Latency spikes. Errors climb. Where do you even start?

When latency or errors spike, the developer knows users are affected. The difficult questions remain: which requests should they inspect, what makes those requests different, and where should the investigation go next?

Millions of traces can hold the clues without making them easy to find. Searching for relevant evidence and interpreting it adds work while customers are still experiencing the problem.

I owned the long-term Distributed Tracing strategy around two connected needs: make more services visible, and make the information useful during an investigation. Entity Coverage Ratio became a measure of that reach, with a roadmap target of 34% to 80% by FY28.

The coverage figure is a future target. The adoption results below reflect work already delivered.

Built features like Intelligent Exemplar to solve the needle-in-a-haystack problem: surface representative traces while preserving the context of the investigation.

Owned prototyping, customer validation, evaluation and launch work, making the proposed experience concrete enough to test with the people using it.

Attribute Correlation Analysis compares anomalous traces with a healthy baseline to identify attributes associated with errors or slow responses. AI-generated explanations help developers interpret those patterns and choose the next step.

I vibe-coded an ACA prototype, gathered customer feedback, and owned prompt design, output evaluation and iterative improvements. The work connected hands-on experimentation to a customer-facing investigation experience.

Improvements across the investigation experience, including Intelligent Exemplar and ACA, contributed to 16% growth in monthly active users from August 2024 to June 2026.

The contribution was a more useful path through the data: a place to start, patterns to examine, and explanations grounded in evidence.

Public documentation labels both capabilities as preview. The MAU result covers the broader investigation experience.