Skip to main content

Dashboards

You'll learn how to create and share Databricks AI/BI dashboards in ~10 min.

Prereqs: Query and explore

Why this matters​

A dashboard is how most people in your org actually see the data. AI/BI dashboards query Unity Catalog tables directly, so there are no extracts to refresh and no stale copies drifting out of sync with the source. The other win is governance: row- and column-level security from Unity Catalog carries straight through, so a viewer sees only the rows they are allowed to see. You don't rebuild access rules in the BI layer.

How it works​

You assemble a dashboard from charts, tables, and counters, each backed by a SQL query. The editor is drag-and-drop. Add filters and parameters, then publish to share. You can schedule a dashboard to run on its own so viewers land on fresh cached results instead of waiting on a query.

Create the first dashboard​

This video builds a dashboard from scratch:

Embed dashboards in external applications​

You can embed a dashboard in an internal portal, a wiki, or a custom app with an iframe, which puts it in front of people who never open the Databricks workspace.

When to use / when not to​

Use dashboards when:

  • Stakeholders need a visual, regularly refreshed view of the same key metrics.
  • You want Unity Catalog row and column security to apply at the presentation layer for free.
  • The audience expects a standard reporting experience: charts, tables, filters.

Use Genie Spaces instead when:

  • Users want to ask their own ad-hoc questions, not browse a fixed set of charts.
  • The questions change often enough that a static dashboard would need constant rebuilding.

Common pitfalls​

  • Too many queries on one dashboard. Every visualization runs its own query, so a board with 20-plus charts crawls on load. Group related metrics and split a big board into focused pages.
  • No scheduled refresh. By default a dashboard queries live data on every page load, which queues warehouse compute each visit. For stable reporting, schedule a periodic refresh so viewers hit cached results instead.

Next​