Platform as Product Failed at Our Company—We Had All the Right Pieces. What Went Wrong?

I need to share a failure story that’s been eating at me for the past six months.

Our Series B SaaS company invested over $2M in building an internal developer platform. We had everything the Gartner reports said we needed: a dedicated platform team of six talented engineers, full executive support, modern tech stack (Kubernetes, ArgoCD, internal developer portal). We studied the success stories from Spotify, Airbnb, and Netflix. We were going to be the 80% that Gartner predicted would have platform teams by 2026.

Eighteen months later, we have less than 10% adoption. Our platform team is demoralized. Our developers are still using kubectl and ad-hoc bash scripts. The beautiful internal developer portal we built gets maybe a dozen pageviews per week, mostly from the platform team checking their own work.

What We Got Wrong

Looking back with brutal honesty, here are the mistakes that killed us:

1. We Built “Enterprise-Grade” Before Anyone Used It

We spent the first nine months building RBAC, audit logging, multi-tenancy, and cost allocation. You know, all the things an “enterprise-grade” platform needs. By the time we shipped something developers could actually use, they’d already built their own workarounds. The window of pain had closed.

2. We Assumed We Knew What Developers Needed

Our platform team was staffed entirely with infrastructure engineers—brilliant people, but they approached this as a technical challenge, not a product. We never did user interviews. We never asked developers what their biggest pain points were. We built what we thought would be impressive, not what would be useful.

3. We Treated It As a Technical Project, Not a Product

We had engineering sprints, technical roadmaps, and architecture reviews. What we didn’t have: customer development, usage metrics, adoption funnels, or satisfaction surveys. We measured story points completed, not developer time saved.

4. We Spent Six Months Building a Portal

We invested enormous effort into a beautiful internal developer portal—dashboards, service catalogs, documentation sites. It became a way for the platform team to feel productive while not actually solving the core deployment bottleneck that was costing developers 4 hours per deploy.

5. We Had No Measurement Strategy

When executives asked “Is it working?”, we pointed to features shipped and uptime percentages. We couldn’t answer “How many developers choose to use this?” or “How much time are we saving?” because we’d never defined those metrics.

The Painful Realization

After reading some of the research on platform engineering failures, I realized we’d fallen into almost every trap. The 60-70% failure rate within 18 months? We’re part of that statistic. The average 10% internal adoption rate outside of Spotify-like companies? That’s us. The “golden cage syndrome” where you build too much governance before delivering value? Check.

The most painful realization: We built what we thought was impressive, not what was useful.

Our developers didn’t need a comprehensive platform with enterprise features. They needed someone to solve the single biggest bottleneck in their workflow—the 4-hour deployment cycle. If we’d solved just that one problem in six weeks instead of building a comprehensive platform in eighteen months, we’d be in a completely different position.

The Path Forward

I’m having hard conversations with leadership now about whether to pivot or wind down. Both feel like valid options. The sunk cost fallacy is real—there’s pressure to “just push for adoption” or “mandate usage”—but I’ve seen enough research to know that mandated adoption fails when developers have alternatives.

My Questions for This Community

  1. Has anyone else been through a failed platform initiative? What did you do?
  2. If you were in my position, would you try to pivot (focus on solving one real problem) or wind it down and learn the lessons?
  3. For those who’ve successfully built platforms: What did you do in the first 6-8 weeks that created momentum?

I’m sharing this because I wish someone had been this honest with me 18 months ago. Maybe it’ll help someone else avoid these mistakes.

I’ve seen this exact pattern from the other side—I inherited a failed platform effort at my previous company, and your post could have been written about that experience. The pain in your words is real, and I appreciate your honesty.

The Infrastructure-Only Staffing Red Flag

Your second point hits hard: “staffed entirely with infrastructure engineers.” This is such a common pattern, and I’ve been guilty of perpetuating it myself. We hire brilliant SREs and platform engineers because, well, we’re building infrastructure, right? But that’s the fundamental mistake.

