PometryDocs
GitHub

Raphtory + Snowflake

Run graph intelligence on your Snowflake data warehouse.

Load transaction data, detect fraud rings, and write results back to Snowflake for BI dashboards - all in Python.

Why Integrate?

  • Leverage existing data: No ETL required, read directly from Snowflake.
  • Scale intelligence: Process billions of interactions.
  • Unified analytics: Combine SQL with graph intelligence.
  • BI-ready results: Write back to Snowflake for Tableau/Looker.

Setup

Example: Daily Fraud Detection

1. Connect to Snowflake

Initialize the connection using the Snowflake Python connector.

2. Load Transactions

Query directly from your warehouse into a Pandas DataFrame.

3. Build Temporal Graph

Ingest the data into Raphtory.

4. Detect Coordinated Behavior

Run community detection and temporal analysis to find suspicious rings.

5. Write Back Results

Export your findings back to a Snowflake table for downstream consumption.

Best Practices

  1. Incremental Loading: Query only new data since your last run to minimize warehouse costs.
  2. Result Tables: Write to dedicated alerts tables and use indexes for fast dashboarding.
  3. Warehouse Sizing: Use a MEDIUM or LARGE warehouse for highly parallel queries if processing large time windows.