Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Design for Manufacturability
Transcript
- Lucas: When the original Roomba launched in 2002, it had about fifty parts. The team at iRobot had deliberately designed it so that every single screw served at least two functions — the chassis was a structural element, a heat sink, and a mounting bracket all in one plastic molding. Luna: Wait — that's not an accident, right? That's a specific design philosophy. Lucas: Exactly. It's called Design for Manufacturability — DFM. And it's probably the single biggest reason hardware startups fail to scale. They design a brilliant prototype, but nobody asks 'can this be built for ten dollars at a thousand units per hour?' Luna: Wow. And I think a lot of founders hear 'prototype' and 'production' and think they're the same thing with more volume. Lucas: They're not. A prototype is a proof of concept. A production design is a proof of manufacturability. And if you don't design for how parts are actually molded, assembled, and tested, your unit economics will kill you. Luna: This is exactly the kind of conversation I wish I'd had before my first hardware project. And if it's useful to you too, honestly, if today was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. Lucas: Yeah, it's a small way to keep the show ad-free and focused on stuff that actually helps builders. So — back to DFM. One of the classic rules is: every fastener you add costs more than the price of the screw. Luna: Because it takes time to pick it up, align it, drive it. Lucas: Right. A ten-cent screw can add a dollar to the build cost when you account for labor, tool wear, and quality risk. So the Roomba team used snap-fits and self-locating features wherever possible. Luna: And that also reduces parts count. Fewer SKUs to manage, fewer suppliers. Lucas: Exactly. iRobot's engineers counted every single part. They had a target — I think it was under sixty unique components for the whole robot. And they hit it. That discipline is what let them sell the Roomba at two hundred dollars when competing robots cost five times that. Luna: So what's the first thing a startup should look at when they do a DFM review? Lucas: Part consolidation. Can two or three parts be combined into one? For example, a plastic enclosure could have molded-in bosses, ribs, and snap latches instead of separate brackets and screws. Luna: But won't that make the mold more complex? Lucas: Yes. It's a trade-off. You might pay more for the tool, but you save on every single unit thereafter. For high-volume products — say over ten thousand units — the amortization usually favors consolidation. Luna: And for lower volumes? Lucas: Then you want to avoid complex molds with slides and lifters. Use simple two-plate molds and assemble separate pieces. But even then, you can still use snap-fits and reduce fasteners. Luna: Right. Let's talk about tolerances. I've seen startups specify plus or minus one millimeter on a plastic part that they then expect to snap together perfectly. Lucas: That's a recipe for scrap. Tighter tolerances cost more — it's not linear, it's exponential. A tolerance of plus or minus 0.1 millimeter might cost twice as much as plus or minus 0.2 millimeter. Luna: So where should you put the tight tolerances? Lucas: Only on mating surfaces that affect function. If two parts need to align precisely, that's a tight tolerance zone. Everywhere else, open it up. And design for 'self-locating' — use features like tapered edges and alignment pins that guide parts into place. Luna: Like how a USB plug only fits one way, but it's forgiving. Lucas: Exactly. That's 'foolproofing' — poka-yoke in lean manufacturing. If a part can be assembled incorrectly, eventually it will be. So design so it can only go together one way. Luna: What about materials? A lot of startups choose exotic plastics because they look cool or feel premium. Lucas: And then they discover that material has a six-week lead time and requires a special mold temperature. Stick to commodity materials — ABS, polycarbonate, nylon — unless you absolutely need something else. The Ninja blender uses ABS for its base, and it's a fantastic product. Luna: Speaking of Ninja — the SharkNinja team is famous for their DFM discipline. I read that they design their blenders so the motor mount, the base plate, and the control housing all share the same screw pattern. That way they can use the same line for multiple models. Lucas: That's brilliant. It's designing for a product platform, not just a single product. If you can reuse the same sub-assembly across three products, your volume per part goes up, your cost per part goes down. Luna: And you're also reducing risk. If one product fails, you still have demand for that part. Lucas: Right. Another DFM principle is to design for the assembly process. If you're planning to use automated pick and place for electronics, don't put components on both sides of the board unless you have to. That requires a second reflow pass or wave soldering, which adds time and cost. Luna: And for mechanical assembly, you want to design so that all parts are inserted from the top — 'top-down assembly.' Gravity helps. Lucas: Exactly. If an operator has to rotate the product to add a screw, that's time. If you can add all fasteners from the same direction, you save seconds per unit. And seconds multiply. At a hundred thousand units, ten seconds per unit is almost three hundred hours of labor. Luna: So what's the one mistake you see most often in early-stage hardware? Lucas: Not involving the manufacturing partner early enough. Founders design the whole thing, then send it to a contract manufacturer and ask 'can you build this?' And the CM says 'sure, but it'll cost double.' Luna: Instead, you want to have a DFM review with your CM before you finalize the design. They'll tell you 'this wall thickness is too thin, this radius is too tight, this draft angle is zero.' Lucas: Exactly. And those changes are cheap on a CAD model. They're expensive on a steel mold. So bring your CM in at, say, the alpha prototype stage, not after you've ordered tooling. Luna: And if you're bootstrapped and can't afford a dedicated manufacturing engineer? Lucas: Then use checklists. There are established DFM guidelines from places like the Society of Manufacturing Engineers. Start with the basics: minimize part count, standardize components, avoid undercuts, use uniform wall thickness, add generous fillets. Luna: One thing I've found helpful is to physically assemble a prototype of your product with the actual tolerances you'll have in production. If you can't snap it together easily, your factory workers will struggle too. Lucas: Yes. And test with parts that are at the limits of your tolerance — not just nominal dimensions. If your design works with the worst-case part geometry, it'll work on the line. Luna: So to wrap up: DFM isn't a phase you do after the design is done. It's a mindset you apply from day one. Lucas: Exactly. And the payoff is a product that scales. The Roomba wouldn't have become a category-defining product if it cost eight hundred dollars to build. DFM made it possible to sell at a price that changed how people clean their homes. Luna: And that's the real win — not just a cheaper product, but a product that actually gets out into the world. Lucas: Right. So next time you're designing a part, ask yourself: 'could this be one piece? Does this fastener need to exist? Could this tolerance be looser?' The answers will save you money and headaches. Luna: Great advice. Thanks Lucas. Lucas: Thanks Luna. Talk to you next time.