Platform engineering is maybe 30% infrastructure and 70% product management. The technical problems—Kubernetes, CI/CD pipelines, observability—those are actually the easy parts. The hard part is figuring out what to build and why developers would choose to use it.

What Actually Worked: Start With One Pain Point

When I took over the failed platform at my last company, we did something radical: we stopped building the platform for three weeks. Instead, we just talked to developers. We watched them work. We asked them what sucked the most about their daily workflow.

The answer wasn’t “we need a comprehensive internal developer platform.” It was: “Deployments take 4 hours and I have to babysit them.”

So we focused on just that. One problem. We built a single script that automated the deployment process—no portal, no RBAC, no fancy UI. Just a command that took deployment from 4 hours to 45 minutes.

We shipped it in three weeks.

Two teams tried it. Both kept using it, because it actually saved them time. Word spread. Other teams started asking for access. We added features based on real usage patterns, not hypothetical requirements. Six months later, we had 15 teams using it voluntarily.

The Missing Ingredient: Product Discipline

Here’s what I learned: Platform engineering without product discipline is just infrastructure with a portal.

You need:

  • User research: What problems are developers actually facing?
  • Prioritization: Which single problem, if solved, would have the biggest impact?
  • Iterative delivery: Can you ship something useful in 2-4 weeks?
  • Measurement: Are you saving developers time? Are they choosing to use it?

When you have those, the technical implementation becomes straightforward. Without them, you end up building technically impressive things that nobody wants.

My Advice: Pivot, Don’t Wind Down

You asked whether to pivot or wind down. My vote: pivot.

You already have the hard parts:

  • Leadership support
  • Budget
  • A team that (hopefully) still has energy
  • Developers with real pain points

What you need to change:

  • Add product management discipline (hire a PM or train someone to think like one)
  • Pick ONE developer pain point to solve
  • Ship something useful in 6 weeks
  • Measure actual adoption and time saved

If it works, great—you’ve found your wedge. If it doesn’t work after a genuine pivot (not just surface changes), then you wind down with real lessons learned, not just “we built it but they didn’t come.”

The Question Nobody Asks

The question I wish more platform teams would ask: “What developer workflow, if we made it 10x better, would cause developers to voluntarily switch to our platform?”

Not “What features should we build?” Not “What does Spotify do?” But: What pain is so acute that solving it would make adoption inevitable?

Find that pain point. Solve it perfectly. Everything else follows.

What’s the single most painful developer workflow at your company right now?

This is so painfully common it hurts to read. I’ve been on both sides of this—once as the VP who greenlit a failed platform initiative, and once as the VP who inherited one and had to make the pivot-or-kill decision. Your vulnerability in sharing this is exactly what this community needs.

The Measurement Gap Is Killing Platforms

Your point about having no measurement strategy resonates deeply. According to recent platform engineering research, 29.6% of platform teams still don’t measure success at all. And get this: 36.6% rely on mandates to drive adoption.

That second stat is the scariest one. Mandated adoption worked in 2020 when developers had fewer options. In 2026? With GitHub Actions, Vercel, Railway, and dozens of other alternatives that developers can spin up with a credit card? Mandates fail spectacularly. Developers route around obstacles, including internal mandates.

The Question Nobody Asked You

Here’s the question that should have been asked on Day 1: “What would make a developer choose to use this platform instead of their current workflow?”

Not “use it because their manager said so.” Not “use it because we sunset the alternatives.” But choose it voluntarily because it makes their life better.

If you can’t answer that in one clear sentence, you’re building the wrong thing.

