When purchasing changes suppliers, the process is nearly always the same: collect quotations, read the website, tour the showroom, flip through customer cases, and finally look at a "sample board". That process filters out the obvious no-hopers, but it is almost useless for the differences that actually decide delivery quality — the sample board was selected, the showroom was arranged, the cases were prepared.
What reveals the real level is the corners of the plant that visitors are normally not taken to. None of the seven details below needs specialist equipment or process expertise; half a day is enough.
One rule of thumb is remarkably reliable: if someone in the workshop is wearing a synthetic jacket and picks up a board bare-handed, you can stop the assessment there. ESD damage does not show up in appearance; it shows up as customer-side failures three months later.
Paste must be stored cold, warmed for long enough before opening, and used within its open time. Every plant does these three things; the difference is whether they are recorded, and whether the records are continuous.
In a high-mix, low-volume plant, changeover happens every day and is also the moment of greatest quality variation. Ask directly for today's changeover record: changeover duration, first-article confirmation time, who confirmed, and where the first board after changeover went. A plant that says "we change over very fast" but cannot produce the record usually means there is no standard — just a senior operator's feel.
A human visual check with a signature and a machine comparing every placement on every panel are two different orders of magnitude. The first catches obvious missing or wrong parts; the second catches offset, polarity and wrong material — the items the eye handles unreliably — and leaves comparable data behind.
The more revealing question is about process interlock: if the first article has not been confirmed, can the line start volume production by itself? If the answer is "yes, but we would not", the control depends on discipline rather than on the system.
Many plants now have AOI; few can state their false-call rate. Three things to look at: who tunes the program, roughly what the false-call rate is, and who performs the review. Whether false calls that were passed are recorded matters more than the rate itself — every escape begins with "let it go this time".
Ask for AOI reports from three sampled batches and check whether first-pass yield and review records are continuous. Where the record breaks is where the problem is.
Traceability has three levels:
| Level | What it can answer | Typical fit |
|---|---|---|
| Batch | Which material lots went into this batch of boards | Consumer, general industrial |
| Panel | Where this panel sat and which reflow run it went through | Medium reliability requirements |
| Unit | For this board: reel lot, machine, program revision, operator, time | Medical, automotive, high reliability |
There is a very practical field test: pull an arbitrary board from a finished carton and ask the plant to trace its material lots and placement program revision on the spot. If it comes back in a few minutes, the traceability is real; if the answer is "we will put it together and send it to you", it was usually assembled after the fact.
Look at two places: the quarantine area and the rework bench. The quarantine area needs clear marking, a log, and ideally a lock. Every board on the rework bench should be traceable to a work order, and every reworked unit must be re-inspected with the result recorded.
One question worth asking face to face: what happens to scrapped boards? Harvesting components from scrap is an open secret in the industry. Asking may be uncomfortable; not asking is the actual risk.
Everyone has prepared cases. The real material is last week's production report, the defect-rate trend over the last three months, and the corrective actions from the last three customer complaints. A plant willing to lay those open is usually also unafraid of an unannounced line audit — and that says more than any certificate.
An audit is not about finding fault; it is about finding a partner who can carry the risk for you over the long term. The closer you look up front, the less you worry later.
