Give an example of a time you failed and what you learned from it.

Situation: I pushed a pipeline change (a new join condition) that wasn't covered by integration tests. It introduced a Cartesian product in an edge case, causing incorrect aggregations and a 6-hour data outage affecting 15 downstream reports. Task: Restore service, communicate transparently, and prevent recurrence. Action: I rolled back immediately and notified stakeholders with an ETA. I ran a blameless post-mortem: root cause was insufficient test coverage for join logic. I implemented (1) integration tests for all join paths. (2) A staging-to-prod comparison job that runs before release. (3) A checklist for pipeline changes (test coverage, runbook update). I shared the post-mortem and learnings with the broader team. Result: We've had zero similar outages in 2 years. The testing framework is now standard for all pipeline changes. I learned to never skip integration tests, even under time pressure.