The Metrics Framework I Wish I’d Had Earlier

When I took over platform engineering at my current company, I implemented a simple metrics framework. Four key measurements:

1. Time to First Deploy

Baseline question: How long does it take from git push to production today?
Target: Platform must be faster, or developers won’t switch.
Our numbers: 4 hours → 30 minutes (that’s the threshold where adoption took off)

2. Voluntary Adoption Rate

Not mandated usage. What percentage of teams/developers choose to use the platform when they have alternatives?
Target: >60% voluntary adoption in 12 months
Red flag: If you need mandates to hit 20%, the platform isn’t solving real pain

3. Developer Satisfaction Score

Quarterly surveys, NPS-style. “How likely are you to recommend our platform to another developer?”
Target: DSAT >7/10
Reality check: If developers wouldn’t recommend it, they’re only using it under duress

4. Support Friction

Ticket volume, time to resolve. A good platform has fewer tickets over time, not more.
Target: Decreasing support volume as adoption increases
Warning sign: If you’re hiring more platform engineers just to handle support, something’s wrong

The Critical Realization

In your post, you wrote: “We built what we thought was impressive, not what was useful.”

That sentence deserves to be framed in every platform team’s office.

Developers don’t care about impressive. They care about: Does this save me time? Does this reduce my cognitive load? Can I ship faster without thinking about infrastructure?

If the answer to those questions is “no,” then it doesn’t matter how enterprise-grade, how well-architected, or how feature-complete your platform is.

Pivot vs. Wind Down: Here’s My Framework

You asked for advice on pivot vs. wind down. Here’s the decision tree I’ve used:

Consider pivoting if:

  • Leadership is still genuinely supportive (not just sunk cost fallacy)
  • The platform team hasn’t completely burned out
  • You can identify ONE specific developer pain point with measurable impact
  • You’re willing to hire/develop product management capacity
  • You can commit to 6-week MVP with clear success metrics

Consider winding down if:

  • Leadership support is only about “not admitting failure”
  • The team is demoralized and burned out
  • Developer trust is broken (they’ve been promised solutions before)
  • You can’t articulate a single, clear pain point to solve
  • The organization isn’t ready to treat platform as a product

In your case, based on what you’ve shared, I’d vote for pivot with one major caveat: You need to add product management discipline immediately. Not eventually. Not “once we prove the concept.” Now.

The Uncomfortable Truth

Most platform failures aren’t technical failures. They’re product failures disguised as engineering projects.

You had the technical talent. You had the budget. You had leadership support. What you didn’t have was someone asking: “What customer problem are we solving, and how do we know when we’ve solved it?”

That’s not an engineering question. It’s a product question.

What I’d Do If I Were You

If leadership approves the pivot:

  1. Week 1: Hire or assign a product manager to the platform team
  2. Weeks 1-2: Developer interviews. No building. Just listening. Find the #1 pain point.
  3. Weeks 3-6: Build the absolute minimum solution to that ONE pain point
  4. Week 6: Ship to 1-2 volunteer teams (not mandated)
  5. Week 7-8: Measure: Are they still using it? Are they faster? Would they recommend it?

If the answer to those questions is “yes,” you’ve found your wedge. Expand to more teams.

If the answer is “no,” you’ve learned fast and cheaply. Try a different pain point or wind down with lessons learned.

One More Thing

You wrote: “I wish someone had been this honest with me 18 months ago.”

That sentence is why I’m writing this long response. We need more honest conversations about platform failures. The Spotify and Airbnb success stories are great, but they’re not representative. The 60-70% failure rate is the reality most of us face.

Thank you for starting this conversation. I’m genuinely rooting for your pivot.

What’s the ONE developer workflow that’s the most painful at your company today? If you can answer that in one sentence, you’ve got a shot at turning this around.

This reads like a classic “build for builders” problem, and honestly, it’s giving me flashbacks to a failed design system I built a few years ago. Different domain, exact same mistakes.

The Design Systems Parallel

When I was at my last company, I spent a year building a comprehensive design system. 150 components. Every variant. Every edge case. Beautiful documentation site. Storybook with hundreds of stories. I was so proud.

Adoption rate? Maybe 8 components were actually used. The rest sat there looking impressive but solving zero real problems.

Why? Because I built what I thought designers and developers needed, not what they were actually struggling with. Sound familiar?

The One Question That Changes Everything

Here’s what I wish I’d asked (and what I think you should ask):

“Did anyone on your platform team actually watch developers work?”

