Reflex
Real-Time Competitive Demand-Shock Detection & Response Platform
A distributed data pipeline that detects unplanned e-commerce demand shocks in real time — validated against two real-world events I didn't know about in advance.
The Problem
E-commerce companies plan well for predictable demand spikes like Black Friday, but struggle with unplanned shifts — a competitor's flash sale, a surprise price drop, a product suddenly trending. By the time internal teams notice, the critical early window to respond has already passed. I've seen this firsthand working in e-commerce, and this project is my attempt to actually solve it.
What it Does
Trend — Builds a rolling baseline of expected demand per category
Shock Detection — Flags statistically significant deviations as they happen
Propagation — Tests whether spillover demand follows to related categories
Response — Combines it all into one ranked, actionable output
Key Findings
8 demand shocks detected across 61 days of real transaction data
Real-world validation: the largest anomaly aligns exactly with the Huawei Mate X foldable smartphone launch (Nov 15, 2019); a second shock captures Black Friday
Consistent leading indicator: every shock was driven by smartphones, concentrated among Samsung, Apple, Xiaomi, and Huawei
Honest, nuanced result: spillover to related categories is real but uneven — headphones showed a clean signal, cameras showed none
