Introduction to Sequence Analytics
summary
Argues that event logs should be a native data model rather than pre-aggregated tables, and demonstrates concrete tooling: state machines to classify users into interpretable states, value‑tuned transformers to embed sequences and predict long‑term outcomes, and a custom Motif Analytics query engine because SQL can’t express the needed iterative sequence queries. Shows product examples (Lyft conversion funnels, Duolingo engagement) and an NFL play that adds ~20% win‑prob, notes prototypes and visual explorer work but warns inference latency limits true real‑time use today.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:25Event sequence data
- 8:21Event studies and state machines
- 15:45User journeys and transitions
- 22:43Data representation vs insights
- 28:18Deep learning for sequences
- 33:51Sequence analytics challenges
- 34:14Q&A
related talks
17mCreating & operating ML models from event-based data using feature stor...
31mAccelerating Your Analytics Journey With Real-time AI
16mDetecting Anomalies in Logs with Time Series Analysis
16mOperationalizing AI in Finance: Real-Time Analytics
43mBeyond Logs: Real-Time Anomaly Detection with Machine Learning
58mAI-Enhanced Big Data: Integrating Private LLMs and Vector Databases
18mBehavioral Data Creation for AI - Delivering Value and Insights
45mAdaptive Data Collection via Autoregressive Generation