Not shadow them for an hour. Not send a survey. Not ask them in a meeting what their pain points were. But actually sit with them for a full day and observe:

  • Where do they get stuck?
  • What makes them sigh?
  • What workflows do they complain about?
  • What do they google repeatedly?
  • Where do they context-switch out of flow?

When I finally did this for my design system reboot, I discovered something shocking: designers weren’t struggling with “not enough components.” They were struggling with three specific, recurring patterns that appeared in every single project.

So we rebuilt. Twelve components. Solved those three patterns perfectly. All twelve were adopted within a month.

Portal vs. Pain Point

Your fourth mistake—spending six months building a portal—hits especially hard. I see this in design systems too. Teams build gorgeous documentation sites, elaborate component galleries, interactive demos… but the core workflow still sucks.

It’s procrastination disguised as productivity. Building the meta-layer (portals, docs, galleries) feels like progress because you’re shipping visible things. But you’re not solving the actual problem.

The hard question: If you’d spent those six months shadowing developers and solving their #1 workflow bottleneck instead of building a portal, where would you be now?

The UX Perspective on Developer Tools

Here’s something the platform engineering community doesn’t talk about enough: Developer experience IS user experience. The same UX principles apply:

User Research Before Building
You wouldn’t build a customer-facing product without user research. Why build an internal platform without it?

Usability Testing
Can a new developer deploy something in under 30 minutes using your platform? If not, why not?

Friction Mapping
Where are the moments of frustration in the current workflow? Those are your opportunities.

Progressive Disclosure
Start with the simplest possible solution. Add complexity only when users ask for it.

The difference is, in B2C products, we accept these as table stakes. In internal platforms, we skip them and wonder why adoption fails.

My Failed Startup Taught Me This

I ran a B2B SaaS startup that failed spectacularly. We spent nine months building before talking to customers. When we finally launched, the market yawned. We had built the wrong thing.

The lesson I took from that failure: Ship something that saves 30 minutes TODAY beats something comprehensive in 6 months.

Because if you can save someone 30 minutes today, they’ll use it. They’ll tell their teammates. They’ll ask for more. You’ll learn what they actually need. And six months from now, you’ll have built exactly the right thing based on real usage patterns.

But if you ship something comprehensive in six months, you’ve committed to a vision with no validation. And by the time you learn it’s wrong, you’ve lost momentum, trust, and probably some team members.

Questions for Your Platform Team

If you decide to pivot (and I hope you do), here are the questions I’d want answered:

  1. Who has watched developers work? Not shadowed. Watched. For full days.

  2. What’s the single workflow that causes the most visible frustration? Not the most technically interesting problem. The one where developers actually curse.

  3. Can you solve that workflow in 4-6 weeks? If not, scope it down until you can.

  4. How will you know if you’ve solved it? What’s the measurable outcome?

  5. Will developers use it without being told to? If you have to mandate it, you’ve failed the product test.

The Empathy Gap

The core issue I see in your story (and in my design system failure) is an empathy gap. We built for the abstract concept of “developers” rather than for the specific humans experiencing specific pain.

The fix isn’t more research. It’s sitting with one developer for one full day and experiencing their pain firsthand. You’ll learn more in that one day than in a month of stakeholder meetings.

One More Thing

You wrote that you wish someone had been honest with you 18 months ago. This community is here for exactly that reason—to have these hard, honest conversations.

But here’s the flip side: You’re 18 months ahead of someone else who’s about to make the same mistakes. By sharing this, you might have just saved another team from your exact failure path.

That has value, even if it hurts right now.

What’s the one developer workflow that, if you made it 10x better, would cause developers to voluntarily switch to your platform?

I’ve killed two platform initiatives and successfully launched one. The difference between success and failure wasn’t technical—it was brutal honesty about why we were building it and what success looked like.

Your post deserves a CTO perspective because these decisions ultimately land on our desks, and I want to share what I’ve learned from both sides of the pivot-or-wind-down decision.

The Three Platform Experiences

Failed Attempt #1 (Microsoft, 2015): Eight infrastructure engineers, no PM, no UX researcher. Built what they thought was technically impressive. 5% adoption after 12 months. Wound down.

