AI
Why Platform Teams Build for the Wrong People
Most engineers on a bell curve sit in the middle. They are not Kubernetes experts, not Terraform power users. They are the 80% who need the platform most. Platform engineers and the vocal experts sit at the edges. Because those edge users are the loudest, platform teams get pulled toward them. The average engineer falls behind. Reichenbach calls this the inverse productivity curve: the platform becomes expert-grade and the middle 80% are left with workarounds.
Exploratory Interviews: Finding the Unknown Unknowns
Exploratory interviews go broad across three to five people per round, mixing one or two experts with three silent average users. The user shares their screen and walks through a full day of work, from spec to deploy. The interviewer’s only job is to listen. Reichenbach puts the ratio at 95% listening and 5% talking. Correcting how someone uses the platform kills the signal. Run these every 6 to 12 months to keep blind spots visible. They surface problems the platform team had no idea existed.
Surveys That Actually Get Filled Out
A survey posted to a dead Slack channel gets ignored. At 1KOMMA5°, surveys are built from hypotheses collected during user interviews. Each question tests something specific, not just fishing for opinions. Making surveys non-anonymous lets the team send personalized follow-up messages via Cursor or Claude. That tactic pushes fill rates to 70 to 80%. For time-sensitive data, Reichenbach’s team ran surveys inside meetings and hit 90 to 100%. Their Terraform refactor survey confirmed users modified stacks once a month or less, which let the team move on and build something else.
Measuring Success Before Writing Code
The standard build-measure-learn order hides a problem: you only find out if you built the right thing after building it. Reichenbach’s colleague Klaus flipped the order to measure-learn-build. Before touching the Terraform scaffolding, the team wrote down three metrics: zero manual steps, zero failures, under 15 minutes end-to-end. The baseline was nine manual steps, seven failures, and 1,248 minutes. Targeted interviews during each iteration brought users in to rate clarity and ease from one to 10. The standing question after every step: “What would make this a 10 out of 10 for you?”
Products Evolve, Projects End
A project ships and the team moves on. A product never stops receiving signal. Reichenbach frames platform teams the same way startups frame product-market fit: if you are not solving the right problems, no one uses your solutions. The scaffolding tool today lacks its full feature set, and that is fine. When a user asks for a new scaffold type next quarter, the team adds it to the product backlog. Change lead time is the team’s north star metric, but Dora metrics alone are too lagging and too broad to guide individual feature decisions. Specific, hypothesis-driven metrics fill the gap.
Q&A
How do you keep expert users happy while building for the middle? Allow experts to bypass abstractions entirely, for example letting them write raw Terraform while average users get a YAML interface. ▶ 23:52
How do you convince skeptical platform engineers to invest time in user interviews? Start with one interview and let them discover their own blind spots firsthand. ▶ 25:41
How do you handle low-adoption libraries used by roughly 50 developers? First understand why they have not migrated, then help automate the migration rather than simply cutting support. ▶ 28:00
How do you reach the quiet 80% who do not want to be interviewed? Start over coffee, pull them in gradually, and create an alpha-tester Slack channel so willing users self-select for ongoing feedback. ▶ 33:46
Notable Quotes
It’s 95% listening and 5% talking. And your role is not to explain it to them. Michael Reichenbach · ▶ 07:44
“What would make this a 10 out of 10 for you?” And you will be surprised what answers you get there Michael Reichenbach · ▶ 17:54
ship products, not projects, and evolve your platform. Michael Reichenbach · ▶ 22:10
we started to build for the middle, for the 80%, instead of just the experts. Michael Reichenbach · ▶ 22:33
Key Takeaways
- Experts are loud but represent a fraction; the silent middle 80% need the platform most.
- Exploratory interviews every 6 to 12 months surface blind spots no survey would reveal.
- Define success metrics before building: zero manual steps beats vague quality goals.
About the Speaker(s)
Michael Reichenbach is a Senior Platform Engineer at 1KOMMA5°. He builds the developer platform behind Europe’s largest virtual power plant, supporting software that manages over 50,000 decentralized energy systems.