Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / Why Hardware Startups Need Design for Assembly from Day One
Transcript
- Lucas: So you've got a prototype that works. The electronics are solid, the firmware boots, the enclosure looks clean. You're ready to go to production. And then your contract manufacturer sends you a quote that gives you actual heart palpitations. Luna: I've heard this story so many times. The quote comes back and suddenly the unit cost is three times what you modeled. Lucas: Exactly. And a lot of the time, the culprit isn't the BOM cost for the components — it's the assembly. The labor, the fixturing, the cycle time, the rework. And there's a whole discipline for preventing that shock. It's called Design for Assembly, or DFA. Luna: And DFA is not the same as DFM — design for manufacturability — right? People lump them together. Lucas: They're related but distinct. DFM is about making individual parts easier to produce — like choosing the right injection molding draft angles. DFA is about making the whole product easier to put together. Fewer parts, simpler motions, less chance for error. The classic framework comes from Boothroyd and Dewhurst — two professors who literally wrote the textbook on it back in the eighties. Luna: And it's still the standard. I've seen their software used at a few contract manufacturers. It's basically a checklist that forces you to ask: does this part need to be separate? Can it be combined? Can it be inserted from one direction? Lucas: Right. The core metric is the DFA index — theoretical minimum assembly time divided by estimated actual assembly time. A high index means a well-designed product. A low index means you're going to spend a lot on labor. I want to talk about a specific example that really drove this home for me. Luna: I'm listening. Lucas: There's a smart thermostat startup — I won't name them, but they were in Y Combinator a few years back. Their first production run was a nightmare. The device had a metal backplate that mounted to the wall, and inside there were four screws, two plastic standoffs, and a separate grounding wire all going into that plate. Assembly took about ninety seconds per unit just for that sub-assembly. Luna: Ninety seconds just for the backplate. Ouch. Lucas: And they were projecting volume of fifty thousand units. So that's one thousand two hundred fifty hours of labor just for that one step. At thirty dollars an hour loaded labor cost, that's thirty-seven thousand five hundred dollars in assembly cost for one part of one sub-assembly. And they hadn't even looked at the rest of the product yet. Luna: So what did they change? Lucas: They hired a mechanical engineer who specialized in DFA. He looked at the backplate and asked: why are the standoffs separate? Why are there four screws when you could use two? Why do you need a dedicated grounding wire when you could use a captive screw that makes contact through the plate? They redesigned it so the standoffs were molded into the plastic inner frame, and they used two self-tapping screws instead of four machine screws with separate nuts. Luna: So they eliminated six parts — four screws, two standoffs — and replaced them with two self-tapping screws and a molded feature. That's a textbook DFA move. Lucas: Exactly. Assembly time for that sub-assembly dropped from ninety seconds to about twenty-five. And the part cost actually went down because they eliminated the separate standoffs and the nuts. The DFA index for the whole product went from about point three to point seven. Luna: Point three is rough. That means the theoretical minimum assembly time was only thirty percent of what they were actually spending. They were wasting seventy percent of their assembly labor. Lucas: And that's not unusual for a first-pass design. Most hardware startups don't think about assembly until the CM gives them a quote. And by then, the design is locked, the tooling is cut, and changes are expensive. The change order for that backplate redesign — modifying the injection mold for the inner frame — cost them about eight thousand dollars. But it saved them over twenty thousand dollars in assembly labor on the first production run alone. Luna: And it kept saving on every run after that. So the ROI on that one change order was basically immediate. Lucas: Right. And the thing is, the founders weren't bad engineers. They just didn't have manufacturing experience. They optimized for electrical performance and for fit and finish, but they never asked themselves: how will this be assembled? How many hand motions does it take? Can a worker do this without special tools? Luna: I think a lot of first-time hardware founders have this mental model where they picture the factory as this magical place where your design just gets assembled by robots. And then they find out that even with automation, you still need to feed parts in the right orientation, and every extra part adds a feeder or a bowl or a robot gripper change. Lucas: That's a huge point. The Boothroyd-Dewhurst method actually starts with a question: can this part