A team added a Redis caching layer to speed up a slow API endpoint. Response times got worse. The cache was working perfectly — it was caching the wrong thing.
The largest enterprise customer was in renewal negotiations. Dashboard load time was their top complaint. Three weeks of engineering effort made the problem worse instead of better. The pressure to show progress was creating more regressions than improvements.
You're a senior engineer at a SaaS company. The dashboard API endpoint takes 4 seconds under load. The product team is escalating. Your tech lead proposes adding a Redis caching layer to cache the database query results. It'll take about a week to implement. What's your response?
No hints. Just judgment.
Caching the full API response eliminates all backend work and delivers fast responses. But dashboard data has freshness expectations — enterprise customers expect near-real-time metrics. Any TTL that meaningfully reduces backend load also introduces staleness that customers notice. You're trading latency for correctness, and for dashboards used for operational decisions, that's a trade most customers won't accept.