Platform Engineering's Golden Path Should Feel Like Freedom, Not Constraints—Are We Measuring Developer Happiness?

I spend a lot of time thinking about developer experience—probably because I came from design where we obsess over user experience. And I keep seeing this pattern: Platform teams build what they think developers need, not what developers actually want.

The promise of platform engineering is compelling: Golden Path that reduces cognitive load by 50%. Self-service infrastructure. Developers focus on features, not on configuring Kubernetes.

But here’s my uncomfortable question: Are we measuring whether developers actually feel that cognitive load reduction? Or are we just assuming our platforms help?

The Design Systems Lesson I Keep Repeating

When we built our design system 3 years ago, we made a classic mistake:

What we built: Comprehensive component library with every possible configuration option. Beautiful, flexible, technically impressive.

What we thought: Designers will love this! So much power!

What actually happened: 40% adoption rate. Most designers kept building custom components.

Why: Our system was TOO flexible. Designers spent more time figuring out configuration than just building the component themselves.

The fix: Simplified to 3 variants per component. Removed 80% of configuration options. Made the default option obviously the right choice for 90% of use cases.

Result: 85% adoption. Designers happy. Cognitive load actually reduced.

The Platform Engineering Parallel

I’m seeing platform teams make the same mistake:

What they build: Powerful, flexible infrastructure platform with lots of configuration options

What they think: Developers will appreciate the flexibility!

What actually happens: Developers confused by options, stick with what they know, platform adoption is middling

What might work better: Opinionated defaults. One obvious way to deploy. Escape hatches for edge cases, but 90% of use cases work with zero configuration.

Platform as Product: Are We Doing User Research?

Michelle mentioned hiring a Platform PM. I LOVE this. Because platforms are products, and products need product management.

But here’s what I want to know: What user research is that PM doing?

When I do user research for design systems, I ask developers:

  • What parts of building UI frustrate you most?
  • Show me your workflow—where do you slow down?
  • What would “easy” look like?
  • What stops you from using the design system?

For platform teams, equivalent questions might be:

  • What infrastructure tasks waste your time?
  • Show me how you deploy a service—where do you get stuck?
  • What would “easy deployment” look like?
  • What stops you from using the platform golden path?

My hypothesis: Many platform teams aren’t asking these questions. They’re building infrastructure they think is helpful, then measuring success by “did we build it?” instead of “are developers actually happier?”

Measuring What Actually Matters

Keisha mentioned developer satisfaction scores (6.2 → 8.4). I want to see MORE of this.

Platform teams should track:

Adoption metrics:

  • % of teams using platform (vs manual infrastructure)
  • % of deployments using golden path (vs custom scripts)
  • % of new services bootstrapped with platform templates

Happiness metrics:

  • Developer satisfaction with infrastructure (quarterly survey)
  • “How easy is deployment?” (1-10 scale)
  • “How much time do you spend on infrastructure vs features?” (time tracking)
  • Net Promoter Score for platform: “Would you recommend our platform to other developers?”

Behavior metrics:

  • Support ticket volume (if platform is self-service, tickets should drop)
  • Time to first deployment for new services
  • Time to productivity for new engineers

If you’re building a platform but not measuring developer happiness, you might be optimizing for the wrong thing.

The Question I’m Sitting With

Luis mentioned “golden path is default, not mandatory—provides escape hatches” and I think that’s the right philosophy.

But in practice: How do you design escape hatches that don’t become the primary path?

In design systems, we provide escape hatches for truly custom components. But some designers use escape hatches for everything because they don’t trust the system.

Platform engineering same risk: If developers don’t trust the golden path, they’ll use escape hatches for everything, and you lose the standardization benefits.

The solution I think works: Make the golden path SO OBVIOUSLY EASIER that escape hatches feel like extra work.

Example from our design system:

  • Using system component: 1 line of code, instant preview, automatic accessibility
  • Building custom: 50 lines of code, manual accessibility testing, no preview

Guess which one developers choose?

For platforms: If golden path deployment is 2 clicks and custom deployment is 2 hours of Terraform, developers will choose golden path. But if golden path is 10 configuration fields and custom is “copy what I did last time,” custom wins.

The Controversial Take

Platform teams should spend 50% of their time on user research and developer experience, and 50% on infrastructure implementation.

Most platform teams I see: 5% user research, 95% building.

Result: Technically impressive platforms with mediocre adoption.

