Latest / The CTO Podcast with Fexingo: Technical Leadership, Architecture, and Engineering Org / How Notion Rebuilt Its Engineering Trust
Transcript
- Lucas: If these conversations are useful for what you're building or running, the way we keep this network ad-free is listener support — buy me a coffee dot com slash fexingo — that's F. E. X. I. N. G. O. Luna: Thanks for the heads-up on the link spelling. It really helps when you're trying to look it up between meetings. So, what's the angle today? Are we looking at another case study in scaling? Lucas: We are, but not the usual scale-up story. Usually, when companies hit a certain size, they start building internal platforms to serve their product teams. The problem is those platforms often become bottlenecks. Today, I want to look at how Notion handled the reverse: they had to rebuild trust with their own engineers after years of fragmented tooling. Luna: That sounds like a classic product-led growth hangover. Did they have too many different CI/CD pipelines? Lucas: Exactly. By mid-2024, Notion's engineering org was split into roughly twenty different squads, each with their own preferred deployment tools, logging frameworks, and even coding standards. One squad might use Terraform for infrastructure while another wrote raw CloudFormation templates. It wasn't just messy; it was dangerous because knowledge about how things broke was siloed. Luna: So the CTO decided to centralize? That usually creates resentment. Lucas: It did, initially. But here's the twist: they didn't just build a central platform team and force compliance. They built a Developer Experience team whose only KPI was reducing time-to-first-commit for new hires and time to production for existing features. They measured success by how little friction engineers felt, not by how many rules they enforced. Luna: That's a subtle but huge difference. Enforcing rules feels like policing. Reducing friction feels like service. How did they measure that reduction? Lucas: They tracked something called DORA metrics, specifically Deployment Frequency and Lead Time for Changes. Before the overhaul, the median lead time was four days. After rolling out their internal platform, which standardized the build process and automated testing gates, that dropped to eighteen hours. But the real win was in engineer satisfaction surveys. Luna: Did morale actually go up? Or did people feel boxed in? Lucas: Morale went up significantly. In Q3 of last year, their internal pulse survey showed a twenty-two point jump in the question 'I have the tools I need to be productive.' The reason is simple: developers stopped spending thirty percent of their week fighting configuration files and started spending it on actual logic. It’s the difference between driving a car and changing its tires every ten miles. Luna: Thirty percent is a massive chunk of capacity. Was there pushback from senior engineers who liked their custom setups? Lucas: There was some, but they mitigated it by allowing extensions. The platform wasn't a walled garden; it was a highway with exits. If a squad had a unique requirement, they could write a plugin for the platform rather than bypassing it entirely. This kept the core stable while still allowing experimentation. It’s a hybrid model that balances governance with autonomy. Luna: So the platform team acted more like product managers for the engineers themselves? Lucas: Precisely. They treated the internal platform as a product with users. They held office hours, gathered feedback, and iterated based on pain points. This shifted the dynamic from 'IT department' to 'colleagues helping colleagues.' It’s a cultural shift as much as a technical one. Luna: I’ve seen companies fail at this because the platform team becomes arrogant. They think they know better than the product teams. How did Notion avoid that trap? Lucas: By tying their compensation to customer satisfaction, not uptime. The platform team’s bonuses were linked to Net Promoter Score from the engineering org. If the engineers hated the tools, the platform team didn’t get paid well. This aligned incentives perfectly. No more hidden agendas or power plays. Luna: That’s a clever incentive structure. Did they have to deal with any legacy systems that couldn't be migrated? Lucas: Yes, and this is where it got tricky. There was a monolithic reporting service that had been around since the early days. It was fragile, poorly documented, and terrified everyone to touch it. Instead of rewriting it immediately, they wrapped it in an API layer and slowly decomposed it over eighteen months. This allowed them to stabilize the rest of the system without risking the cash cow. Luna: Strangler fig pattern applied to internal tools. Smart. What about security? Centralizing usually means a single point of failure for vulnerabilities. Lucas: Security actually improved. With a centralized platform, they could enforce consistent security scanning across all builds. Previously, some squads skipped security checks to meet deadlines. Now, security is baked into the pipeline. You