One Bad Extrusion, 8,000 Stamped Parts: A Quality Inspector's Story About Automotive Stamping Dies

It started with a phone call on a Tuesday afternoon. A buyer from a mid-sized supplier said, 'We need a bracket assembly. About 200,000 pieces per year. Think you can do it?' I remember looking at the calendar: late March 2024, delivery due in July. We said yes.

Before I go further, here's what I do. I'm the quality and brand compliance manager at Delphi, an automotive metal forming company. We design and build automotive stamping dies in-house, we stamp parts, and we run CNC machining for finishing operations. I review every deliverable before it reaches customers—roughly 120 part numbers and die packages a year. In 2024, I rejected about 8% of first deliveries for tolerance mismatches, bad certifications, or incomplete documentation. The bracket assembly was not one of those first deliveries. It was worse.

Background: The Bracket Project

The assembly had two main components: a stamped steel bracket from a new progressive die, and an aluminum extrusion that we'd CNC machine into a secondary mounting piece. The customer called it 'a simple bracket.' It wasn't simple. It was a mix of three processes: stamping, extrusion, and CNC. That combination is where the trouble usually hides.

The stamped piece was straightforward. We designed an auto stamping die with a progressive layout, ran tryout, and hit dimensions on the first article. The first article report was clean. We checked hole position, flatness, and edge condition. The die produced 500 parts before we ran a five-axis CMM check. It was boring, which is exactly how you want it. The extrusion was another story.

Automotive aluminum extrusions are great for lightweighting, but they behave differently than sheet metal. You're not cutting a flat blank from a coil. You're pushing hot aluminum through a die, then stretching, cooling, and cutting it. That process creates natural twist and bow. If your drawing doesn't specify tolerances for those, the extruder will ship whatever they can within their standard limits.

One thing I've learned: stamping car parts is not just about the press. The die is where quality is decided. But a well-built die can still fail at assembly if the mating part—especially an extrusion—isn't held to the same logic.

Where It Started to Go Wrong

The customer's drawing for the extrusion included overall length, hole positions, and a note: 'Material: 6061-T6.' It did not specify twist. It did not specify flatness. Our quote included a redline: 'Add twist and flatness requirements per ASTM B221 or provide a GD&T callout.' The customer replied, 'We'll handle it.' Fine.

We ordered the extrusion from a supplier we'd worked with before. They delivered 2,000 feet of profile. The stamping side was ready. Our CNC team set up the fixture to machine the extrusion ends and drill mounting holes. Then the tooling engineer came to my desk holding a piece of aluminum.

'Look at this,' he said. He set it on the granite surface plate. It rocked like a seesaw. I used a height gauge to measure the high corner: about 0.4 mm off the plate over a 500 mm span. Converted to angle, that's roughly 3.2 degrees of twist. Our internal limit for that feature was 1 degree. The drawing the customer gave us said nothing. The extruder's own standard was 3 degrees. So technically, they shipped to their standard—but it wouldn't work in our fixture.

I called the buyer. 'The extrusion isn't qualified for our fixture,' I said. 'We need a tighter twist requirement or a straightening step.'

'It's within industry standard,' he said. 'The print doesn't have a twist callout.'

He was right. The print didn't. But his assembly would fail if we machined those extrusions as-is. We had 8,000 stamped blanks already sitting in finished goods. If the extrusion couldn't be used, the stamped parts couldn't ship either. That's when the project stopped being 'a simple bracket.'

I still kick myself for not making our redline a contractual condition. If I'd refused to order material until the drawing was updated, we'd have avoided a three-week delay. The engineer said, 'We'll handle it,' and I let that be enough. My fault as much as theirs. I do not say that to blame the buyer. The drawing was incomplete, and I knew it.

The Twist That Stopped the Line

Here's a side note. I see searches for 'CNC cars' in our analytics. I think the intent is 'CNC machine car parts.' That's what we do: machined brackets, mounts, flanges, often from aluminum extrusions or forged blanks. If you're looking for a supplier, search for that. But before you send a drawing, ask yourself whether your specs actually cover the raw material state. Because a CNC machine can't fix a part that twists before it ever gets clamped.

Honestly, I'm not sure why so many buyers understand tolerance on sheet metal but ignore it on aluminum extrusions. My best guess is that they think of extrusion as a 'profile' not a 'part.' It's a mental gap. They also assume the CNC operation will pull everything into line. But clamping a twisted extrusion into a rigid fixture only masks the problem. The moment you release the part, it springs back. So the machined features may be right in the fixture, but wrong in the assembly.

We ended up doing three things. First, we sent the extrusion back to the supplier and asked for a straightening pass. They had a stretch straightener, but they hadn't run the profile through it because 'the drawing didn't require extra work.' We paid for that operation. Second, we revised the purchase order to reference a twist limit of 1 degree per 300 mm and included a fixture acceptance sample. Third, we wrote a short note for the customer explaining what had happened and why the new requirement belonged on their part drawing.

The customer was defensive at first. Then the engineer asked, 'Can you send me the measurement method?' I did. A few weeks later, he sent back a revised drawing with twist and flatness tolerance added. He also asked us to approve the extrusion supplier going forward. That's exactly what I wanted.

The final delivery was six weeks late—I want to say six, but don't quote me on the exact number. It felt longer. The customer had to pay their own expediting costs. But the parts assembled cleanly, and the same bracket went into continuous production without the assembly issues we'd just found. They also gave us the next revision of the assembly for 2025. We didn't lose the account. We lost the illusion that a drawing can depend on 'common understanding.'

What I'd Tell Any Buyer of Stamping Dies or Machined Parts

So what did I take away from this?

• An auto stamping die can be perfect, but a part assembly fails as a system. The die, the extrusion, and the CNC work all need to share the same tolerance logic.
• If you're buying automotive aluminum extrusions for a machined car part, specify twist and flatness. Don't assume the extruder knows your fixture.
• If a supplier redlines a drawing and says 'add a tolerance,' don't say 'we'll handle it.' Ask them to put the requirement in writing.

According to IATF 16949, the automotive quality management standard, production part approval is a controlled process. But the standard can't make you define every tolerance. You have to do that yourself. I'd rather spend an hour explaining twist tolerance than three weeks arguing about a rejected batch. An informed customer asks better questions and gets better parts.

That's my experience based on about 200 die qualification projects with OEMs and tier suppliers in North America. If you're working in low-volume motorsport or one-piece prototype work, your world might be different. And that's fine.

The person who writes the clearest spec wins. That sounds simple, but it's kinda rare. Not because they're the smartest, but because they've turned assumptions into numbers. The rest of us can't read minds. We can only read prints.