The Test Set by Posit Your VP Is Doing a Rogue Analysis in Cursor Right Now — with Nell Thomas
Mar 23, 2026
Nell Thomas, VP of Data at Shopify who has led data teams at Etsy, Facebook, and the DNC, maps the modern data stack and the roles that make it work. She talks data quality, building trust with scrutiny and blameless postmortems. She describes scaling a 400-person org, why semantic layers are so hard, and how AI is changing experiments and metric investigations.
AI Snips
Chapters
Transcript
Episode notes
Build Trust By Going Far Left On Data
- Investing far-left in data (collection, ingestion, pipeline) multiplies leverage across analytics and production ML.
- Shopify focuses on correct capture, cost efficiency, privacy, and an MCP for the warehouse to enable self-service.
Get An End To End Data Map Immediately
- Map an end-to-end data flow and ownership as your first onboarding step at a company.
- Trace events from source to dashboard to reveal hops, latency, and hidden owners of assumptions.
Analytics And Production Diverge Later Not Sooner
- Analytics use cases (batch, human-facing) and production use cases (low-latency, machine-facing) fork requirements but should share upstream context.
- Forking too early causes duplication, cost, and lost knowledge.