Our growth team needed a new events table to track a funnel they were launching in two weeks. The PM didn't have a full spec ready, so I sat down with her, got enough context to start, and built out the pipeline. I kept the schema simple — one row per event, user ID, event type, timestamp — and got it into production within three days. The team was able to run their analysis on launch day. It went smoothly and they were happy with the data.
The growth PM needed an events table for a new funnel, launching in ten days, with no spec beyond 'we need to track drop-off.' The real ambiguity was cardinality: events could have one to many properties depending on type. I made a deliberate call — a typed event schema with a JSON properties column for extensible attributes rather than a wide table — so the DS team could query structured fields fast while the PM could add new event types without a schema migration. Pipeline was in production in four days. At launch, the DS team ran same-day funnel analysis and identified a thirty-percent drop-off at step two, which the PM used to reprioritize the sprint.