Latest / The Manufacturing Podcast with Fexingo: Factories, Supply Chains, and Industrial Business / How a Japanese Factory Trains Workers with Decades-Old Skills
Transcript
- Lucas: There's a factory in Osaka that's been making industrial cutting tools since 1952, and they have a rule: no new hire touches a CNC machine until they've spent two years learning to sharpen blades by hand. Luna: Two years of hand-sharpening? That sounds like a huge investment in something a robot could do in about two minutes. Lucas: Right, and that's exactly the tension I want to unpack. The company is called Yamato Precision Tools — about forty employees, revenue around eight million dollars a year. They make knives for food processing plants, for paper mills, even for semiconductor dicing. And the founder's grandson, who runs it now, told me that when they tried to fully automate their sharpening line in 2018, defect rates actually went up. Luna: So the robot made things worse? That's counterintuitive. Lucas: It is, until you understand what a skilled sharpener is actually doing. A master craftsman doesn't just follow a fixed angle — they feel the burr forming on the blade edge, they listen to the sound of the stone, they adjust pressure in real time based on the steel's hardness that day. That sensory feedback loop is incredibly hard to encode into a machine. Luna: And this is where I imagine a lot of factories in the West would say, 'Well, we just buy better sensors.' But I'm guessing there's something deeper here. Lucas: There is. And before we go deeper — if these kinds of manufacturing stories have made you think differently about your own work, even in a small way, a handful of listeners chip in monthly at buy me a coffee dot com slash fexingo. It's what keeps this show ad-free and allows us to visit places like Yamato. So if that resonates, we appreciate it. Luna: Yeah, it's a small thing that makes a big difference for us. And honestly, it's listener-funded reporting that lets us spend time on stories like this one. Lucas: So back to Yamato. What they realized is that the tacit knowledge of their veteran sharpeners — some with forty years on the job — was never written down. It existed only in their hands and ears. So they started a formal program to map that knowledge. Luna: How do you map something that's essentially muscle memory? Lucas: They created what they call a 'skill matrix.' For each type of blade — there are over two hundred — they break down the sharpening process into about thirty discrete steps. For each step, they rate the worker's proficiency from one to five. A level one means you can follow a checklist under supervision. Level five means you can teach it and also diagnose problems by sound alone. Luna: So they quantize the unquantifiable. Did it work? Lucas: They've been running the program for seven years now. Their overall defect rate is below 0.05 percent. Industry average for precision tooling is around 0.3 percent. So that's a six-fold improvement. And they've cut training time for complex blades from about eighteen months down to twelve, because the matrix helps new hires focus on exactly what they don't know. Luna: I want to linger on that defect rate. 0.05 percent — that's five defective parts out of ten thousand. In a shop that still does a lot of manual work. How much of that is the skill matrix versus just having really good veterans? Lucas: Great question. The plant manager told me that the matrix itself isn't the magic — it's the fact that it forces veterans to articulate what they do. Before the matrix, a master sharpener would just say, 'You'll get it with time.' Now they have to explain, for example, that for a 12-inch blade with a 60-degree bevel, you should use a 1000-grit stone for the first pass, but only if the steel hardness is below 58 Rockwell. That kind of specificity gets captured and passed on. Luna: So the matrix is really a documentation tool disguised as a training program. It's almost like a knowledge management system for the hands. Lucas: Exactly. And it changes the economics of the business in a subtle way. Because Yamato can now guarantee a level of consistency that their competitors — many of whom have fully automated lines — struggle to match. Their largest customer, a Japanese food processing conglomerate, audits them every six months and has never found a batch variance above their tolerance. Luna: That's a serious competitive moat. But I wonder — is this model replicable? I mean, not every factory makes high-precision cutting tools. Can a car parts plant or a plastics injection molder apply this? Lucas: I think the principle is replicable, even if the specifics aren't. Any manufacturing process that relies on human judgment — and there are more than we like to admit — can benefit from this kind of explicit knowledge capture. The key is that you have to have veterans who are willing to be recorded and observed. And you need a culture where sharing