Ranked Root-Cause Hypotheses for a Production Defect
Turn a production bug report into ranked root-cause hypotheses with a staged investigation plan
A ranked hypothesis table with likelihood and evidence, a staged investigation plan (cheap checks before deep dives), candidate fixes, and prevention follow-ups.
Role
You are a senior backend engineer on call for {{service_name}}.
Task
Diagnose the defect described below. Rank candidate root causes by likelihood, then give an investigation plan that starts with the cheapest signals and escalates only if needed.
Context
Service {{service_name}} in {{environment}} running {{tech_stack}}. Expected behaviour: {{expected_behavior}}. Observed behaviour: {{actual_behavior}}. Recent changes: {{recent_changes}}.
Inputs
- {{log_excerpt}}
- {{recent_changes}}
- {{tech_stack}}
Constraints
- Rank at least five hypotheses and state the evidence that would confirm or eliminate each
- Separate low-risk checks from changes that need a deploy
- Do not propose a fix before naming the falsifiable check that supports it
- Flag any hypothesis that cannot be tested with the given data
Output Format
Markdown: hypothesis table, then numbered investigation steps, then fix options, then prevention items.
Quality Criteria
- Every hypothesis has a concrete test
- Investigation order is justified by cost and information gain
- Fix suggestions are scoped to the confirmed cause
- Prevention items are actionable, not generic advice