be eliminated? And if not, can it be combined with another part? Because every part you add has a cost that goes beyond the material. It has a logistics cost — you have to inventory it, handle it, orient it. It has a quality cost — more parts means more potential failure points. And it has an assembly cost — each part requires a motion, and motions take time. Luna: There's a famous example from Ford back in the day. They redesigned a car door inner panel and reduced the number of parts from something like ninety to forty, and the assembly time dropped by over sixty percent. And that was just from combining stamped metal parts into fewer, more complex stampings. Lucas: Right. And that same thinking applies at the hardware startup scale. It's not about making parts cheaper — it's about making fewer parts. A single molded plastic bracket that combines three functions — locating, fastening, and grounding — is almost always cheaper and faster to assemble than three separate parts that do each thing individually. Luna: But there's a tension, right? Because combining functions means making the mold more complex. So you're trading tooling cost for assembly labor. And for low volumes, the tooling cost might not pay back. Lucas: That's the key judgment call. DFA isn't about minimizing parts at all costs. It's about minimizing total cost — part cost plus tooling cost plus assembly cost plus quality cost. For a startup doing five thousand units, it might make sense to keep parts separate because the tooling savings outweigh the labor. But for fifty thousand or more, the balance shifts. Luna: So what's the practical takeaway? When should a hardware startup start thinking about DFA? Lucas: Day one. Before you even finalize the industrial design. Get a manufacturing engineer in the room during the concept phase. They'll look at your CAD and say things like 'that screw boss should be on the other side so the screw can be driven from the same direction' or 'you can combine these two injection molded parts into one if you use a living hinge.' Those changes are free if you make them before you cut steel. Luna: And after you cut steel, they're expensive. I've seen startups pay fifty thousand dollars for a mold modification that would have been a five-minute CAD change if they'd thought about it earlier. Lucas: Exactly. The thermostat startup I mentioned — they learned that lesson the hard way. But they caught it before their first production run, which is earlier than most. A lot of companies don't realize the problem until they're on the line and the assemblers are struggling. Luna: Yeah, and by then you're losing money every minute. There's a concept called the 'cost of change curve' — it's much cheaper to change a line on a drawing than to change a physical part on the factory floor. DFA is basically about moving those changes to the left on that curve. Lucas: That's a great way to put it. And if you want a quick self-assessment, there's a simple heuristic: count the number of unique fasteners in your design. If you have more than three types of screws — say, Phillips pan head, Phillips flat head, and Torx — you've already added handling complexity. A worker has to change bits or grab a different driver. Each change adds a few seconds, but over thousands of units, those seconds become days. Luna: And if you can eliminate fasteners entirely — snap fits, press fits, ultrasonic welding — that's even better. Though of course, those have their own design constraints. Lucas: Right. But the first step is just to ask the question. The thermostat startup went from a DFA index of point three to point seven just by asking 'do we really need these four screws?' And that single question saved them tens of thousands of dollars. Luna: It's one of those things that seems obvious in hindsight, but in the middle of a hardware development sprint, you're so focused on functionality and reliability that assembly just doesn't get attention. Lucas: And that's exactly why we talk about it here. If these conversations are useful for what you're building or running, a couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Luna: Yeah, it really does make a difference. Even a few contributions help cover hosting and research time. Lucas: So back to DFA — the last thing I'll say is that the Boothroyd-Dewhurst method is available in a free online calculator. You can input your part count, your assembly time estimates, and get a DFA index. It's a fifteen-minute exercise that could save you months of pain. Luna: And if you want to go deeper, there's a great book called 'Product Design for Manufacture and Assembly' by Boothroyd, Dewhurst, and Knight. It's the bible for this stuff. Lucas: It is. And the core message is simple: design for assembly isn't an afterthought — it's a design parameter you set before you start. If you treat assembly cost as a constraint from the beginning, you end up with a product that's cheaper, more reliable, and faster to bring to market. Luna: And less likely to give you heart palpitations when the CM quote comes in. Lucas: Exactly. That's the goal.