Skip to main content
Beton
Postgres
Data Source

Beton + Postgres

Behavioral signal discovery on top of your raw warehouse

Connect any Postgres warehouse as a behavioral data source. Replicas, Supabase, self-hosted — Beton reads schema and discovers signals directly.

How It Works

1

Connect a read-only role

Provide a connection string. Beton uses a single read-only role and never writes back. Your Postgres takes no extra load — Beton schedules its own scans on a quiet replica or off-peak window.

2

Beton proposes hypotheses

The agent reads schema (tables, columns, foreign keys) and proposes testable behavioral hypotheses tailored to YOUR table layout — no generic templates.

3

Backtest before promotion

Each hypothesis is validated against historical rows: precision, recall, lift over baseline. Promote only the ones that beat your bar.

Setup

Provision a read-only role for Beton (run in psql)

bash
CREATE ROLE beton_readonly LOGIN PASSWORD '<random>';
GRANT CONNECT ON DATABASE your_db TO beton_readonly;
GRANT USAGE ON SCHEMA public TO beton_readonly;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO beton_readonly;
ALTER DEFAULT PRIVILEGES IN SCHEMA public
  GRANT SELECT ON TABLES TO beton_readonly;

Features

  • Read-only connection — zero impact on production
  • Works with managed Postgres (RDS, Cloud SQL), Supabase, and self-hosted
  • Automatic schema introspection — no manual table mapping
  • Compatible with warehouse loaders like Fivetran, Segment, RudderStack
  • Same hypothesis-and-backtest flow as PostHog
  • No data export or migration required

Use Cases

  • Run signal discovery on a production replica
  • Use Beton on warehouse data piped from Segment / Fivetran / RudderStack
  • Cross-reference event data with internal product tables (subscriptions, deals, support tickets)
  • Run signal hunts on Supabase-backed apps

Ready to connect Postgres?

Start detecting revenue signals and routing them to Postgres in minutes.