Failed Attempt #2 (Startup, 2019): Six SREs, technical excellence, developer-hostile UX. Kubernetes expertise through the roof, but developers preferred Heroku even though it cost 3x more. Killed after 18 months.

Successful Launch (Current company, 2024): Four infrastructure engineers, two product managers, one part-time UX researcher. 80% voluntary adoption in 12 months. ROI positive within 6 months.

The difference? Not the technology. Not the budget. Not even the talent level.

The difference: We treated it as a product with customers, not as an infrastructure project with users.

The Strategic Mistakes That Kill Platforms

Looking at your situation, I see four strategic failures that go beyond the tactical mistakes you’ve already identified:

1. Problem Definition Was Wrong

You said you were building an “internal developer platform.” But why? What specific, measurable problem were you solving?

In my failed attempts, the answer was always vague: “Modernize our infrastructure.” “Improve developer productivity.” “Reduce toil.”

In my successful attempt, it was specific: “Reduce time from git push to production from 4 hours to under 30 minutes for our top 3 services, which account for 80% of deploys.”

See the difference? One is a vision. The other is a problem with a measurable outcome.

2. No Clear Success Criteria Upfront

I’m willing to bet that when you pitched this to leadership 18 months ago, the business case was something like:

  • Platform team: $2M/year
  • Expected benefits: “Improved developer productivity,” “Faster time to market,” “Reduced infrastructure costs”
  • Timeline: 12-18 months to “full platform”
  • Success metric: “Adoption” or “Feature completeness”

Am I close?

Here’s what the pitch should have been:

  • Platform team: $2M/year (6 people)
  • Specific problem: Deployments take 4 hours, blocking 40 developers × 4 hours/week = 160 hours/week wasted
  • Target: Reduce deployment time to <30 minutes within 8 weeks for ONE service
  • Success criteria:
    • Time saved: 150+ hours/week within 3 months
    • Voluntary adoption: 60% of teams within 6 months
    • Developer satisfaction: NPS >7 within 6 months
  • Cost-benefit: Save 150 hours × $100/hour = $15K/week = $780K/year. ROI positive in 15 months.

With specific success criteria, you can make a pivot-or-kill decision based on data, not sunk cost fallacy.

3. Team Composition Was Infrastructure-Only

“Staffed entirely with infrastructure engineers”—this is the single biggest predictor of platform failure I’ve seen.

It’s not that infrastructure engineers can’t build platforms. It’s that building a platform is 70% product management and 30% infrastructure.

You need:

  • Product managers who treat internal developers as customers
  • UX researchers who understand developer workflows
  • Technical writers who create usable documentation
  • DevEx engineers who think about user experience, not just technical architecture

Without product discipline, you get technically impressive things nobody wants to use.

4. Timeline Screamed “Big Bang”

18 months to ship anything useful. That timeline itself was a red flag.

Successful platforms ship value every 4-6 weeks. They start with one team, one workflow, one pain point. They prove value, gather feedback, iterate, expand.

An 18-month timeline says: “We’re building everything before we validate anything.” That’s not a platform strategy. That’s a bet. And based on the 60-70% failure rate, it’s a bad bet.

The Uncomfortable Truth

Most platform failures aren’t technical failures. They’re product failures disguised as engineering projects.

You had the budget. You had the talent. You had leadership support. What you didn’t have:

  • Someone asking: “What customer problem are we solving?”
  • Someone measuring: “How do we know when we’ve solved it?”
  • Someone with the courage to say: “This isn’t working. We need to pivot.”

That last one is what separates good leaders from stubborn ones.

