Platform Telemetry
Instrumentation of the platform itself — who uses which capability, how long it takes and where it fails — used as product evidence, not just operations.
Platform teams routinely instrument the workloads they host and neglect to instrument the platform, which leaves them managing a product with no usage data and arguing about priorities from anecdote.
The questions the telemetry must answer are product questions. Which capabilities are used, by whom and how often. Where developers abandon a flow part-way, which is where the friction is. How long each self-service operation takes at p50 and p95, since the tail is what people remember. Which services are on deprecated versions and who owns them. What proportion of teams are on the golden path versus off it.
This data is what turns a deprecation from an announcement into a plan, what settles roadmap arguments with evidence, and what demonstrates value to the funders — the platform's case is strongest when it can show consuming teams' lead time and change failure rate improving after adoption.
The one caution: usage telemetry across internal teams is still data about people's work, and publishing per-team comparisons tends to produce gaming and defensiveness. Aggregate for reporting, segment privately for improvement.