Cycle time, cognitive load and the metrics that actually move delivery.
Measure friction, not sentiment alone
Developer experience is the time and effort required to make a safe change. Surveys reveal perception, while delivery data reveals where work waits. Both are necessary.
Track environment setup, build and test duration, review delay, deployment lead time, change failure and time to recovery. Segment results so averages do not hide one severely constrained team.
Follow the workflow
Map a change from idea to production and mark queues, repeated approvals, handoffs and unreliable tools. Interview developers around specific recent work rather than general satisfaction.
Separate necessary controls from accidental friction. Security and governance can remain strong while their implementation becomes clearer and more automated.
Improve and verify
Choose one bottleneck, define the expected effect and measure again after the change. Platform work should have product-style outcomes, owners and users.
Do not turn metrics into individual performance scores. That encourages gaming and weakens the trust needed to discover systemic problems.