When UFO Lands! Analyzing User Feedback During a Massive Customer Migration
summary
A lightweight, LLM‑driven pipeline ingests in‑product feedback and uninstall surveys into BigQuery, labels and translates comments with the Gemini API, and surfaces a daily User Feedback Overview plus interactive dashboards. It scales from a 20k canary to tens of millions (30 million migrated users), processes about 6,000 comments per day for roughly $6, and exposes concrete issues (firewall/printer failures), reducing blind spots and speeding troubleshooting.
AI-written summary · about
chapters (7)
- 0:00UFO and user feedback
- 3:20Migration challenges
- 8:24Manual labeling to AI
- 13:55AI feedback system
- 21:27Replicating the solution
- 24:42Quick start guide
- 25:06Q&A
related talks
30mMulti-tenant Conversational Analytics
29mCrash-reporting taken to next-level
30mReliable Data Processing with Minimal Toil
14mAI-accelerated migrations at scale
38mOptimizing UX and Performance
37mHow Small Changes Can Make Huge Waves
24mHow to create and scale AI Products, MLOPs and Generative AI
31mCloud Migration Without Fear - A...