What a Nadcap heat-treat audit actually walks through
The uncomfortable thing about a Nadcap heat-treat audit is that it is not really an inspection of your process. Your process is probably fine. It is an inspection of whether you can prove your process was fine on a specific day you do not get to choose.
That distinction decides how the week goes. A shop with good practice and poor records has a hard audit. A shop with good practice and records that reconstruct themselves has a boring one, and boring is the goal.
What follows is the shape of the walk, and where it tends to go wrong. It is not a substitute for the checklist — AC7102 and its slash sheets are the authority, and your auditor has read them more recently than anyone.
It starts with a job, not with your system
The auditor picks a job. Often one that shipped some months ago, sometimes one they noticed in your order book, occasionally one a customer flagged. From there the direction of travel is backwards, and every step asks the same question in a different costume: show me the record, and show me it was valid at the time.
The certificate. What did you tell the customer? Which specification, which revision, what quantity, which lot and heat. The certificate is the claim; everything after this is checking the claim was true.
The load. Which furnace charge did those parts run in? Not which furnace — which charge, on which date, with what else in it. If parts from one order line went through more than one charge, all of them.
The cycle. What recipe did that charge run, at what setpoints and tolerances, and what actually happened? Not the recipe as it exists today — the recipe as it was when the load ran. Recipes get revised. An auditor who finds you comparing a 2024 load against a 2026 recipe has found something.
The furnace, on that date. Now the pyrometry. Was the system accuracy test current? Was the uniformity survey current, and was the cycle temperature inside the range that survey qualified? Were the controller, recorder and over-temperature instruments in calibration, and passing?
The thermocouples. Which ones, from which lot, within calibration, within their permitted number of uses.
The people. Was whoever ran it qualified to, on that date.
Every step is answerable. The difficulty is that answering the fourth step requires knowing what was true in the past, and most systems only know what is true now.
Where it goes wrong
Not at the certificate. Nearly every shop can produce the certificate.
It goes wrong at the point where the auditor asks for the SAT that was in force when the load ran, and the system offers the most recent one. They are not the same record, and the difference is the entire question.
What usually happens next is a folder. Someone goes to a shared drive, opens a spreadsheet of SAT dates, and cross-references by hand against a load date written on a traveler. It works — plenty of shops pass this way. But three things are now true:
- The reconstruction is manual, so it is only as good as the person doing it under pressure.
- The spreadsheet is now in scope. You have just shown the auditor that your compliance position lives outside your controlled system.
- It takes hours per job, and they may ask about several.
The second one is the expensive one, and it is rarely the finding people expect.
The questions worth rehearsing
Not because an auditor will phrase them this way, but because if you can answer them cold, the walk above holds together:
- For a load that ran a year ago, which SAT, TUS and calibrations were in force on that date?
- Was the cycle temperature inside the range the then-current survey qualified?
- Which thermocouple lot was installed, and was it inside both its date and its use count?
- If a calibration was recorded last week, does last year’s load still show the calibration that applied then — or has the record silently updated?
That last one is the quiet killer. A system that overwrites in place will answer every question confidently and some of them wrongly, which is worse than a system that admits it does not know.
What “good” looks like
A shop that finds audits boring usually has three properties, and none of them is a dashboard:
Records are dated and never edited in place. A correction is a new record, not a change to an old one. History is a sequence, not a current value.
Evidence is captured at the moment of use. When the charge is built, what was true about that furnace is written onto the charge and frozen. Later changes cannot rewrite it, because it is no longer a reference — it is a copy.
The reconstruction is a query, not a project. Ask for a job, get the whole chain back: receiving, the charges, the cycle actuals, the pyrometry as of the run date, the tests, the certificate.
We built the audit-evidence pack to do exactly that reconstruction — for a certificate, reproduce the SAT, TUS and calibrations in force when each load ran, re-derive whether they were valid on that date, and turn gaps into findings rather than silence. The pack is only complete when every record checks out, which means “complete” is a claim the system will stand behind rather than a label someone applied.
The goal is not to impress an auditor. It is to make the answer to every question in the walk above something you can produce in front of them, in a minute, without opening a spreadsheet.