Free · Lesson 1 of 9
Metabase turns a database you already trust into questions and dashboards people can open without writing SQL.
Sign in to save your place.
Metabase is the shop window. The database is the stockroom. People ask questions at the window. They do not walk into the stockroom and rearrange the shelves. A question is one saved request. A dashboard is a wall of those requests, with knobs that change the request for everyone looking at that wall.
Metabase is a business-intelligence application. It connects to a database you already have, such as PostgreSQL, MySQL, SQL Server, BigQuery, Snowflake, or Redshift, and lets people explore that data. It does not replace the warehouse. Your tables, backups, and query engine stay where they are.
You can run Metabase Cloud, which Metabase hosts, or you can run the same application yourself with Docker or the JAR file. The open-source application is free to run. Paid plans add sign-in with your company identity provider, row-level sandboxes, a usage audit, and deeper embedding. This first lesson is free. The rest of the course is Pro.
| Piece | What it is | Where it lives |
|---|---|---|
| Question | A query plus a chart | A collection |
| Model | A curated question people should start from | A collection |
| Dashboard | Several questions on one page, with filters | A collection |
| Collection | A folder with permissions | Metabase, not the warehouse |
| Application database | Stores questions, dashboards, and users | Metabase's own database |
The useful habit is to name the grain before you chart it. "Orders" can mean one row per order, one row per line, or one row per day. The chart cannot fix a grain you have not chosen.
Metabase Cloud is the short path: you create an instance, invite people, and add a database connection. You do not patch the application. You do still choose which database Metabase may read, and you still own the credentials you paste in.
Self-hosting is the path when the data cannot leave your network, or when you want the application database on infrastructure you already operate. A trial can use the built-in H2 file. A real instance should store Metabase's own data in PostgreSQL or MySQL. That application database is not your warehouse. It is the notebook of saved questions. Lose it, and the charts are gone even if the warehouse is fine.
# Trial only. Production should set MB_DB_TYPE to postgres.
docker run -d -p 3000:3000 --name metabase metabase/metabaseA chart in Metabase is a question sent to your own database every time someone opens it.
Step 1 of 3Your data stays in your database, for example PostgreSQL. Metabase keeps no copy of it; it asks the database each time.
SELECT SUM(total) FROM orders WHERE created_at >= '2026-10-01';Step 2 of 3Metabase turns choices in the query builder into SQL, sends it, and caches the answer only if caching is switched on.
New → Question → Orders → Sum of Total
Step 3 of 3The reader sees the result as a chart, limited by their group's permissions, and can click a bar to see the rows behind it.
Click a bar → View these Orders
Open the next lesson and stand up an instance, or open an existing Cloud instance if you already have one. Then connect a database you are allowed to read. Do not connect a production writer account. A read-only user is enough for every lesson in this course.
Checked against vendor documentation in October 2026. Screens and plan names still change, so confirm the control in your own instance.