Inventory History as Pure Functions
summary
Stitch Fix builds Tracer to answer exact inventory state at any timestamp by storing delta transitions (per-item quantity changes) plus periodic hourly snapshots, avoiding expensive full-table reads and subsecond polling. Implementation runs on Spark with a Scala core and Python end-user APIs, indexes only 24 snapshots per day, handles tens of thousands of SKUs, and reconstructs state by applying deltas to the nearest snapshot so queries are fast and production DB load stays low.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:55Overview
- 3:08Why Inventory
- 4:32Recommendation
- 6:48Design
- 11:39Extreme
- 13:13Fixed Interval
- 14:36T1 T2
- 17:42Summary
- 18:14Implementation
related talks
30mBuilding a metadata ecosystem using the Hive Metastore
07mAdvanced analytics and the mystery of the missing jeans
35mBuilding a successful inventory planning solution with Amazon Forecast
24mRetail Transactions at scale for AI - From PoC to production
40mProductionizing Big Data - Stories from the Trenches
50mOur experience at Etsy com with building “Big Data” analytic capabilitie
34mApache Iceberg.
23mData Fabric