Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / The Hardest Part of Hardware Is the Last Ten Percent
Transcript
- Lucas: You look at a successful hardware startup and it looks inevitable. The product is polished, the packaging is tight, the reviews are glowing. But what you don't see is the six to eighteen months between the working prototype and the product that actually ships without catching fire or bricking itself on the first charge. Luna: Right — the 'last ten percent' problem. I've heard that phrase before, but I'm not sure most people understand how brutal that stretch actually is. Lucas: It's the part of the journey that kills more hardware startups than bad ideas or weak product-market fit. You can have a demonstrable prototype, a thousand preorders, glowing press — and still die because you couldn't get from unit 1,000 to unit 10,000 without a defect rate that bleeds you dry. Luna: And the canonical example here has to be Pebble, right? The smartwatch that basically invented the category before Apple showed up? Lucas: Pebble is the textbook. They raised over 10 million dollars on Kickstarter back in 2012 — at the time, the most funded campaign in history. They shipped over a million watches. They had passionate users, a vibrant app ecosystem, and they still folded in 2016 and sold their assets to Fitbit for pennies. Luna: So what broke? They clearly had demand. They had the engineering. What was the bottleneck? Lucas: The bottleneck was that the last ten percent of the product — the part that makes a prototype into a reliable, shippable, return rate manageable device — requires a completely different skill set than the first ninety percent. Designing a circuit board that works on a bench is different from designing one that survives a drop test, an electrostatic discharge, and a hundred thousand button presses. And even if you nail the design, you then have to navigate the foundry, the injection molder, the final assembly line — and every one of those partners is calibrated for volume. Luna: So the contract manufacturers just don't want to deal with small batches? Lucas: They'll take your money, but they'll quote you a per-unit price that assumes fifteen percent scrap, and they'll prioritize their Toyota or Apple contracts when line time gets tight. If you're a startup ordering 20,000 units, you're a nuisance. And if you have a quality issue, you don't have an engineer stationed at the factory to catch it — so you discover the defect after 5,000 bad units have already shipped. Luna: That was literally Pebble's story with the original Pebble Steel. I remember reading that the first batch had a high rate of screens detaching from the body. They had to replace thousands of units — which for a startup with thin margins was devastating. Lucas: Exactly. And the economics are brutal. If your bill of materials is, say, 80 dollars per unit, and you retail at 150, your gross margin is about 70 dollars. But if ten percent of units come back as defective, you're not just losing the 80 dollars of materials — you're eating the cost of return shipping, the refurbishment, and the lost sale. One bad batch can wipe out the profit on the entire production run. Luna: So how do you avoid that trap if you're a hardware founder today? Because it's not like the ecosystem has gotten more forgiving — supply chains are still tight, and investors are still impatient. Lucas: The smartest founders I've seen are the ones who treat the journey from prototype to production as a separate engineering discipline — not an afterthought. They hire a 'design for manufacturing' engineer before they've even finalized the industrial design. They run pilot runs of 500 units, not 5,000, and they personally inspect every single one for failure modes. And they build relationships with small, nimble contract manufacturers in places like Shenzhen or Guadalajara that specialize in exactly this high-mix, low-volume niche. Luna: So the classic advice is 'fail fast' — but in hardware, failing fast on a production run can bankrupt you. Lucas: Right. And that's the paradox. Software can deploy a buggy feature and patch it overnight. Hardware ships its mistakes in boxes, and each mistake costs real dollars. I think one of the hardest lessons for software-background founders is that hardware doesn't forgive optimism. Luna: Makes me think of Framework, the laptop company. They seem to have sidestepped some of this by designing for repairability from day one — but even they had a rough launch with battery issues and supply chain delays. Lucas: Framework is a great example of transparency as a strategy. They published their teardown guides, they sold replacement parts directly, they basically said 'we know this isn't perfect, but you can fix it yourself.' That builds trust and reduces the cost of defects because a user can swap a bad battery instead of returning the whole laptop. But even then, their first batch was delayed by months because a single connector supplier couldn't meet spec. Luna: That's the thing — no matter how good your design is, you're dependent on suppliers you can't control. One capacitor shortage, one mold that's out of tolerance, and your timeline blows up. Lucas: And the investors who funded you on a 12-month timeline start getting nervous. That's actually what killed Pebble in the end. They had a great product, but they burned through cash paying for returns and factory tooling. When they tried to raise a Series B, the VCs looked at