How to Audit a Foundry Supplier: A 25-Point Checklist for On-Site Verification

A foundry audit fails in two directions. It fails when the buyer walks the floor, nods at the machines, accepts a brochure, and signs. It also fails when the buyer arrives with a hundred-page questionnaire that nobody can answer in a day, so the visit becomes a paperwork session and the floor never gets looked at properly.
What follows is the middle path: a visit you can actually complete, with a scoring sheet that turns observation into a decision. It assumes you are technical enough to know what a spectrometer is, and that you have limited time and one shot at the supplier.

What This Audit Decides
A supplier audit answers one question: can this foundry make my part, repeatedly, and will it tell me the truth when it cannot?
That is it. It is not a price negotiation, not a capability wish-list, and not a relationship-building exercise. Every check below maps back to one of those two halves, and if a check does not, it does not belong on the sheet.
The output is a score out of 50. Not a feeling, not a “they seemed professional.” A number, with the evidence written next to it, so that when someone asks you in three months why you chose this supplier you can answer without reconstructing your memory of the visit.
Three Different Things People Call an Audit
These get conflated constantly, and conflating them wastes time.
Screening is desk research. Certificates, references, a first quote. It costs an afternoon and it eliminates maybe half the field. It is not an audit.
An audit is a visit. You verify that the documents describe a real operation and that the operation can do your work. The output is a decision to proceed or stop.
First article approval is a part. You approve a physical casting against the drawing. It verifies the process for one part; the audit verifies the process in general.
You need all three, in that order. Suppliers who offer to skip straight to a part are usually the ones who would rather you never saw the floor.
Before the Visit: The Documents That Tell You Most
Request these a week ahead. How a supplier responds to this list is itself data: a prepared foundry sends most of it in a day, a disorganized one sends excuses.
| # | Document | What you are actually checking |
|---|---|---|
| 1 | ISO 9001 certificate | Validity, and that the scope covers casting |
| 2 | Quality manual or control plan | Whether one exists for work like yours |
| 3 | Process flow diagram | Whether they think in process steps or in jobs |
| 4 | Inspection equipment list and calibration certificates | That measuring devices are controlled |
| 5 | Sand and binder specifications | That materials are specified, not bought ad hoc |
| 6 | Alloy specifications and suppliers | Traceability of charge material |
| 7 | Pattern and tooling inventory for your part type | Whether the work is in their normal range |
| 8 | Reject rate for the last 6 months | Whether they measure it at all |
| 9 | Customer complaint log, anonymized | Whether problems are recorded and closed |
| 10 | A corrective action report (CAR) example | The single most revealing document on this list |
| 11 | Equipment list with ages and capacities | Whether the capability claim matches the plant |
| 12 | Heat treatment procedure and calibration records | If they heat treat, whether it is controlled |
| 13 | Inspection and test plan template | What they propose to check, and how often |
| 14 | Sample material certificate | That traceability actually works end to end |
| 15 | Insurance and certification for special processes (NDT, welding) | That subcontracted skills are certified |
If you only read one document, read item 10. A corrective action report shows you how the foundry behaves when something has already gone wrong: whether the root cause is real or performative, whether the fix was verified, whether the customer was told early. Any foundry can look good when nothing is broken. The CAR is where you find out what happens when things are.
A warning about reject rates. A foundry that says “we do not track that” is telling you something. A foundry that quotes a suspiciously low number, say under 2% for a complex part, should be asked how it is measured. If the answer is “we only count what the customer rejects,” the number is not a reject rate.
On the Floor: What to Watch, Step by Step
Walk the process in the order metal moves through it. Do not let the tour start at the office and end at the shipping door.
Melting. Is the spectrometer switched on and in use? Are charge materials segregated by alloy, or is there a pile of mixed scrap? Ask to see the last week of furnace logs. Ask who records them and when, because logs completed at the end of a shift from memory are decoration.
Molding and core making. Watch a mold being made. Is the sand being tested at the station, or does the sample go to a lab and come back tomorrow? Ask what the compactability target is and what happens when it drifts. If the answer is a number, good. If the answer is a shrug, the process is running on habit.
