Snowflake Cortex Analyst

Snowflake Cortex Analyst Semantic Views SQL NLP

A natural language analytics layer over a real estate portfolio data model — plain English questions resolved to auditable, SQL-backed answers without writing a single query.

Overview

Business stakeholders typically wait days for ad-hoc data questions to be answered. The bottleneck isn't the data — it's the translation layer between a business question and a SQL query. This POC removes that bottleneck: a real estate star schema with a Semantic View on top, so Cortex Analyst can answer questions about agent rankings, market trends, and budget variances with no SQL knowledge from the user. Every answer includes the SQL that generated it, so technical stakeholders can verify it.

💬

Natural Language → SQL

Non-technical users ask questions in plain English. Cortex Analyst generates correct SQL against your actual schema with no prompt engineering needed.

🧠

Semantic Views

A SQL-native semantic layer describing tables, relationships, dimensions, and metrics so Cortex understands your business context precisely.

🔍

Fully Auditable

Every answer includes the SQL that generated it. Stakeholders can inspect, validate, and trust the output rather than treating it as a black box.

🏗️

Zero Extra Infrastructure

Runs entirely inside Snowflake. No external endpoints, no data movement, governed by your existing Snowflake roles and permissions.

Sample Questions

"Which agents generated the most commission this year?" → groups by agent_name, aggregates total_commission_revenue, ranks descending

"What is the average days on market by property type?" → groups by property_type dimension, returns avg_days_on_market metric

"Which markets have the highest price-to-list ratio?" → joins property market, aggregates avg_price_to_list_ratio per market

"What expense categories are over budget?" → filters on negative budget_variance in operating expenses fact table

Roadmap

✓ Done

Real estate POC — data model, semantic view, Cortex Analyst

Full star schema, sample data generation, semantic view, and Cortex Analyst tested with business questions.

In Progress

Snowflake Intelligence chat interface

Wrapping Cortex Analyst in Snowflake Intelligence for a conversational UI accessible to non-technical stakeholders.

Planned

Cortex Search + RAG pattern

Combining Cortex Search (vector search on documents) with Cortex Analyst in a single Snowflake Intelligence agent.

Planned

Streamlit in Snowflake dashboard

Interactive charts hosted inside Snowflake built on the same semantic view — no external BI tool required.