Latest / The CTO Podcast with Fexingo: Technical Leadership, Architecture, and Engineering Org / How Figma Scaled Design Systems Without Losing Creativity
Transcript
- Lucas: There is a persistent myth in tech leadership that scaling a product requires slowing it down. You hit a certain size, you add layers of approval, you introduce governance, and suddenly the thing that was fast becomes heavy. But if you look at what Figma did between two thousand twenty-two and now, the data tells a different story. They actually doubled their feature release velocity while simultaneously tightening their brand consistency by nearly forty percent. Luna: That sounds counterintuitive. Usually, when I see velocity go up like that, quality goes down. How did they prevent the chaos? Lucas: It wasn't magic. It was structural. They stopped treating design and engineering as separate silos with a handoff ceremony in the middle. Instead, they built what they call an engineering liaison model. Every major product squad has a dedicated engineer whose sole job is to translate design intent into technical constraints before any code is written. It shifts the conversation from 'can we build this?' to 'how do we build this efficiently?' Luna: So it's less about gatekeeping and more about early alignment. Does that mean designers have to learn code? Lucas: Not necessarily. But they have to speak the language of constraints. The key tool here is their internal design token system. You know how most companies have a style guide that lives in a PDF nobody reads? Figma embedded those tokens directly into their development environment. So when a designer changes a color variable, it propagates to the codebase automatically. It eliminates the drift that usually happens during implementation. Luna: I've seen that work for small teams, but does it hold up when you have thousands of contributors? Doesn't that create version control nightmares? Lucas: That was the exact concern. Their solution was to treat design tokens as first-class citizens in their source control. They implemented a merge-check process where any change to a core token triggers an automated regression test across all active components. If a new primary blue breaks a button in the dark mode theme, the pipeline rejects it. It forces discipline without requiring human review for every single change. Luna: Okay, so automation handles the consistency. But what about the creative side? Do designers feel stifled by all these rules? Lucas: This is the part people get wrong. They think design systems are about restriction. In reality, good systems are about liberation. When the basics are handled by the platform, designers can focus on the novel parts of the experience. Figma reported that their internal NPS score for design tools went up because creators spent less time fighting CSS and more time solving user problems. They traded micro-management for macro-freedom. Luna: That makes sense. Less busy work means more room for innovation. But how do they keep the system fresh? Systems tend to become stagnant over time. Lucas: They run what they call quarterly architecture audits. A rotating team of senior engineers and designers reviews the token library to identify deprecated patterns. If a component hasn't been used in three releases, it gets flagged for removal. It prevents the bloat that kills performance. You end up with a leaner, faster system that only contains what is actually needed. Luna: Removing dead weight. That’s hard to sell to the person who built that component originally though. Human ego is always in the way of clean code. Lucas: Exactly. So they gamified it slightly. Engineers who contribute to the cleanup effort get recognition in their performance reviews. It’s not just about keeping things tidy; it’s about contributing to the health of the platform. It shifts the culture from ownership of assets to stewardship of the ecosystem. Luna: Stewardship over ownership. That’s a powerful mental shift for individual contributors who are used to owning their features. Lucas: And it scales. Because once the foundation is stable, you can spin up new squads without rebuilding the wheel. When Figma entered the AI space last year, they didn't start from scratch. They leveraged the existing interaction patterns and token structure, which cut their time to market by six months compared to previous major initiatives. Luna: Six months is huge in this market. So the ROI isn't just cleaner code, it's actual speed to revenue. That’s what executives care about. Lucas: Right. And it reduces risk. Fewer custom implementations mean fewer bugs in production. Their incident rate dropped by thirty percent after they standardized the component library. Stability attracts users. If your app feels consistent and reliable, people trust it more with their data. That’s a competitive moat that doesn’t show up on a balance sheet immediately, but it matters immensely. Luna: Trust as a product feature. I like that. But is this model replicable for smaller startups who can’t afford dedicated liaison roles? Lucas: You don't need a dedicated role at the start. You need the mindset. Even a two-person team can implement a simple token dictionary and a weekly sync to align on design decisions. The principle is universal: communicate early, automate the boring stuff, and treat your design system as a product itself. If you ignore it, it will eventually ignore you, and then everything slows down. Luna: So the lesson is: invest in the infrastructure early, even if it feels like overhead. It pays off later. Lucas: Precisely. Overhead is just interest you pay on future chaos. Paying it early keeps the principal manageable. Luna: I think a lot of listeners are going to relate to that. We’ve all paid that interest the hard way. Lucas: We have. And looking at Figma’s trajectory, it seems like the gap between design and engineering is closing permanently. The best products are being built by hybrid teams that share a common vocabulary. Luna: It’s less about titles and more about shared outcomes. When everyone cares about the same metrics, the friction disappears. Lucas: Absolutely. Shared outcomes drive shared solutions. And that’s where the real efficiency gains come from. It’s not about working harder; it’s about removing the barriers that make work hard in the first place. Luna: Well said. It’s a reminder that structure, when done right, is an enabler, not a constraint. Lucas: If these conversations are useful for what you're building or running, we deliberately don't run ads on these. If you want to support that choice, the link is buy me a coffee dot com slash fexingo — that's F. E. X. I. N. G. O. Luna: Thanks for keeping this independent. It lets us dig into the details without worrying about sponsor agendas. Lucas: Back to the topic. One final thought on the liaison model. It works best when those liaisons rotate. You don't want the same person explaining design to engineering forever. Rotation spreads knowledge and prevents bottlenecks. Luna: Rotation ensures that the knowledge stays in the team, not in one person's head. That’s crucial for resilience. Lucas: Exactly. Resilience comes from redundancy. If one person leaves, the system doesn't collapse. That’s the mark of a mature engineering org. Luna: And that maturity takes time. It’s not something you bolt on overnight. It’s a cultural evolution. Lucas: It is. And for leaders watching, the question isn't whether you can afford to invest in it. It's whether you can afford not to. The cost of technical debt compounds faster than you think. Luna: So start small. Align early. Automate the basics. Let the rest follow. Lucas: Start small, but start now. Because the longer you wait, the steeper the climb gets. What’s one small alignment step you could take in your current sprint?