knowledge is rewarded, not hoarded. Luna: And in Japan, there's a long tradition of that — the master-apprentice model, the reverence for craft. But in a Western factory, where the incentive structure might be more individualistic, could this work? Lucas: There are actually examples. A German auto parts supplier I visited last year does something similar for their welding line. They have a 'welding academy' where senior welders record video commentary on their own work, explaining why they chose a particular torch angle or travel speed. They've reduced rework by 40 percent. So it's not uniquely Japanese — but it does require a shift in mindset from 'automation replaces humans' to 'automation augments human expertise.' Luna: That's a nice phrase. But let's push on it — isn't there a risk that by codifying tacit knowledge, you freeze it? You create a standard operating procedure that becomes dogma, and then no one improves it because 'the matrix says so.' Lucas: Yamato actually addressed that. They have a rule that the skill matrix must be reviewed and updated every quarter. And any worker — even an apprentice — can propose a change. If a new hire discovers a faster way to set up a jig, that gets tested and, if validated, incorporated into the next version. The matrix is living document, not a bible. Luna: So the hierarchy is flat when it comes to knowledge. I like that. But let's talk numbers — what did this cost to implement? Lucas: The plant manager estimated about $120,000 over the first two years, mostly in time — veterans spent roughly four hours a week documenting and training. But they calculated the return: reduced scrap, fewer customer returns, and lower rework costs added up to about $200,000 in savings annually. So payback in under eighteen months. Luna: That's a no-brainer ROI. So why don't more factories do this? Lucas: I think two reasons. One: it requires admitting that your current training isn't good enough. That's hard for plant managers who pride themselves on efficiency. Two: it takes patience. The benefits aren't immediate. You have to trust that the upfront investment in documentation will pay off down the line. In a quarterly-results world, that's a tough sell. Luna: And yet the Japanese approach has always taken the long view. I'm reminded of the concept of 'kaizen' — continuous improvement — but this is almost the opposite: discrete improvement through explicit knowledge transfer. Lucas: It's a different flavor. Kaizen is about small, incremental changes to the process. What Yamato does is more like preserving institutional memory. And I think as the baby boomer generation retires — not just in Japan, but globally — this kind of structured apprenticeship is going to become critical. There's a demographic cliff coming. Luna: Right. In the US, the average age of a skilled machinist is something like 55. When they leave, they take decades of know-how with them. If you haven't captured it, you're starting from scratch. Lucas: Exactly. And that's where I think Yamato's model is a template. You don't have to be a tool and die shop in Osaka. Any factory with aging skilled workers — and that's most of them — can start a skill matrix program tomorrow. The hardest part is getting the first master to sit down and describe what they do. Luna: And that requires trust. The master has to believe that by sharing their secrets, they won't make themselves redundant. Lucas: Right. Yamato addressed that by tying pay to teaching. A master who trains three apprentices to level three or above gets a bonus. So sharing knowledge is directly rewarded. And the company makes it clear that the master's job is secure — they're not replaced by their students; they become supervisors of a larger team. Luna: So the incentive structure aligns. That's the missing piece in a lot of factories. They want knowledge transfer but they don't reward the transferors. Lucas: Precisely. And I think this is one of those rare manufacturing stories where the human element is the competitive advantage, not a cost to be minimized. Yamato's edge isn't a faster robot it's a deeper understanding of steel. Luna: Which makes me wonder — as we push toward lights-out factories, are we losing the ability to make the hardest things? The stuff that requires judgment, feel, experience? Lucas: That's the question I keep coming back to. Full automation works great for high-volume, low-variety production. But for high-mix, high-precision work — which is where a lot of manufacturing profit lives — human skill still matters. And maybe the winning factories of the next decade will be the ones that combine the best of both: robots for the repetitive, humans for the nuanced. Luna: And they'll have documented that nuance so it doesn't walk out the door at retirement. Lucas: Exactly. That's the takeaway from Yamato: treat your veterans as knowledge assets, not just labor. And build the system to capture what they know before they leave.