Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Digital Twins to De-Risk Manufacturing Scale-Up
Transcript
- Lucas: Last week I was talking to the CTO of a robotics startup that hit a wall trying to go from fifty units to five thousand. One number caught my attention: they spent almost two hundred thousand dollars on tooling for a production line that never ran correctly. Luna: Two hundred grand just to find out the process didn't work? Lucas: Exactly. And that's where digital twins come in. Instead of building physical prototype lines, you simulate the entire assembly process in software first. It's a practice that aerospace and automotive have used for years, but now it's filtering down to startups. Luna: So you can catch problems before spending real money? Lucas: Right. There's a company called FlexSim, or Siemens offers Tecnomatix—both are established platforms. But there are also cheaper, more accessible options. For a hardware startup, using a digital twin can cut the number of physical iterations by half. One study from Deloitte showed that early adopters saw a thirty percent reduction in manufacturing costs. Luna: Thirty percent is huge. But how much time does it take to set up a digital twin? I imagine it's not trivial. Lucas: It depends on complexity. For that robotics startup, they spent about three weeks building the model and running simulations. That three weeks saved them months of physical rework and the two hundred thousand in scrapped tooling. The key is to focus on the bottleneck processes—don't simulate the entire factory floor at first. Luna: So you're saying you don't need a full-scale model to get value? Lucas: Not at all. Start with the most expensive or riskiest step. For a lot of hardware startups, that's the assembly line where you integrate multiple components. Or the curing process for adhesives. Identify the step where errors are most costly, and model that first. Luna: That makes sense. I know a founder who used digital twin software to optimize their PCB assembly line. They found that changing the order of component placement saved thirty seconds per board. Over ten thousand boards, that's huge. Lucas: Exactly. Small optimizations compound quickly. And now we have tools that are specifically designed for small and medium manufacturers. Companies like Visual Components or Simio offer lower-cost licensing. There are even open-source options, though they require more technical know-how. Luna: Which open-source ones are worth looking at? Lucas: There's OpenModelica for systems modeling, and for discrete event simulation, you can use JaamSim. They won't have the polish of commercial software, but if your team has modeling skills, they can work. I'd say if your startup is bootstrapped, start with a free trial of FlexSim—they have a student version—and see if it fits. Lucas: Before we go deeper, I want to mention something. If these conversations are useful for what you're building or running, a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo. That's literally what funds making this many of these episodes. Luna: Yeah, it's an honest model—no ads, just listeners who find value. It keeps us free to explore the niche topics like this one. Lucas: Exactly. So back to digital twins—another angle is using them for supply chain simulation. You can model how a delay in a component delivery might ripple through your production schedule. That's something startups rarely plan for, but a digital twin can surface those dependencies early. Luna: Can you give a concrete example? Lucas: Sure. A medical device startup I know used a digital twin to simulate different supplier lead times. They found that if one specific capacitor was delayed by two weeks, they'd have to idle the entire line for a day. That awareness let them stockpile that component and negotiate better terms. Luna: That's a smart use case. But I imagine the data input is a challenge. How accurate does the twin need to be? Lucas: It doesn't have to be perfect. Start with rough estimates from your machine specs or historical data. The simulation will still show you relative improvements. Over time, as you gather real data, you refine the model. It's iterative. Luna: So it's more about directional insight than absolute precision? Lucas: Exactly. If the simulation says you can increase throughput by fifteen percent by rearranging workstations, that's probably directionally correct even if the exact number is off. And once you implement it, you measure and update. Lucas: A lot of founders ask me about the software learning curve. I'd say the bigger barrier isn't the tool—it's defining your process clearly. You have to know the cycle times, the failure rates, the material flows. If you have that data, building the twin is straightforward. Luna: So the real work is in understanding your own manufacturing process deeply. Lucas: Right. And that's valuable whether you build a digital twin or not. The act of mapping the process forces you to think through every step. One founder told me that just creating the process flow diagram revealed three obvious bottlenecks they hadn't noticed. Luna: Interesting. So even if you never run a simulation, the exercise is worth it. Lucas: Exactly. But if you do run a simulation, you can test scenarios that would be impossible in the real world. What happens if you double the speed of a conveyor? What if you add a second QA station? Those are cheap to test in software. Luna: And that leads to better capital allocation. Instead of buying equipment and hoping it works, you simulate the ROI first. Lucas: Precisely. I've seen startups avoid buying an expensive robot arm because the simulation showed it would become a bottleneck elsewhere. They saved a hundred thousand dollars. Luna: That's the kind of story that makes digital twins a no-brainer for anyone scaling hardware. What's your take on the future? Will this become standard practice? Lucas: I think within five years, any hardware startup that raises Series A will be expected to have a digital twin of their production process. Investors want to see that you've de-risked the scale-up. It's becoming a hygiene factor. Luna: So the question for founders listening might be: have you mapped your production process as a model yet? Lucas: That's exactly the question. And if you haven't, start small. Pick your most expensive step, model it in a free tool, and run a few what-if scenarios. You'll likely find something worth changing. Luna: It's a low-risk way to make high-stakes decisions. I'm all for that. Lucas: Same here. And that's the real takeaway: simulation doesn't replace engineering judgment, but it makes it a lot more informed.