January 2024, Monday morning. I was standing in our receiving bay with cold coffee and a colder feeling in my stomach. The first batch of forged steel flanges from a new supplier had just arrived—1,200 pieces.
Spec called for ASTM A105 carbon steel, 150mm outer diameter, ±0.1mm tolerance. My calipers bounced between 149.72mm and 150.31mm across the batch.
Look, I'm not trying to be dramatic. A 0.6mm swing on a high-pressure piping system isn't nitpicking—it's a fit problem. And a fit problem on our line becomes a field failure for our customer.
We rejected the whole batch. The redo cost the supplier $18,000 and pushed our project back a week.
That was the fourth rejection in six months. And the moment I realized the problem wasn't really about any one supplier—at least not entirely. It was about how we evaluated them.
Let me back up. I'm a quality and compliance manager at a mid-sized aerospace components company. I review every metal batch before it reaches our production floor—roughly 350 to 400 batches a year. In 2024, we rejected about 12% of first deliveries, mostly for dimensional inconsistency or mismatched material certs.
Q1 2024, we landed a new contract requiring a mix of parts: brass CNC turned parts for a fuel assembly, aluminum forging blanks for structural brackets, and a run of brass Y hose connectors for a cooling system. Our customer asked if we could find one vendor to handle both the brass forging and the CNC turning.
Simple enough, right?
It was not simple.
The Aluminum Forging That Looked Fine—Until It Didn't
We started with aluminum forging. One supplier—call them Vendor A—quoted competitively. Their samples passed every check. Surface finish was clean, hardness tested within spec, dimensional report was spotless.
First batch arrived. Looked good. I approved it.
Three weeks later, our CNC shop flagged internal stress issues in the forging blanks. After machining, nearly 30% of the parts warped during secondary operations.
I called the vendor. Their response: "Industry standards allow for this kind of variance. You're being too strict."
Maybe they were right. Maybe I was too strict. But our tolerances didn't care about industry standards—they cared about whether the part fit.
We scrapped the batch. $14,000 in rework. The lesson? A good-looking sample tells you nothing about production consistency if the supplier doesn't control their own process.
The Brass CNC Parts and the Vendor Who Said Yes to Everything
After the flange fiasco, we switched suppliers. Vendor B answered every question fast—almost suspiciously fast. Their qualifications looked solid. Price was right. They took on the brass CNC turned parts and said they could handle the brass Y hose connectors too.
First batch of brass CNC turned parts arrived. Dimensions checked out. Surface finish was acceptable. I approved.
Second batch, same thing.
Third batch—tolerance drift began. Nothing catastrophic, but each piece was just barely outside spec. We measured 40 pieces. Seventeen were marginal.
I called their quality contact. Long pause. "We may have subcontracted part of the run," they admitted.
Subcontracted. To a shop running older cam-operated lathes that couldn't hold the tolerance.
Here's something vendors won't tell you: when a supplier says "we can do that" without any hesitation, they're either impossibly good or they don't actually know their own limits. Vendor B was the second kind.
We lost six weeks and roughly $8,000 between rework and expedited shipping on that one.
The Supplier Who Said "We're Not the Best Fit for That"
That rejection hit hard (ugh, again). But it also reset my thinking.
We had a smaller batch of brass Y hose connectors coming up—about 400 pieces. High precision on the internal bore and thread engagement. Not a complex part, but tolerance-sensitive.
This time, I tried a different approach. I described the part in detail to three suppliers. The first two said yes immediately. The third—Vendor C—asked questions for twenty minutes. Then their sales engineer said: "We can do this. But if you ever need high-volume forging with quick turnaround, we're not your best option. We're set up for precision turning on small-to-medium runs."
That honest boundary earned my trust for everything else.
We gave Vendor C the job. First article inspection report came back without us asking. They included process capability data (Cpk 1.67 across critical dimensions). A short video of the machining setup. Everything.
200-piece lot inspection: zero rejections. Zero.
I kept second-guessing myself for the first two days after approving them. What if the second batch drifts like the others? But they didn't. Batch two, same numbers. Batch three, same.
What I Wish I'd Known Earlier
Most buyers focus on price per piece and completely miss process control capability. I was guilty of that. We'd been treating "can quote a part" as the same as "can hold tolerance at volume." They're wildly different things.
The question everyone asks is "what's your best price?" The question they should ask is "what do you do best—and what should I go elsewhere for?"
After four rejections across three suppliers, here's what I changed in our process:
- Written acceptance criteria for every batch. Not "industry standard." Actual measurements, actual sampling plan, actual disposition process for non-conforming material. No more judgment calls at receiving.
- Ask the uncomfortable question. "What's your weakest process?" If a supplier can't answer honestly, that tells me more than their ISO certificate.
- Track rejection data by supplier and batch. We built a simple spreadsheet—took maybe two hours to set up. Peaked at 12% rejection rate in Q1, down to under 2% by Q4. The data didn't lie.
Total cost of that learning curve: about $22,000 in rework, expedited freight, and lost production time. But the bigger cost was the trust we burned with our customer on the first delayed shipment (note to self: we've since improved that relationship, but it took a year).
So the flange rejection? It wasn't the supplier's fault. It was mine for expecting a generalist to be a specialist.
We still use vendors for different strengths now. Vendor C gets all our precision brass turning. For larger forging runs, we work with a different shop that only does forging—and tells us when a job is too small for their setup.
Would I have believed any of this before those four rejections? Probably not. Sometimes you have to measure a thousand wrong parts before you learn to ask the right question.