Il pleut à verse. Ask for the pouring temperature range for your alloy and where it is recorded. Watch a pour if you can. A foundry that pours without recording temperature is not controlling one of the two variables that decides whether your part fills.
Heat treatment. Look at the chart recorder. Is the paper or the data file current, and does the record match the parts in the furnace? Ask who reviews the charts. If nobody reviews them, they are being generated, not used.
Cleaning and finishing. This is where cost hides. Watch how gates are removed and how grinding is decided. If the rule is “grind until it looks right,” your part’s finish depends on who is holding the grinder that day.
Inspection. Ask to see a first article inspection actually being performed. Check that the instrument used is on the calibration list. Ask what happens to a part that fails: is it tagged, quarantined, and does it appear in the reject log you were shown before the visit?
One habit worth adopting. Ask the same question of two different people. Ask the molders what the sand target is, then ask a supervisor. If the two answers differ, the control is in the paperwork and not on the floor.

The 25-Point Scorecard
Score each item 0, 1 or 2.
- 0 means absent, or claimed but not evidenced.
- 1 means present but informal, inconsistent, or person-dependent.
- 2 means documented, evidence available, and someone owns it.
Maximum is 50. Write the evidence next to each score as you go, because the evidence is what makes the score defensible later.
Quality system
| # | Check | 0 / 1 / 2 |
|---|---|---|
| 1 | ISO 9001 certificate valid, and the scope covers casting | |
| 2 | A control plan or equivalent exists for work similar to yours | |
| 3 | Inspection equipment calibrated, certificates current and available | |
| 4 | Nonconformance records kept, with closure evidence | |
| 5 | Reject rate measured and reported, with a defined method |
Process capability
| # | Check | 0 / 1 / 2 |
|---|---|---|
| 6 | Spectrometer in use, calibrated, with recent records | |
| 7 | Charge materials segregated by alloy and traceable | |
| 8 | Sand tested at the molding station, per shift, with records | |
| 9 | Sand parameter limits defined, with action taken when exceeded | |
| 10 | Core curing time and temperature controlled and logged | |
| 11 | Pouring temperature recorded for every heat | |
| 12 | Heat treatment charted and reviewed by a named person |
Traceability
| # | Check | 0 / 1 / 2 |
|---|---|---|
| 13 | Heat number traceable from material certificate to shipped casting | |
| 14 | Pattern or tooling identity linked to production lots | |
| 15 | Test bars cast from the same heat, not separately | |
| 16 | Shipping documents tie back to the production record |
Capacity and honesty
| # | Check | 0 / 1 / 2 |
|---|---|---|
| 17 | Quoted lead time consistent with the order book you can see | |
| 18 | Equipment list matches what is actually installed and running | |
| 19 | Any subcontracting declared up front, not discovered later | |
| 20 | Staffing consistent with the claimed output |
Commercial and behavioral
| # | Check | 0 / 1 / 2 |
|---|---|---|
| 21 | Tooling ownership and transfer terms stated in writing | |
| 22 | Design change and revision process defined | |
| 23 | Willingness to run a paid pilot with a first article report | |
| 24 | Problems volunteered before you find them | |
| 25 | The uncomfortable question gets a direct answer |
How to read the total:
| Score | Reading |
|---|---|
| 40 to 50 | Proceed. Run a pilot on your hardest part, not your easiest. |
| 30 to 39 | Conditional. Identify the specific gaps and require evidence of closure before the first production order. |
| Below 30 | Stop. The gap is systemic, and it will show up in your parts. |
Do not average away a zero. Items 1, 3, 6, 13 and 25 are structural. A foundry with no calibrated instruments, no spectro chemical control, or no traceability is not a partial supplier, whatever the other 20 checks say. Weight those five as pass or fail and score the rest.
Documents Describe Intent, the Floor Describes Reality
The most useful thing you can do on a visit is look for the gap between the two.
