo, obvious in a report.
- Systematic tolerance errors. A fixture offset or a wrong program makes every part of the batch wrong the same way. One measured part finds it; a container of parts proves it.
- Finish problems. Coating thickness under spec on the "invisible" faces, or anodize color off between batches.
The tricks — and the right response
- "Report to follow." Parts photographed, report promised later, signature requested by message. Response: no signature without the full pack. Approval of paper you haven't seen isn't approval.
- The cherry-picked sample. Five first articles made, best one sent. Ask which setup the others ran on — and keep the right to a random pull from the first production batch to re-verify.
- Deviation diplomacy. "Minor deviation, please approve to keep schedule." A deviation can be legitimate — but it's your call, priced as your call: the schedule pressure is not an engineering argument. Approve in writing with the deviation documented, or reject; never let silence be the answer.
- The generic cert. A material certificate with no heat-lot number and no reference to your order is wallpaper. It proves a certificate exists, nothing more.
FAI differs by process
The report discipline is the same; the risk concentration moves with the process:
- Die castings: the tool dominates. Check porosity evidence on critical sections (X-ray or sectioned samples), verify parting-line and ejector-pin locations against the approved tool design, and confirm the alloy matches the spec — ADC12 vs A380 differences hide inside a casting that looks fine. Weight is a surprisingly good tripwire: a casting 4% under nominal weight usually means voids or short fill.
- CNC machined parts: the setup dominates. Confirm the first article came from the production fixture and program — re-run one dimension yourself on a second article if you can. Watch inside-corner radii (did they use the tool the drawing assumed?) and surface finish on the faces that matter.
- Sheet metal: flat-pattern correctness, bend angles, and hole-to-bend distances. Also the finish: powder thickness on every face, including the ones pointing down in the rack.
- Extrusions: cross-section dimensions along the length (twist and bow grow with length), temper verification, and aging records for the batch.
A five-minute FAI triage
Before the full review, run this triage — most packs reveal their quality immediately:
- Count the balloons. Fewer balloons than drawing dimensions? Something isn't being checked. Reject the pack, don't cherry-pick.
- Scan the "actual" column. Too many values identical to nominal means transcribed, not measured.
- Match tolerances to your drawing. Any substitution of "general tolerance" where you specified a number is a rewrite.
- Find the heat-lot number on the material cert. No lot number, no traceability, no approval.
- Check who measured. CMM named for tight features, hand tools only for the loose ones. The reverse is a red flag.
Packs that survive this triage deserve the line-by-line review. Packs that don't have told you enough already.
After you sign off
Approval releases production — and sets the baseline. Three rules keep it meaningful:
- The first articles become retained samples. Both sides keep signed parts; any later dispute gets measured against them.
- Changes trigger a re-FAI. New mold, new material source, new fixture, major repair — the article gets re-approved before the batch resumes. "It's basically the same" is not a process change policy.
- FAI ≠ batch inspection. A perfect first article doesn't excuse pre-shipment AQL sampling on the actual order. The two checks answer different questions: can you make it right, and did you make it right this time.
Where I fit in
I review first articles for clients as a standard gate: checking the report against the drawing feature by feature, spotting the rewritten tolerances and the too-perfect columns, and physically re-measuring critical dimensions here in China before anything ships. When your signature is needed, it's an informed one — and when something's wrong, you find out from me before production, not from your customer after delivery.
See where the FAI gate sits in the full sourcing process →
Start earlier: the DFM checklist that keeps first articles right →
Send a drawing — first-article review included →
Common questions
- Do first articles cost money? Often yes for custom parts — they carry setup time and may destroy material. But the FAI report is part of doing the job, not an add-on. A factory that bills separately for the paperwork is telling you how it thinks about quality.
- How long does an FAI take? The measurement itself is days, not weeks. What takes time is everything upstream — tooling, setup, programming. Plan the FAI into the schedule instead of letting it become the schedule.
- Is FAI the same as ISO 9001's requirements? ISO 9001 requires controlled production but doesn't mandate a formal FAI; AS9102 (aerospace) does. For most industrial parts, a pragmatic ballooned-drawing FAI like the one above gives you the same protection without the full standard's overhead.