Most internal developer platforms never reach their intended audience. Teams burn months configuring portals like Backstage, only to watch adoption flatline. According to Max Körbächer, CNCF Ambassador and founder of Liquid Reply, the root cause is not technical. It is cultural.
Körbächer, who has spent the last five to six years in the open source community and co-authored Platform Engineering for Architects, argues that platform engineering teams routinely treat their work as a project with a deadline rather than a product with users. The result is infrastructure that looks impressive on a slide deck but solves no one’s actual problems.
The Trap of Infrastructure-First Thinking
Körbächer identifies what he calls "infrastructure-first thinking" as the single biggest killer of platform adoption. Most platform engineers come from cloud or infrastructure backgrounds. Their instinct is to build something technically elegant, then hope users show up. It rarely works.
He draws a parallel to his own early career in demand management, where he was forced to interview end users and challenge their requirements. "It was a painful process because it is not my nature to go out and ask people, what do you really want?" he admitted during a recent QCon presentation. That discomfort, he says, is exactly what platform teams must embrace.
The consequences of skipping this step are predictable. Companies install Backstage because Spotify built it, then discover it is an empty shell until five engineers spend months on extensions. Or they return from KubeCon excited about a new tool and drop it into the platform without asking whether anyone needs it. Körbächer calls this "randomeering," a blend of random and engineering that produces bloated, unused infrastructure.
Why "Shift Left" Often Means "Shit Left"
The push to shift responsibility onto developers has compounded the problem. Körbächer quotes another conference speaker who rebranded the concept as "shit left": pushing CI/CD, observability, tracing, security, and compliance down the pipeline to individual teams without giving them the tools to handle it.
In large enterprises, this scales badly. Different teams build their own patch solutions. Reorganizations wipe out harmonization efforts. Industry hypes, from IoT to AI, break whatever fragile alignment existed. "You are wasting more money than you actually bring success to a company," Körbächer said, pointing to studies showing AI and LLM initiatives are failing more often than they succeed.
A Product Mindset, Not a Project Mindset
The fix, he argues, starts with treating the platform as a product. That means understanding the customer, who may not be a developer at all. One of the most successful platforms his consultancy built has no fancy frontend. Almost half of all VPs check a GitHub report on platform usage weekly, delivered straight to their inbox. "A platform without adoption is just infrastructure, and just infrastructure is usually very unsexy," he noted.
To get there, Körbächer recommends rethinking the classic iron triangle of feasibility, desirability, and viability. Most teams can optimize two of the three at the expense of the third. The only way to balance all three is to reach a certain scale threshold, which takes time and deliberate investment.
Measuring What Actually Matters
Körbächer is skeptical of DORA metrics as a management weapon. When leaders use deployment frequency or change failure rate to pressure teams into shipping faster, quality drops and stress rises. He prefers SPACE metrics (satisfaction, performance, activity, communication, efficiency) and DevEx metrics focused on flow state and cognitive load.
The latter is critical. Research shows that every five minutes of interruption can cost an hour of mental processing time. A platform that forces context switching, or one that adds meetings and reporting overhead, is actively harming productivity. The only way to know is to ask users directly, repeatedly, and in writing. "Numbers can lie," Körbächer warned. "Two numbers which look correlated do not need to be correlated."
Technical Debt and the Sunk Cost Trap
Even well-intentioned platforms accumulate technical debt fast, especially on Kubernetes. Körbächer blames the sunk cost fallacy, reinforced by accounting practices that pressure teams to keep running obsolete tools until tax write-offs are complete. His advice: define depreciation criteria from day one. If a tool is hated by the overwhelming majority of users, cut it, even if you personally love it.
The Road Ahead
Körbächer closes with a simple compass for platform teams: clear purpose, continuous user research, technical debt management, and community building. The most successful platforms he has seen invite contributions from across the organization, sometimes with small rewards, always with genuine openness. "The success of platforms is not in its code," he said. "The technology does not matter so much. You will fail if you just focus on the tech and ignore people, culture, and process."
The next frontier for platform engineering may be less about tooling and more about organizational change management. Teams that can bridge that gap will define the standard for the next decade.