Some specific gaps worth hunting:
Records that are too tidy. A month of sand logs completed in the same pen, in the same handwriting, with no corrections and no gaps, was written once afterwards. Real records have crossings-out and initials.
Equipment that is furniture. A spectrometer under a dust cover, a chart recorder with a flat line, a calibration sticker that expired last year. None of these are proof of fraud; all of them are proof that the control is not live.
A lab with nobody in it. Ask who runs the inspection and when they work. If the answer is a person who is also doing three other jobs, you have learned the real inspection frequency.
Capacity arithmetic that does not close. Multiply the machines by plausible cycle times and compare with the claimed monthly output. If the numbers require 30-hour days, the lead time you were quoted is not real.
Stock that does not match the story. If they claim to pour 200 tonnes a month, the raw material on the floor should be consistent with that. It usually is not, in both directions.
None of this requires accusing anyone of anything. You are checking whether the management system is alive or archived, and you can do that politely in twenty minutes.

One-Strike Disqualifiers
Some findings end the visit. Not because the supplier is dishonest, but because the risk cannot be managed within your project.
Refusal of a site visit. Full stop. If they will not let you see the plant before you award work, the audit is over and so is the relationship.
No in-house chemical analysis. No spectrometer means alloy composition is taken on trust from the charge supplier. For anything structural, that is not acceptable.
No heat traceability. If a casting cannot be connected to the heat it came from, you cannot investigate a failure later, and neither can they.
Subcontracting that was not declared. Not the subcontracting itself, which is normal, but the silence about it. You are auditing the process, and you were shown the wrong one.
Calibration certificates that do not exist. Measuring with uncontrolled instruments means the inspection records mean nothing, including the ones on your pilot.
Design changes accepted without a drawing revision. If they will modify a part on a verbal request from your buyer, they will modify it without telling you.
Every answer is yes. A foundry that can do everything has either misunderstood your part or is not being straight with you. The credible answer to “can you hold this tolerance?” is often “on these dimensions yes, on that one we would machine it.”
Auditing a Supplier You Cannot Fly To
Most buyers now audit at least some suppliers remotely, and a remote audit is a weaker instrument. It can still be worth doing properly.
Reject recorded sessions. A pre-recorded walkthrough shows you what they chose to film. Insist on a live video call, and ask to see things at a moment’s notice rather than to a schedule.
Send a witness part. Print the drawing, and send a part you already know the answer to. Ask them to measure it and return the report. Do this before revealing the true dimensions. It tests both their metrology and their honesty.
Use a third-party inspector. For a first order on a new overseas supplier, an independent inspection at the foundry before shipment is inexpensive relative to the risk. Specify what to check; a generic inspection report is not much better than no report.
Split the payment on evidence. Structure terms so that a portion is released against the inspection report and a portion against delivery. Suppliers who are confident in their process rarely object.
Get the traceability documents before shipment, not after. Once the parts are on a ship, your leverage over missing paperwork drops to zero.
Scoring by Region: What to Weight
Where a supplier sits changes which checks deserve the most attention. Not because capability varies in principle, but because the failure modes cluster differently.
Western Europe. Weight the environmental and certification checks. CE marking, waste sand handling and emissions compliance are heavily enforced, and documentation is generally strong. The most common surprise is capacity honesty: the plant may be smaller than the sales presentation implies.
India and South Asia. Weight capacity and schedule realism hardest. Process capability is often genuine and the metrology culture is improving quickly, but order books are frequently over-committed. Check the arithmetic in the capacity section carefully, and ask for the delivery record on the last three similar orders rather than the average.
Southeast Asia. Weight subcontracting transparency. Multi-tier supply chains are common and not always declared, which means the process you audited may not be the process that runs your order.
Turkey and Eastern Europe. Weight the tooling and change-control checks, and the export documentation. These suppliers are often strong on execution and export-oriented, and the audit value is in confirming that change control is formal rather than verbal.
China. Weight the metrology, traceability and honesty checks, and spend your time on the floor rather than in the conference room. The best Chinese foundries are genuinely excellent and will hand you the CAR example before you ask. The gap between the best and the average is wider than in Europe, which is exactly why the audit earns its cost.
