Why Headcount Growth Didn't Move the Productivity Needle
In 2016, Spotify was in the middle of a hiring boom. On paper, the company looked like it was on a smooth upward trajectory. But the numbers told a different story: despite the surge in engineers, overall productivity wasn't improving at all.
Director of Engineering and Head of Platform Developer Experience Pia Nilsson shared on the Thoughtworks podcast that the real diagnostic was an onboarding metric: how long it took a new engineer to land their tenth pull request. At the peak of Spotify's rapid growth, that figure ballooned to over 60 days. Something in the engineering organization was fundamentally broken.
The feedback Nilsson and her team collected from engineers pointed to two main culprits:
- Context switching from fragmentation. Spotify's culture treats each squad like a "little startup" that can charge ahead independently. That autonomy was great for speed in the early days, but as the company scaled, it created a wildly fragmented ecosystem. Engineers were drowning in cognitive load just managing the sprawl.
- Findability was nearly impossible. Engineers couldn't tell which service to integrate with. Should they use the user data service built by the customer service team, the slightly different one from the premium team, or just build their own? The result was more duplication, which made the fragmentation worse.
A Path Between Autonomy and Chaos
Spotify's famous autonomous culture made the obvious fix unworkable. Mandating that every team adopt the same technologies and frameworks would have turned the organization into a monolith — and would have erased the very freedom that made engineers want to work there in the first place.
The solution had to be developer-centric, giving engineers a single place to find everything they needed without imposing rigid rules on how teams worked. That was the genesis of Backstage.
Backstage was designed around three core actions: find stuff, manage stuff, and create stuff.
Find Stuff
Before Backstage, Spotify engineers burned hours each week hunting for documentation, platforms, systems, and their owners across the internet. Backstage consolidates all of that into one unified location, making discovery a matter of seconds rather than a scavenger hunt.
Manage Stuff
Rather than hopping between tabs to check the health of a Kubernetes cluster or the status of a deployment, engineers can now oversee their monitoring, logging, and CI/CD pipeline in one place. Backstage brings those operational tools together, cutting down on the context switching that was sapping productivity.
Create Stuff
When engineers need to spin up a new ML model, data pipeline, or microservice, Backstage lets them do it without reinventing the wheel. Instead of burning time on boilerplate code that duplicates what already exists elsewhere in the ecosystem, they generate new components that are automatically set up according to Spotify's Golden Paths — the company's standard for best practices and tech conventions.
This approach lets Spotify keep its decentralized, autonomous squads while gently steering them toward consistency. Backstage functions as a nudge rather than a mandate: teams stay free to make their own choices, but the platform pushes alignment and prevents further fragmentation. And because Backstage is open source, the plugin ecosystem continues to grow with features beyond Spotify's own use cases.
The Numbers That Justified the Investment
The proof came when Spotify rechecked the onboarding metric. After Backstage was introduced, the time to a new engineer's tenth pull request dropped from over 60 days to just 20 days.
That kind of measurable shift makes the case for investing in developer experience on its own. As Nilsson put it, "If you have numbers like that in your organization, I find that it's easy to get buy-in for investments in developer experience."
For more context on how Backstage addresses developer effectiveness, the full Thoughtworks podcast episode — and Spotify's guide on getting started with Backstage — are good starting points.
