IntermediatePROBLEM_SOLVING
During the Brand Relaunch Campaign (Project) you drove large increases in brand awareness; describe a time you diagnosed a tracking or data-consistency bug in campaign metrics, how you investigated the issue (logs, instrumentation, test cases), and what code-level or infrastructure changes you proposed or made.
Software Engineer
General

Sample Answer

Mid-campaign our week-over-week CTR jumped 40% overnight — but revenue didn't change. I dug into the pipeline and found duplicate impression events inflating denominators. I reviewed raw logs in S3, compared event_id hashes, and wrote a small Spark job to surface duplicates by (user_hash, creative_id, minute). Reproducing the issue in staging showed a race where client retries and a flaky CDN callback created double-posts. I worked with engineers to add idempotency keys on the client and server, moved callback acknowledgement to the edge, and added a lightweight Kafka de-duplication step using message keys. Within 48 hours duplicates dropped from 12% to 0.3% and our KPI dashboards realigned with revenue within the next reporting cycle.

Keywords

Used logs and a Spark job to quantify duplicates (12% → 0.3%)Identified race condition between client retries and CDN callbacksImplemented idempotency keys, edge ack, and Kafka de-duplicationRestored metric integrity within 48 hours and corrected dashboards