Latest / The CTO Podcast with Fexingo: Technical Leadership, Architecture, and Engineering Org / How Wix Rebuilt Its Website Builder for 200 Million Users
Transcript
- Lucas: Wix has two hundred million users. Which is a staggering number for a website builder — but for years, that platform ran on a single monolithic PHP codebase. Luna: I remember using Wix back in like 2012. It worked, but it was definitely... sluggish. Lucas: Exactly. And the engineering team knew that to keep growing — and to keep users happy — they had to rebuild the entire editing experience from the ground up. Not just a facelift, but the core architecture. Luna: So where did they start? Lucas: They started with the editor. The original Wix editor was a flash-based app — yes, Flash. Then they moved to HTML and JavaScript, but it was still a giant single-page app with a lot of tightly coupled logic. They decided to break it into micro-frontends and move the backend to a service-oriented architecture. Luna: Micro-frontends — that's a term that gets thrown around. How did Wix actually implement it? Lucas: They built what they call 'Velvet' — a framework that lets different teams own distinct parts of the editor. So the team that owns the text editor ships their code independently, the team that owns the image gallery ships theirs, and they all integrate into a common shell. Think of it like a lego system where each brick is a self-contained feature. Luna: And the backend? That was the monolith, right? Lucas: Right. The backend was a single PHP application handling everything — user data, site storage, rendering, everything. They migrated to a set of microservices written in Node.js and Go. But the key insight: they didn't do a big-bang rewrite. They used the strangler fig pattern — extracting one service at a time while the monolith still ran. Luna: So users didn't notice anything changing from day to day. Lucas: Exactly. They'd extract, say, the image optimization service into its own service, test it, route traffic gradually, and only then cut over. Over five years, they extracted dozens of services. They also built a custom API gateway to route requests between old and new systems transparently. Luna: That's a long migration. Five years? Did the business ever get impatient? Lucas: I'm sure there was pressure. But Wix's CTO at the time, Avishai Abrahami, made the case that this was essential for future growth. And they had a concrete number to prove the benefits: page load times dropped by 40 percent after the new editor was fully rolled out. Luna: Forty percent — that's huge for user retention. Especially on mobile. Lucas: Yeah, and mobile was a big motivator. The old monolith was built for desktop browsers. To make the editor work fast on phones, they needed to rethink how they rendered pages. That's where WebAssembly came in. Luna: WebAssembly in a website builder? That's interesting. What part? Lucas: They used WebAssembly to run the layout engine client-side. Instead of sending every user interaction to the server and waiting for a response, the browser can compute the new layout locally using precompiled C++ code. That made dragging and dropping elements feel nearly instantaneous — even on lower-end devices. Luna: So the editor feels native. Smart. But maintaining backward compatibility with millions of existing sites — that must have been a nightmare. Lucas: It was one of the hardest parts. Every existing Wix site was built using the old editor's data model. The new editor uses a completely different data schema. They built a translation layer that converts old site data into the new format on the fly, without breaking any existing functionality. And they had to ensure that every feature from the old editor — even obscure ones — worked identically in the new one. Luna: That's the kind of engineering that's invisible to users but absolutely critical. If one widget breaks, you lose trust. Lucas: Right. And they tested this at scale — they ran automated regression tests against thousands of real production sites. They also gave users the option to switch back to the old editor for months after launch, just in case something felt off. Luna: Speaking of trust — one thing I love about this show is that it's ad-free. You don't have to sit through sponsor breaks to get this kind of deep dive. Lucas: Exactly. We keep it that way intentionally. If these conversations are useful for what you're building or running, and you want to support that choice, the link is buy me a coffee dot com slash fexingo. No pressure, ever. Luna: And it genuinely helps. Every contribution means we can spend more time on episodes like this one. So thank you. Lucas: Alright — back to the architecture. One other clever thing Wix did was to build what they call 'Corvid' — now called Velo — a pro-level development platform on top of their editor. But that required yet another architectural layer. Luna: Velo lets users write custom JavaScript and connect to external APIs, right? Lucas: Yes. So they needed to support server-side code execution for millions of sites. They built a sandboxed Node.js runtime that runs in a dedicated cluster, isolated from the main editor infrastructure. Each site gets its own container, and they use a custom orchestration layer to manage scaling. Luna: That sounds like a lot of complexity. How do they keep it manageable? Lucas: They invested heavily in observability. They use distributed tracing across all their services — every request gets a trace ID that flows from the API gateway through to the backend services. When something breaks, they can pinpoint the exact service and even the exact line of code. Luna: So they built a culture of monitoring from day one of the migration. Lucas: Absolutely. They also implemented feature flags so they could roll out new capabilities gradually. For example, the new editor was initially available only to new users, then they enabled it for existing users in waves. That way if something went wrong, they could pause the rollout without affecting everyone. Luna: That's a textbook approach to risk management in platform engineering. Lucas: It is. And it paid off. The new architecture now handles over a hundred billion API requests per day. And the team can ship updates to the editor multiple times a day — compared to once every few weeks with the monolith. Luna: So the migration wasn't just about scaling — it transformed their velocity. Lucas: Exactly. And I think that's the real lesson for anyone running a platform at scale: the migration itself becomes an investment in future speed. Wix spent years rebuilding, but now they can iterate much faster than competitors who are still on older architectures. Luna: What about the cost? Running microservices and WebAssembly runtimes isn't cheap. Lucas: They've been transparent about that. The infrastructure costs went up — more servers, more services. But the performance improvements led to higher user engagement and lower churn. They also optimized aggressively: they use spot instances for non-critical workloads and built a custom caching layer that reduces backend calls by 60 percent. Luna: So it's not just a tech play — it's a business decision. Lucas: Right. And the numbers back it up. Wix's revenue has grown steadily as they've expanded their user base. The new architecture let them introduce features like ai powered site design and multi-language support much faster. Luna: AI design? That's another layer. Lucas: It is. They have an AI that asks you a few questions and generates a whole site. That kind of feature is only possible because the backend services are decoupled and can be orchestrated independently. Luna: So the architectural transformation enabled product innovation down the line. Lucas: Exactly. And that's the takeaway I'd leave listeners with: when you're planning a major rebuild, think about what product capabilities it unlocks in the long term. The migration itself is painful, but if you do it right, it's the foundation for your next decade of growth. Luna: And they did it with two hundred million users watching. Lucas: That's the part that impresses me most. Wix managed to rebuild the engine while the plane was in the air — and the passengers never felt a thing.