Skip to main content

CDP BI-tool connectors

The BI-tool connector catalog is the answer to “how do our analysts pull Orbit data into the BI tool they already use?” Reverse-ETL destinations move CDP data out to martech; the warehouse export loads contacts, events, audiences, and profile traits into a warehouse you own. This catalog is the other half: it tells a BI tool how to read those tables, without CSV exports or hand-written SQL.

1. Pull into BI vs push via reverse ETL

Orbit’s export families split along two directions:
  • Reverse-ETL destinations (push) — Orbit writes your CDP entities into a destination (a warehouse, a martech endpoint, a webhook). Wire these when you fan data outward on a schedule.
  • BI-tool connectors (pull) — your BI tool connects natively to the warehouse the reverse-ETL export already loads (BigQuery, Snowflake, Redshift, Postgres, or Databricks) and reads the curated datasets from there. No new pipeline, no new export — the warehouse export hub is the substrate, and the connector catalog is metadata over it.
Because the BI tool connects to your warehouse with your credentials, Orbit never sits in the query path and never proxies warehouse access. The catalog supplies the connection profile; the credentials stay tenant-owned. For the warehouse export itself, start with the reverse-ETL warehouse exports guide or the CDP reverse ETL operator walkthrough.

2. Catalog listing

List the supported connectors:
Filter the list with three optional query parameters:
  • ?status= — ga, beta, or coming_soon.
  • ?warehouse= — restrict to tools that connect to one warehouse (bigquery, snowflake, redshift, postgres, databricks).
  • ?search= — a case-insensitive substring over the tool name, vendor, or description (≤100 characters).
The response returns the filtered connectors plus three discovery blocks:
  • statuses — per-status facet counts over the full catalog (the filter sidebar).
  • warehouses — the live warehouse list with each warehouse’s driver name.
  • datasets — the curated analytics datasets (contacts, events, audiences, and campaigns) with their labels.
A bad status or warehouse value returns a 400 rather than silently an empty list.

3. Per-tool metadata

Fetch one tool’s full catalog entry before you wire it:
The response carries the vendor, the supported warehouses, the connector status, and the tool’s own documentation URL. An unknown tool_id returns 404.

4. Connection-profile generation

Generate the exact connection shape an analyst follows to point the tool at your warehouse — driver, connection fields, schema-qualified datasets, and ordered setup steps:
Request body:
  • warehouse (required) — a live Orbit warehouse: bigquery, snowflake, redshift, postgres, or databricks.
  • schema (optional) — the schema the export writes to; defaults to orbit_cdp.
  • dataset_keys (optional) — restrict the profile to a subset of the curated datasets (≤20 keys); omit to include all four.
Three properties carry the contract:
  • datasets are schema-qualified. Each entry names the fully qualified table (orbit_cdp.orbit_profiles for contacts, orbit_cdp.orbit_events for events, orbit_cdp.orbit_audiences for audiences). Campaign engagement reads from the events table filtered on the campaign_id property — it is a view over the event stream, not a separately loaded table, so the catalog describes it as such.
  • No secrets, ever. connectionFields are labelled placeholders (account, host, database, schema — whatever the warehouse’s driver asks for), and the setup steps remind the analyst to authenticate with their own credentials. Orbit never proxies warehouse access, so the profile is safe to paste into a ticket or a runbook.
  • Honest maturity. Every response echoes tool_id, status, and connection_available. A coming_soon connector still returns a profile for discovery, but the dashboard renders a request-access action rather than a connect button.
Error shape: 404 for an unknown tool; 400 for a body parse failure; 400 field error when the tool does not natively connect to the chosen warehouse, the schema name is not a valid SQL identifier, or a requested dataset key is unknown. The generation route is rate-limited tighter than the reads (30/min vs 60/min) because it builds a payload rather than serving a directory.

5. Auth and roles

All three routes require either a Clerk session or an API key, and a role of owner, admin, or developer — the same gate as the sibling destination catalog. The payload is static product metadata (no tenant or customer data, no secrets), so no additional scope is required. Reads run at 60 requests per minute per tenant; the profile-generation POST runs at 30 per minute.

6. Supported tools

Full-detail metadata plus connection-profile generation are live for: The catalog also carries Sigma, Mode, and Amazon QuickSight — QuickSight pairs with Redshift natively — and a tool enters ga only when it natively connects to at least one warehouse Orbit’s export actually loads, so a GA badge always maps to a real, serviceable connection.