Nordstrom Conversation Analytics Data Product (POC)

This repo is a compact, production-style analytics project focused on conversation analytics with a governed BI layer.

It demonstrates:

• warehouse modeling (staging → marts) with production SQL
• a Looker-aligned semantic layer (LookML-style metric definitions)
• conversation enrichment (LLM-style classification into structured fields)
• decision-ready outputs (executive brief + dashboard mockups)
• reproducible local runs (generates synthetic data and builds modeled tables)

Quickstart

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

python scripts/run_pipeline.py

Outputs:

• warehouse/out/*.csv (modeled tables)
• reports/metrics_snapshot.json
• reports/dashboard_mockups.pdf
• reports/executive_brief.pdf

Notes

• Default enrichment is rule-based so it runs offline.
• An optional LLM hook is documented in llm_enrichment/enrich_conversations.py.