What I’d love to see more platform teams do:

  1. Weekly developer feedback sessions
  2. Observation studies (watch developers use platform, note where they struggle)
  3. A/B testing different UX approaches
  4. Measuring cognitive load (time to complete tasks, error rates, satisfaction)
  5. Treating low adoption as a UX problem, not a training problem

For platform teams: Are you building FOR developers or WITH them? How often do you actually talk to the developers using your platform?

Because reducing cognitive load is the promise. But if we’re not measuring it, we don’t know if we’re delivering.

Maya, YES! This is exactly the conversation we need to have.

We run quarterly developer satisfaction surveys, and the results are illuminating. When we asked “What stops you from using the platform?” the #1 answer wasn’t “missing features”—it was “I don’t understand how to use it.”

That’s a documentation and UX problem, not a technical problem.

Your point about 50% user research / 50% implementation resonates. We implemented bi-weekly “platform office hours” where developers can ask questions, and our platform PM observes where people get stuck. The patterns we’ve found:

  • Developers want examples, not configuration docs
  • “Quick start” guides used 10x more than comprehensive docs
  • Self-service is great IF it’s actually faster than asking for help

The escape hatch question is crucial. We track “% of deployments using golden path” and it’s currently 78%. The 22% using custom approaches? We interview them to understand why. Usually it’s either edge cases (legitimate) or lack of trust in the platform (need to fix).

Measuring cognitive load is hard. We proxy it with: time to first deploy, support ticket volume, and survey questions. But I’d love better methods. How do design teams measure cognitive load quantitatively?

This thread is making me rethink how we evaluate our platform team’s success.

We’ve been measuring: “Did platform team ship the features on their roadmap?” But Maya’s right—we should be measuring: “Are developers actually happier?”

The platform-as-a-product mindset makes total sense from a product perspective. Internal products need product management just like external products.

Question for Michelle and Keisha: Your Platform PMs—what are their success metrics? Is it developer adoption? Satisfaction scores? Time saved?

Because if platform PM is measured on “features shipped” they’ll optimize for that. If measured on “developer happiness” they’ll optimize for UX and adoption.

At our company, we don’t have a platform PM. Our platform tech lead makes product decisions. Is that a mistake? Should platform teams always have dedicated product management?

My hypothesis: Platform teams without PMs build technically impressive but hard-to-use systems. Platform teams with PMs build simpler, more opinionated, higher-adoption systems. Is that accurate?

David asked about Platform PM metrics—great question.

Our Platform PM (reports to me) has these success metrics:

  1. Developer satisfaction score (target: 8/10 or higher)
  2. Platform adoption rate (target: 80% of teams)
  3. Time to first deployment for new services (target: under 4 hours)
  4. Support ticket volume trend (should decrease as platform improves)

NOT measured on: features shipped, lines of code, technical complexity.

This was intentional. If we measured features shipped, she’d optimize for quantity over usability. By measuring adoption and satisfaction, she optimizes for building what developers actually need.

Maya’s point about opinionated defaults is critical. Our PM’s biggest contribution was REMOVING options. Platform team wanted to support 5 CI/CD tools. PM said “Pick one that works for 90% of use cases, make it great.” We went with GitHub Actions, heavy opinions, 90% adoption.

The 10% with edge cases? They use escape hatches, and that’s fine. Better to delight 90% than confuse 100%.

To answer David’s question: Yes, platform teams should have product management. Either dedicated PM or engineer with product thinking. Technical excellence without product sense creates low-adoption platforms.

Maya’s design systems parallel is brilliant and it’s making me question our platform approach.

We’ve been measuring technical metrics (deployment frequency, MTTR, etc.) but not happiness metrics. After reading this, we’re adding to our Q2 platform roadmap:

New metrics to track:

  • Developer NPS for platform
  • Quarterly “How easy is infrastructure?” survey
  • Platform adoption rate (currently don’t track this!)
  • Ratio of golden path deployments vs custom

New processes:

  • Monthly platform feedback sessions with rotating developer attendees
  • Observation studies (watch developers use platform, note friction points)
  • Exit interviews when developers choose custom infrastructure instead of platform

The cultural change management piece is key. We involved developers in platform design from day 1, but Maya’s point about 50% user research is a level beyond what we’re doing.

Challenge: Our platform engineers are infrastructure experts who love building systems. Asking them to spend 50% of time on user research is a mindset shift. It requires either training infrastructure engineers in product thinking, or hiring people with both skills (rare and expensive).

What’s working: Pairing platform engineers with our design team’s user researchers. The researchers run the studies, platform engineers observe and learn.

Question for Maya: How do you measure cognitive load quantitatively in design systems? Can that methodology apply to platform engineering?