can't deploy without passing the scan. It’s automated enforcement, which removes the human element of cutting corners. Luna: Automated enforcement is key. Humans will always optimize for speed over safety if given the chance. Does this model work for smaller startups too, or is it only viable for large orgs? Lucas: Smaller startups can benefit, but they shouldn't copy the scale. For a team of fifty, you don't need a dedicated platform team. You need a designated owner who rotates responsibility. The principle remains: standardize the boring stuff so everyone can focus on the hard stuff. Don't reinvent the wheel for every sprint. Luna: So it’s about cognitive load. Letting developers think about problems, not plumbing. Lucas: Exactly. Cognitive load is the enemy of innovation. When your brain is full of configuration syntax, you have less bandwidth for creative solutions. Reducing that load frees up mental energy for what actually moves the needle for the business. Luna: It reminds me of how Apple designs hardware. They handle the complexity internally so the user just sees simplicity. Is this the same idea applied to software development? Lucas: Very similar. Apple hides the engineering so the experience shines. Notion hid the DevOps so the product features shine. It’s a philosophy of abstraction. Hide the complexity, expose the value. It’s why good platforms feel invisible until they break. Luna: And when they do break, how transparent is Notion about incidents? Do they blame individuals? Lucas: No, they practice blameless post-mortems. Every outage is treated as a system failure, not a person failure. They publish detailed post-mortems internally, sometimes externally. This builds psychological safety. Engineers aren't afraid to report bugs because they know the focus is on fixing the process, not punishing the messenger. Luna: That trust is priceless. It encourages faster reporting and quicker fixes. Without it, people hide mistakes until they become disasters. Lucas: Right. And in a high-velocity environment, hiding mistakes is fatal. Transparency accelerates learning. If everyone knows what went wrong, no one repeats it. It turns failures into institutional knowledge. Luna: I’m curious about the ROI. Beyond speed, what’s the financial impact of this DX investment? Lucas: The financial impact is hard to isolate directly, but the indirect savings are enormous. Reduced churn of senior engineers alone saves millions in recruiting and onboarding costs. Plus, faster feature delivery means earlier monetization. In tech, speed is literally money. Every day saved in deployment is a day closer to revenue. Luna: So it’s an efficiency play that doubles as a retention strategy. Two birds, one stone. Lucas: Exactly. Happy engineers build better products, and happy products drive growth. It’s a virtuous cycle. The key is recognizing that engineering culture isn't soft HR fluff; it's a core operational lever. Luna: Most CEOs still treat it as secondary. Do you think that view is changing in 2026? Lucas: It has to. As AI handles more routine coding tasks, the value shifts to architecture and system design. Those are high-leverage activities that require deep focus and trust. If your culture is toxic or chaotic, AI won't save you. You’ll just automate inefficiencies faster. Luna: So investing in culture is now a hedge against automation risk. Interesting framing. Lucas: Precisely. You want humans doing the thinking, not the typing. That requires an environment where thinking is valued and supported. It’s a strategic imperative, not a nice to have. Luna: One last thing on the platform side. Did they open-source any of their internal tools? Lucas: They did. They released parts of their CI/CD toolkit to the community. This attracts talent and gives back to the ecosystem. It also validates their approach. If other companies adopt your tools, you know you’re solving a real problem. Luna: Good move for employer branding too. Developers love working with modern, open stacks. Lucas: Absolutely. It signals that they care about engineering excellence, not just shipping features. It’s a signal to the market that they’re serious about long-term stability. Luna: Do you think other productivity tools will follow Notion’s lead? Slack or Monday.com? Lucas: Probably. As the market matures, competition shifts from features to reliability. Notion proved that reliability is a feature. Others will have to catch up or risk losing developer goodwill. Luna: It’s becoming a competitive advantage. Like customer support used to be in the nineties. Lucas: Spot on. Developer experience is the new customer experience. Treat your engineers well, and they’ll build products your customers love. It’s that simple, yet so many miss it. Luna: Simple in theory, hard in execution. Thanks for breaking down the Notion playbook. Lucas: Anytime. Next time you log into your repo, ask yourself: is this helping me build, or is this just noise? Luna: A good question to end on. Let’s see if our own dev setup passes the test. Lucas: Let’s find out. Until then, keep building.