First article inspection: proving the process as configured can make the drawing

Every requirement, not the ones that are easy to reach

The distinguishing feature of this exercise is completeness. Each dimension, note, finish call-out, material specification and marking requirement on the drawing gets an entry, a result and a reference to how it was verified. That total sweep is precisely why it is worth doing: the characteristics that get forgotten in routine checking are the ones nobody thought about, and this is the one occasion the drawing is read line by line. Numbering the requirements first, so results can be matched back to them, turns an argument about what was covered into a document anybody can audit.

It proves the configuration, not the population

A conforming first article establishes that the tooling, programme, fixture, material and settings in combination can produce an acceptable part. It says nothing about spread, drift or the behaviour of the process over a run, which is a separate question answered by capability evidence gathered over time. Confusing the two leads to a familiar disappointment: an approved part, followed by a production run that misses the same feature repeatedly, because the first article happened to be a good one. Where a characteristic is known to be difficult, the approval package should carry data from a series of parts alongside the single article.

The events that should trigger a repeat

Approval attaches to a configuration, so anything that changes the configuration invalidates it. Replacement or major refurbishment of tooling, a move to a different machine or a different site, a change of material supplier or grade, a design revision, a change of subcontracted operation, and a long dormancy where the setup has been broken down and reassembled from memory. That last one catches plants out most often, because nothing was formally changed but nothing survived either. Agreeing the trigger list with the customer at the outset avoids the harder conversation about whether a change should have been notified.

Who signs, and what the signature is worth

Two signatures matter. Yours states that the results are true and the part was made by the process described. The customer's states that the evidence has been reviewed and production may proceed. Where the customer waives review, get the waiver in writing, because a waived approval has a habit of being remembered differently after a failure. Internally, the person signing should not be the person who made the part, and should have physically seen the article rather than approving a spreadsheet. A package assembled by an engineer who never handled the part reliably contains transcription errors.

What the package is really for after approval

The value does not end at the sign-off. The approved article becomes the reference point for every later dispute about what was agreed: which revision the process was proven against, how a form was measured, what the datum scheme was, which interpretation of an ambiguous note both parties accepted. Retain the article itself where practical, along with the results and the drawing revision used. Years later, when a customer asserts that a feature has always been out, the ability to produce the approved part and its measurements settles the question in an afternoon.

Frequently asked questions

Does a passed first article mean production parts will conform?

No, and the gap between those two statements causes most of the disappointment around the exercise. It demonstrates the configuration is capable of producing an acceptable part under known conditions. Variation over a run, tool wear, material lot differences and operator effects are outside its scope entirely. For characteristics that are tight relative to what the process can hold, capability evidence gathered across a series of parts is the appropriate companion, and it should be planned at the same time.

When does a first article have to be repeated?

Whenever the configuration it proved no longer describes reality. New or substantially refurbished tooling, transfer to another machine or plant, a design change, a change in material source, a newly subcontracted operation, and any restart after a long production gap all qualify. The commonly missed trigger is the last of these, because nothing was deliberately altered while everything was quietly reassembled. Agree the list with the customer early rather than debating notification obligations after a problem appears.

Who should carry out the measurements for a first article?

Somebody independent of the person who made the part, using gauges whose calibration status is recorded on the report. Where the measurement is complex, an accredited laboratory adds weight and removes the argument about method, but it is not necessary for straightforward features. What matters more is that the report states how each characteristic was measured, since a disagreement between customer and supplier on a form feature is nearly always a difference in fixturing rather than in the part.