If you want a supplier directory rather than an audit method, this is not that article. Our Guide pour choisir un prestataire de services de moulage au sable covers how to build a shortlist. The audit is what you do to the shortlist.
From Score to Decision
The score is not the decision. It is the input to one.
A high score with a weak item 25 is not a high score. A supplier that scores 44 but answered your questions evasively will manage the relationship the way it managed the audit. Buy the behavior, not the certificate.
A mid score with a clear improvement path is often the right answer. A score of 34 that comes from four identified, closable gaps is a different proposition from a score of 34 that comes from a diffuse absence of control. Ask for a written closure plan with dates before you place a production order.
Always run a pilot, even after a 48. The audit tells you the process exists. The pilot tells you it can hold your tolerances. Take your hardest part, not the one you are most confident about, and specify the inspection before the parts are made.

Foire aux questions
What is the difference between a supplier audit and a first article inspection?
An audit verifies the process in general: the systems, the equipment, the records, the people. A first article inspection verifies one physical part against one drawing. You need both, and the audit comes first, because there is no point approving a part from a process you have not verified. The audit also tells you what the first article inspection should focus on.
How long should a foundry audit take?
A full day on site is enough for most foundries if you prepare properly. Half a day is workable for a small plant or a narrow scope. What matters more than duration is that you walk the process in order and speak to the people doing the work, not only to the sales team. A two-hour visit that stays in the conference room is a meeting, not an audit.
What should I do if the foundry refuses a site visit?
Treat it as a failed audit. A supplier that will not let a serious prospective customer see the plant before an award is either unable to show you something or does not expect the relationship to last. Neither is a good basis for placing work. If travel is genuinely impossible, propose a live video walkthrough as a minimum, and treat a refusal of that as the same answer.
Should I use a third-party inspection company?
For a first order from a supplier you cannot visit, yes. A pre-shipment inspection by an independent firm costs a fraction of a bad first order and gives you an evidence trail. It is not a substitute for an audit, because an inspector checks what you tell them to check, and specifying that list is exactly the work an audit teaches you to do.
What is the single most useful document to request?
The corrective action report from a real past problem. Every other document shows intent; the CAR shows behavior under stress. You want to see whether the root cause is specific or generic, whether the fix was verified, and whether the customer was informed early or late. Ask for one that has been anonymized, and ask why it happened.
How do I audit a supplier’s claimed capacity?
Do the arithmetic yourself. Take the number of molding lines, the cycle time for a part in your size range, and the shifts worked, and compare the result with the claimed monthly tonnage. Then ask for the delivery performance on the last three orders of a similar part rather than an annual average, because averages hide the orders that went wrong. If the arithmetic requires superhuman uptime, the quoted lead time is not real.
The Audit You Run Once
A well-run supplier audit is not a recurring ceremony. Run it thoroughly once, score it, write down the evidence, and then manage the relationship on delivery performance and quality data rather than on visits.
Two things are worth carrying forward from the visit. The first is that score, with your notes, because it is the only record of what the process looked like before you committed. The second is the impression you formed of item 25. Whether a supplier told you the uncomfortable truth when it would have been easier not to is the best available predictor of how they will behave the first time a delivery slips.
If you would rather not run the audit at all, there is a shortcut worth considering: ask for a pilot on your hardest part and see what happens. A foundry that volunteers the problems with your drawing, tells you which tolerance is marginal, and returns a first article report without being chased has just passed the audit in a different form.
Send us your hardest part file and find out
Articles connexes
- How to Choose a Sand Casting Service Provider
- Small Batch Casting Procurement: How to Buy 5 to 500 Parts
- Casting Defects: Porosity, Shrinkage and Gas Defects in Aluminum and Iron
- Sand Casting Process Guide: Green Sand, Gating Systems and Patterns
- Sand 3D Printing Accuracy and Tolerance
- Rapid Prototype Casting: From CAD to Casting in 5 Steps