Pivot or Wind Down: The Decision Framework

You asked for advice. Here’s the framework I use:

Vote PIVOT if:

:white_check_mark: Leadership is genuinely supportive (not just avoiding admitting failure)
:white_check_mark: Platform team isn’t completely burned out
:white_check_mark: You can identify ONE specific developer pain point with measurable impact
:white_check_mark: You’re willing to hire/develop product management capacity
:white_check_mark: You can commit to 6-week MVP with clear success/fail criteria
:white_check_mark: Developer trust isn’t irreparably broken

Vote WIND DOWN if:

:cross_mark: Leadership support is just sunk cost fallacy
:cross_mark: Team is demoralized and burned out
:cross_mark: Developers have lost trust (“they promised solutions before”)
:cross_mark: You can’t articulate one clear pain point to solve
:cross_mark: Organization isn’t ready to treat platform as product
:cross_mark: The $2M would create more value elsewhere

My Recommendation: Conditional Pivot

Based on your post, here’s what I’d do:

Recommendation: Pivot with a 6-week experiment and a kill switch.

The Pitch to Leadership:

“We’ve invested $2M and 18 months in a comprehensive platform that has 10% adoption. We can either wind it down now and redeploy the team, or we can try a focused 6-week pivot to solve one specific problem. If the pivot works, we’ll know by Week 8. If it doesn’t, we wind down with clear lessons learned. Either way, we make a decision based on data in 8 weeks, not 8 months.”

The 6-Week Pivot Plan:

  • Week 0: Hire product manager (ex-engineer with product experience)
  • Week 1: Developer interviews to find #1 pain point
  • Weeks 2-5: Build minimum solution to that ONE pain point
  • Week 6: Ship to 1-2 volunteer teams (not mandated)
  • Weeks 7-8: Measure:
    • Are they still using it without reminders?
    • Has it measurably reduced time/pain?
    • Would they recommend it to other teams? (NPS)

Success criteria:

  • Time saved: >50% reduction in that workflow
  • Voluntary usage: Both pilot teams use it without mandate
  • NPS: >7/10 from pilot teams

If success criteria met: Expand to 2 more teams. Repeat.
If criteria not met: Wind down, redeploy team, document lessons learned.

The Hardest Part of Your Job Right Now

As the leader who greenlit this, you now have the hardest conversation ahead: Do you pivot or wind down?

Both are valid options. Both require courage.

Pivoting requires:

  • Admitting the current approach failed
  • Convincing leadership to invest more in a new direction
  • Getting the team to believe in a different approach
  • Actually changing behavior (not just saying you’ll change)

Winding down requires:

  • Admitting the investment didn’t pan out
  • Having the difficult conversations with the team about what’s next
  • Extracting and sharing the lessons learned
  • Moving on without the sunk cost fallacy

Either way, the worst option is to keep doing what you’ve been doing and hope for different results.

Air Cover and Execution

If you choose to pivot, your job as the leader changes:

Protect the team from scope creep. Say NO to “while you’re at it…” requests.

Manage executive expectations. Frame this as a high-confidence experiment, not a guaranteed success.

Celebrate learning over delivery. If you learn this approach doesn’t work, that’s success too.

Shield from politics. Platform team needs psychological safety to try something radically different.

The team will execute the pivot. Your job is to give them the air cover to do it without interference, second-guessing, or scope creep.

The Question That Matters

You asked: “What would you do?”

I’d pivot. But with three conditions:

  1. Hire a product manager immediately. Not eventually. Now.
  2. Commit to 6-week experiment with clear kill criteria. No moving the goalposts.
  3. Get real about what ONE problem you’re solving. Not a platform. One workflow.

If you can’t meet those three conditions, wind it down. The half-measures won’t work.

Final Thought

You wrote: “I’m having hard conversations with leadership now about whether to pivot or wind down.”

That sentence tells me you’re already ahead of most leaders in this position. You’re asking the hard questions. You’re not doubling down on failure. You’re seeking input.

That’s leadership.

Regardless of whether you pivot or wind down, you’ve already done the hardest part: admitting the current approach isn’t working and being willing to change course.

What’s the ONE developer workflow that’s the most painful at your company today? If you can answer that in one sentence, you have a shot at a successful pivot. If you can’t, wind it down and redeploy the team to something with clearer value.

Either way, I’m rooting for you. Please report back in 6-8 weeks with what you learned.