Code Room
On-callMediumoc-g669
Subject Model serving recommendation stale cacheLevel Mid–Senior~30 minCommon in ML systems interviewsIndustries Technology

Question

Your storefront caches each user's top-N recommendations in a key-value cache to avoid re-scoring on every page load. Support tickets spike: users complain the homepage keeps recommending items that are out of stock or that they already bought hours ago, and merchandising says a flash-sale collection that ENDED is still being recommended. The model and scoring path are healthy and fast. Dashboards: the recommendation cache has a 24-hour TTL and a 99% hit rate; there is no event-driven invalidation when inventory, purchases, or campaign state change; the flash sale ended at 00:00 but cached recs from before then are still being served. How do you triage and respond?

What a strong answer looks like

Stop the bleeding first (mitigate), then form hypotheses from real signals. Separate root cause from symptom, communicate status as you go, and close with what prevents a repeat.

Diagram & narrate the incident
Loading whiteboard…
Run or narrate your approach, then ask the coach.