An audit exists to reduce uncertainty about how a supplier actually operates, not to produce a certificate. The certificate is a by-product. If the process produces a clean report and no change in understanding, the audit failed regardless of the score.
Most supplier audits are conducted against a checklist that was written for a different industry, or for a much larger plant. The checklist asks whether a documented procedure exists, and the supplier produces one. Nothing is learned, because a document proves intent, not practice.
The useful question is always the same: show me this happening, now, on work that is in progress. Everything else is theatre.
Checklists go soft in three predictable places.
The first is document control. A procedure with a recent revision date proves that someone edited a file. It does not prove that the version on the floor matches, or that the operator has read it. Ask to see the copy at the machine and compare revisions.
The second is calibration. A calibration sticker proves that an instrument was sent out. It does not prove it was used correctly, or that the measurement plan matches the drawing. Ask to see a measurement being taken on a part that is currently in production.
The third is traceability. A certificate pack proves that material certificates exist. It does not prove that the certificate belongs to the material in the rack. Ask to follow one heat number from the rack to the finished part, and time how long it takes.
Good audit questions are specific, answerable, and hard to prepare for in advance.
Ask what happens when a first article fails. The answer reveals whether there is a quarantine process or whether the parts are reworked quietly. Ask who can stop production, and whether that person has ever used the authority. Ask how a rework is recorded, and whether the record follows the part.
Ask about the last customer complaint, what caused it, and what changed as a result. A supplier with no complaints is either very good or not recording them. The follow-up question "what changed" is the one that separates the two.
Ask to see the schedule for the current week and the previous week, and compare. A schedule that was rewritten entirely is a schedule that is not being followed.
Much of what matters is visible without asking anything. Look at how work is stacked. Work in progress piled beside a machine usually means a bottleneck upstream or a batch released too early. Empty floor beside a machine that is running means the opposite.
Look at the measurement tools. If the only instruments at the bench are the ones that are easy to use, the harder measurements are being deferred to inspection, which adds a queue. Look at the drawing at the machine: is it the current revision, and is it legible, or is it a photocopy of a photocopy.
Look at how scrap is handled. Scrap in a labelled bin with a tag is a controlled process. Scrap on the floor is a process that has given up. Neither is decisive alone, but the pattern across a walk is more informative than any single observation.
Standard audits concentrate on the steady state: calibration, document control, training records, non-conformance handling. They are much weaker on the two things that decide whether a new programme succeeds. Both are central to custom work, where every part is a new introduction.
The first is ramp-up. An audit rarely asks how a new part is introduced, who proves the process, how many pieces are run before full release, and what happens to the pieces made during learning. These are the questions that predict whether the first delivery arrives on time.
The second is tooling. Ownership, storage, condition, and revision are often outside the checklist entirely, even though a failing tool is one of the most common causes of a missed date. An audit that ignores tooling is auditing the easy half of the operation.
If an audit has one day, these five items use it well.
An audit that ends with a score has wasted its findings. The output should be a short list of specific risks, each with a question the supplier has agreed to answer by a date. That list becomes the basis for the next conversation, and eventually for the next audit.
It also gives the buyer something to monitor. If the main risk was tooling condition, the follow-up is a register review in three months. If it was ramp-up discipline, the follow-up is the first article record on the next new part. Monitoring a named risk is far more effective than re-scoring a checklist.
An audit is a snapshot, and snapshots go stale. The useful pattern is a light repeat at a defined interval, focused on the risks identified the first time, plus one new area. That keeps the relationship current without turning into an annual performance.
For buyers of custom fabricated work, the areas worth auditing hardest are the ones where the supplier has the most freedom: process routing, tooling, and the introduction of new parts. Those are also the areas where customisation changes the answer. A factory founded in 2009 with 53,819 sq ft and more than 50 employees will have a system worth looking at directly, and the walk will tell you more than the file. The company background, the parts under other parts, stainless exhaust systems and titanium components, and the dealership terms are useful preparation.