Tracing a research insight to shipped code

WHAT THIS TESTS This tests traceability system design plus intellectual honesty about attribution. The interviewer wants stable identifiers flowing through real toolchains and a recognition that insight-to-outcome links are many-to-many, not a clean causal chain.

A GOOD ANSWER COVERS Give every research insight a stable, unique ID in a research repository, with metadata: study, date, theme, severity. Propagate that ID downstream through integrations. When an insight spawns work, the backlog ticket references the insight ID through an issue-tracker integration. Developers include the ticket or insight ID in commit messages and pull requests, so a version-control integration links code to the insight. If the work ships behind a feature flag, tag the flag with the insight ID via the flag platform's API. After launch, connect to analytics to capture the outcome metric tied to that insight's hypothesis. The result is a queryable graph from study to insight to ticket to commit to flag to metric. Crucially, model relationships as many-to-many: one insight influences several features, and one feature reflects several insights, so report influence and correlation, not strict causation.

COMMON WRONG ANSWERS Tracking everything in a manual spreadsheet that goes stale. No stable insight identifiers, so links break. Claiming a clean one-to-one causal mapping from insight to commit. Stopping at the commit and never measuring the outcome. Ignoring that engineers must actually reference the ID for the chain to hold.

LIKELY FOLLOW-UPS How do you handle one insight influencing many features. How do you measure outcome impact without overclaiming causation. What if developers forget to reference the insight ID.

ONE CONCRETE EXAMPLE Insight INS-204, 'users miss the save action', gets an ID in the research repo. A Jira ticket cites INS-204; the developer writes 'Fixes PROJ-88 (INS-204)' in the commit, and the PR links back. The feature ships behind flag save-button-redesign tagged INS-204. Post-launch, analytics tied to that flag show a rise in save usage. Leadership queries the graph and sees INS-204 traced from study to a measurable lift, while the system also shows INS-204 contributed to two other tickets, so impact is reported as influence across several changes rather than a single causal line.