their margins and their inventory and walked away. Luna: So the VC model — which is built for software with zero marginal cost — just doesn't fit hardware. Lucas: Not without adaptation. There are funds now that specialize in hardware, and they underwrite based on supply chain maturity, not just user growth. They expect a founder to have a manufacturing partner lined up before they write the check. And they're patient about timelines — they know hardware takes 24 months, not six. Luna: But what about the startups that don't have that option? The ones bootstrapping or raising small angel rounds? Lucas: Then you have to be ruthless about limiting complexity. Use off-the-shelf components even if they're not perfectly optimized. Design your enclosure to be molded in steel tooling, not aluminum, because steel is cheaper and faster to iterate. And don't try to build a smart home hub with twelve sensors — build a smart plug with one sensor and nail that first. Luna: Simplify until the only thing left is the one thing you can't take out. I like that. Lucas: If today's episode was useful to you and you want to keep it ad-free, buy me a coffee dot com slash fexingo helps. Luna: Yeah, that really does make a difference — keeps us independent, keeps us from having to shoehorn in sponsored segments. Lucas: So back to the last ten percent — I think there's a psychological dimension too. Founders fall in love with the prototype. It's their baby. And when the factory sends back a batch that doesn't work, they want to blame the factory, not the design. But in my experience, if you have a yield problem, it's almost always a design problem that the factory just exposed. Luna: That's a hard pill to swallow. You've been working on this for a year, you've got 50 engineers, and the guy on the factory floor with a micrometer is telling you your beautiful design has a tolerance that's physically impossible to hold. Lucas: And that's where the best hardware founders I've seen differentiate themselves. They fly to the factory. They sit with the line operators. They take the feedback and redesign. They don't send angry emails — they send revised CAD files. And they build a relationship with the production manager, who will then go out of their way to flag a potential issue early. Luna: So the human element is still huge, even in an age of automated assembly lines. Lucas: Enormous. I talked to a founder who makes industrial sensors — he said the single best decision he made was bringing his lead manufacturing engineer to Shenzhen for three months during the pilot run. That engineer caught a misalignment in the pick and place machine that would have caused a 40 percent failure rate. He caught it because he was standing there. Luna: Forty percent. That's an existential threat for a startup. Lucas: Absolutely. And the fix was a 20-cent adjustment to the nozzle. But if he'd been in California, that 20-cent fix would have been a 200,000-dollar disaster. Luna: So the lesson is: if you're starting a hardware company, budget for boots on the ground at the factory. That's not in most pitch decks. Lucas: It's rarely in the budget at all, which is why so many startups run out of money before they reach production quality. They optimize the BOM cost but ignore the cost of rework, scrap, and returns. And those costs are nonlinear — one bad batch can be 10 times more expensive than the batch itself. Luna: Can you give me a concrete number? Like, what's the real cost of a single returned unit for a 150-dollar product? Lucas: Sure. Assume 80 dollars in components and assembly. Add 15 dollars for inbound shipping to the customer. They return it — that's another 10 dollars in return shipping. You inspect it — maybe 5 dollars in labor. If it's repairable, you spend 20 dollars on parts and labor. If it's not, you write off the whole 80 dollars. And you still have to ship out a replacement, which costs another 80 plus 15. So a single return that can't be repaired costs you roughly 190 dollars — more than the original sale price. And if your return rate is 5 percent, that's eating almost all your margin. Luna: Yeah, that math is brutal. And it explains why a startup that looks profitable on paper can be losing money on every unit they sell. Lucas: Right. And that's why the last ten percent is where hardware startups go to die. It's not the idea, it's not the design — it's the execution at scale, with real materials, real tolerances, and real customers who expect the thing to work out of the box. Luna: So what's the one piece of advice you'd give to a founder listening who's about to ship their first production run? Lucas: Ship 50 units to friends and family first. Not 500 to Kickstarter backers. Run those 50 through every use case you can imagine. Drop them. Get them wet. Leave them in a hot car. And then, only then, scale to 500. And when you go to 500, station someone at the factory. It's boring, it's expensive, but it's the cheapest insurance you'll ever buy. Luna: Yeah, that's the kind of advice you only get after you've been burned. Or after listening to this show, hopefully. Lucas: Hopefully. Because the hardware startup graveyard is full of products that worked perfectly on a lab bench but couldn't survive the real world. And the only way to avoid that fate is to fall in love with the problem of manufacturing, not just the problem of design. Luna: Alright, I think that's a good place to leave it. Thanks for the deep dive. Lucas: Thanks, Luna. Talk next time.