In my first year of quality management, I thought my job began at an inspection table. Measure a sample, pull the bad ones aside, write a report, move on. It took a $4,300 mistake to teach me what I now say at the start of every new hire training: by the time a bad part exists, the quality failure has already happened.
To give you some background: I'm the quality manager at GSP, a metal parts manufacturer that makes stamping parts, progressive dies, molds, CNC machined car parts, forging parts, and aluminum extrusions. Most of our orders go to automotive and heavy-duty customers, usually in quantities from a few hundred to several thousand pieces. I've been doing this for nine years. I've also personally approved—or failed to stop—four significant mistakes over that time. Together they cost roughly $10,000 in scrap, rework, and expedited freight. Actually, $9,817.50. I looked up the number while writing this, because rounding it up made it feel worse.
Here's the position I've landed on: the decision that decides whether a part will ever become defective usually happens before the material order is placed. Not at the press, not at the CNC, and not at final inspection.
The mistake that rewired how I think
In the spring of 2022, we quoted an order for 4,500 stamped brackets. The project looked straightforward. Material grade was specified, hole sizes were clear, the bend radius made sense for the tooling we planned. The quote went out, the customer approved it, and we started cutting.
What we didn't realize was that the customer had sent an updated drawing revision that same morning. The new revision changed the width of a mounting slot from 12 millimeters to 10 millimeters. The email with the updated PDF sat unread in our shared mailbox for four working days. By the time someone opened it, we had already stamped the full run.
The parts weren't ugly. They were useless. We scrapped 4,500 good-looking brackets, remachined a die section, and reran the order. Total damage: about $3,900 in material and tooling, plus a week of delay that our customer did not appreciate. The customer didn't shout at us, which somehow made it worse.
That mistake was not caused by a worn die or an out-of-tolerance CNC spindle. It was caused by a gap in the process—one that a two-minute check would have caught before the first blank hit the press.
What most people get wrong about part defects
If you asked someone outside this industry how automotive parts fail, they'd probably guess machines: a die wears, a spindle drifts, a forging cracks. That does happen. But in my experience, the expensive failures usually start in the space between what the customer assumed and what we assumed.
Take something as boring as a thermostat housing. It looks simple—a machined casting with a few holes and a sealing face. But the drawing may not say which surface is the datum for the coolant outlet. It may not say whether a surface finish of Ra 1.6 is a requirement or just a nice-to-have. If both sides assume the other side understood, nobody discovers the problem until the first parts arrive at assembly. Then the line stops, and the conversation gets tense.
Here's something a lot of suppliers won't tell you: when a drawing is ambiguous, there's a huge temptation to say “no problem” and quote it anyway. Asking a question feels like admitting weakness. And nobody wants to lose a quote to the competitor who says yes faster. That is how 4,000-piece mistakes are born.
We have a rule now at GSP: no quotation goes out until someone has listed the open questions that were not answered. It doesn't have to be a long list. Sometimes it's just one email line: “Assumed datum A is primary, pending your confirmation.” That single habit has removed more risk than any optical measuring machine we own.
The part that surprises customers: sometimes we slow you down
This is the part that catches people off guard. My team will occasionally tell a buyer to hold their purchase order for 24 hours while we ask for a clarification.
That sounds like the opposite of good service, especially to a procurement manager whose own boss is chasing them for a lead time. But I'd rather lose a job during quoting than lose it when our name is on a container of bad parts.
In late 2023, a prospective customer sent us an aluminum extrusion design with one note: “surface must be smooth.” That note means almost nothing. A surface can feel smooth to a buyer's hand and still fail the adhesion test that happens after anodizing. We asked three questions and sent our quote a day later than the other bidders. I assumed that was the end of the story.
The customer came back a week later and said, “You were the only supplier that asked. The last one shipped twice and argued twice. We ordered 8,000 units from you.”
I don't tell that story to make us look good. I tell it because I used to be the kind of engineer who would rather look agreeable than look careful.
What about urgent orders?
At this point, somebody always says: “That's fine for normal orders, but we have a line-down situation. We need parts in four days, not four weeks.”
Fair enough. Urgency is real in this industry. But urgency is exactly when the checks get skipped, and skipped checks are how small problems become emergency shipments. The five minutes you spend opening the latest drawing revision, or confirming which hole is the datum, are cheaper than the $9,817.50 I once spent learning that lesson.
Automotive quality already has official gates: first article inspection, PPAP, capability studies. Those systems generate a lot of paperwork. But a report only helps if someone actually reads it before production starts. Signing a report after the run is just documentation. Checking before the run is quality.
I'll also be honest about the limits of what I'm saying. My experience is based on mid-volume automotive orders—hundreds to thousands of parts—not million-part annual production runs. If you're launching a platform that will run for seven years, your process needs to be even more formal than what I'm describing. And if you're making a one-off bracket for a prototype, you probably don't need the same level of checking. What you do need is the habit of asking yourself one question before you approve anything: what am I assuming here?
Why I keep the checklist
I now maintain the pre-production checklist my team uses before any job starts. It began after the 2022 mistake as a half-page of questions:
- Have we all looked at the latest revision, not just the file name?
- Which surfaces are the datums, and does the customer agree with our interpretation?
- Are vague notes like “smooth” or “remove sharp edges” backed by a measurable tolerance?
- Is a material certificate or test report required before we cut?
It's not a brilliant document. It's common sense written down. In the past 18 months, those questions have caught 47 potential issues before they became production problems. I can't prove exactly how much money that saved, but I know what one of those issues would have cost, because I paid for a similar one in 2022.
I don't share this because our shop is perfect. It isn't. We still make mistakes, and I still write them down. But I'm convinced that most quality problems are not prevented at inspection. They're prevented in the five minutes before someone says “that should be fine.”
Check the drawing before you check the shipment. It's cheaper, and it works.
Leave a Reply