Latest / The Manufacturing Podcast with Fexingo: Factories, Supply Chains, and Industrial Business / How a Michigan Factory Cut Scrap Rate to Near Zero
Transcript
- Lucas: So there's a factory outside Detroit — makes aluminum transmission housings for one of the Big Three automakers — and two years ago, their scrap rate was hovering around eight percent. Industry average for that kind of part is maybe three to four, so eight is bad. Really bad. Luna: Eight percent scrap on a high-volume line — that's not just wasted metal. That's rework, lost machine time, delayed deliveries. Lucas: Exactly. The plant manager told me they were throwing away roughly two million dollars a year in raw material alone, not counting the labor. But here's the thing — they didn't buy any new equipment. No robots, no AI inspection systems. They just changed how they thought about quality. And within eighteen months, their scrap rate dropped to under two-tenths of a percent. Luna: Under 0.2 percent? That's basically zero in a production environment. What was the first thing they changed? Lucas: They started with what they call 'root-cause triggers.' Every single time a part got rejected — whether it was dimensional, a surface flaw, whatever — the operator had to stop and fill out a one-page form before they could run the next part. Not a digital form, a paper card taped to the machine. Luna: Paper? That feels almost retro. Did operators push back? Lucas: They hated it for the first three weeks. Felt like paperwork for paperwork's sake. But the plant manager — I'll call him Dave — he told me they'd never actually tracked why parts were failing. They just scrapped them and moved on. Within a month, patterns emerged. One machine on the night shift was responsible for forty percent of all scrap. One specific drill bit geometry was failing after eight hundred cycles instead of the expected two thousand. Luna: So they didn't know what they didn't know. That's almost every factory I've ever visited, honestly. The data exists — it just never gets collected in a way you can act on. Lucas: Right. The second piece was statistical process control — SPC — but not the kind where an engineer comes by once a shift to take measurements. They trained every operator to take a part every thirty minutes, measure three critical dimensions, and plot them on a control chart that was posted right next to the machine. If a measurement drifted toward the upper or lower spec limit, the operator was empowered to tweak the process themselves. Luna: That's a big culture shift. Most factories I've seen, the operator's job is to run the machine, not to adjust it. That's the setup technician's role. Lucas: Exactly. And that's where a lot of continuous improvement efforts stall — the person who touches the part eight hours a day has no authority to change anything. Dave's team spent three months training every operator on basic SPC. They didn't have to become statisticians — just recognize when a run of five points in a row was trending up, and then make a one-dial adjustment. Nothing complicated. Luna: What about the drill bit issue? You mentioned that was the biggest source of scrap. Lucas: That's the third piece, and it's the simplest one. One of the operators — a guy named Marcus who'd been running that line for twelve years — suggested they install a simple infrared sensor that measured the temperature of the drill bit after each cycle. As the bit wears, friction increases, temperature rises. He said, 'If we know when it's about to fail, we can change it proactively.' Cost about two hundred dollars per machine. Luna: Two hundred bucks. And they'd been running bits until they snapped for how long? Lucas: Decades. The engineering team had never even considered the idea because they assumed operators would just change bits at the scheduled interval. But the scheduled interval was conservative — they were changing bits too early on some shifts and too late on others. The sensor gave them real-time feedback. First year, that single change saved the plant one point four million dollars in reduced scrap and fewer tool changes. Luna: So the three pieces: root-cause tracking, operator-level SPC, and a low-cost sensor. None of them are revolutionary individually. But putting them together — and giving operators real authority — that seems to be the hard part. Lucas: It is. And I think that's why this story matters. You read about 'Industry 4.0' and 'smart factories' and million-dollar AI inspection systems, but this plant did it with paper cards, a clipboard, and a two-hundred-dollar sensor. The real innovation was organizational. Luna: Before we go deeper, Lucas — this kind of conversation, where we dig into real-world manufacturing changes that actually work — that's exactly why we do this show ad-free. If you find these episodes useful in your own work, consider supporting the podcast at buy me a coffee dot com slash fexingo. It's a small way to keep these conversations going without any sponsor interruption. Lucas: Yeah, listener support makes a real difference. Even a few bucks helps us cover production costs and keep bringing you stories like this one. Now, back to Marcus and his drill bit sensor — what happened after the first year? Luna: Right, because sustaining improvement is often harder than the initial gain. Did they see any regression? Lucas: They did, actually. Around month fourteen, scrap started creeping back up to about one percent. Not catastrophic, but the trend was clear. Dave called a meeting with all the operators and asked what was going on. Turns out, the bonus structure was the problem. Luna: Wait — they had a bonus tied to production volume, didn't they? Lucas: Exactly. Operators got a quarterly bonus based on parts produced per shift. And when things were running smoothly, they started skipping the thirty-minute SPC checks so they could produce more parts. They knew they were gambling, but the bonus incentive was stronger than the quality incentive. Luna: So the system was working against itself. Classic misaligned incentives in manufacturing. Lucas: Dave restructured the bonus so that sixty percent was based on first-pass yield — parts that passed inspection on the first try — and only forty percent on raw volume. The operators actually pushed for that change because they realized they were spending more time reworking scrap than they were saving by skipping checks. Within two months, scrap was back below 0.3 percent. Luna: That's the piece most consultants miss. You can install all the sensors and training in the world, but if the compensation system rewards the wrong behavior, people will find a way around it. Lucas: And here's the kicker: after the bonus change, productivity actually went up slightly. Because less time was being spent on rework, the net good parts per shift increased. They were making more money, with less waste, and the operators were earning higher bonuses. It was a genuine win-win. Luna: I want to zoom out for a second. This factory is in Michigan, making auto parts. How transferable is this playbook to, say, a food processing plant or a pharmaceutical line? Lucas: Surprisingly transferable. The root-cause trigger card works anywhere you have a reject. The SPC training is generic. The sensor idea is specific to the process, but the principle — find a cheap way to measure the thing that's about to fail — applies everywhere. I talked to a plant manager in Ohio who makes plastic injection-molded parts, and he adapted the same approach using a simple scale to detect when a mold was starting to build up residue. Cost him fifty dollars. Luna: Fifty bucks to catch a problem before it creates thousands in scrap. That's the kind of story I wish more manufacturing conferences would feature instead of yet another case study about AI on the edge. Lucas: Absolutely. And I think that's the real lesson here: the most effective quality systems aren't the ones with the fanciest technology. They're the ones that give the people closest to the work the data and the authority to fix problems in real time. That's what this plant did. And eighteen months later, they're running at 99.8 percent first-pass yield. Luna: That's a number any factory would love to have. Lucas, do you think they can sustain it long-term? I've seen too many lean initiatives fade after the champion leaves. Lucas: That's the million-dollar question. Dave told me they're now training a second layer of operators to be 'quality champions' — basically peer coaches who can train new hires on the SPC system. They're building the discipline into the onboarding process so it survives personnel changes. It's not guaranteed, but they're thinking about it the right way. Luna: So the key is institutionalizing the change rather than relying on a single person's passion. That feels like a good note to end on. Lucas: Yeah. One factory, two years, no new machines — just a better system. If you're listening and thinking about your own floor, maybe start with the paper card. It's cheap, and it might just surprise you.