Putting big data self-service analytics in the hands of business people
summary
Argues that a real‑time analytics stack built on Druid with streaming ingestion (Kafka) plus an intuitive UI (Imply Pivot) delivers subsecond queries on fresh and historical data so frontline marketers, product and ops teams can explore without analyst bottlenecks. Benchmarks claim a ~12x average speedup versus classic warehouses, 2.55x faster than Snowflake in one test and 7–24x more cost efficient in others. Companies deploy it for clickstream exploration, APM/metrics monitoring and live fraud detection (Ibotta, Dream11) with production success.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:07Hot analytics explained
- 3:14Temperature tiers of data
- 6:21Why hot analytics matters
- 8:26Use cases for hot analytics
- 13:59Limitations of data warehouses
- 16:27Building hot analytics capability
- 19:58Performance benchmarks
related talks
20mAnalyzing streaming data with Apache Druid
28mBest of Stream + Batch: When Streaming Analytics Isn’t
22mSituational Intelligence with Analytics in Motion
38mRealtime Data Activation - Analytics, Intelligence & Decision Making
28mThe Heat is On: Architecting for Hot Analytics
32mSome Like It Hot
27mStream, Store, Visualize: Fast-Track Real-Time Analytics the OSS Way
39mApache Druid 101