Apache Superset

Superset in brief

Apache Superset is an open source platform for looking at data. You connect it to a database, define what a chart should be built from, build charts and put them together into dashboards.

It does not hold your data. Queries go to the database you connected, and the results come back to be drawn. Superset stores only its own metadata: the charts, the dashboards, the dataset definitions, the users.

Three layers to understand

Superset is easier to learn once you can see that it has three layers stacked on each other.

The database connection. Where the data actually lives. Configured once by someone with the credentials.

The dataset. What charts are built from. It can be a physical table, or it can be defined by a query when the shape needed is not a table that exists. Metrics and calculated columns are added here.

The chart. One visualisation, built from one dataset, by choosing a chart type and dragging dimensions and metrics into place.

Dashboards then arrange charts on a page and can apply filters across them.

Why the dataset layer earns its place

It is tempting to skip datasets and write a query per chart. The dataset layer exists because that approach falls apart quietly.

If a metric such as active accounts is written inside each chart, twelve charts hold twelve copies of that logic. When the definition changes, some get updated and some do not, and nobody notices until two dashboards disagree.

Defining it once on the dataset means every chart built from that dataset uses the same definition. This is a small piece of governance that saves a large amount of reconciliation.

SQL Lab

Not everything is best expressed by dragging fields into place, and many analysts are faster writing SQL.

SQL Lab is a query workbench inside Superset. Queries can be written, run and saved, results inspected, and a useful query turned into a dataset or a chart without leaving the tool.

The result is that one platform serves both audiences rather than forcing everyone into the same way of working.

What connecting means in practice

Superset connects through Python database drivers, which is why the list of supported databases is long. The documentation has a page per database, covering the connection string format and anything specific to that engine.

Because queries run against the connected database, dashboard responsiveness depends on that database. A slow warehouse produces a slow dashboard, and the fix belongs there rather than in Superset.

Who Superset suits

Superset suits organisations that want analytics available broadly without per user licensing shaping the decision, and teams comfortable running the platform themselves. Both chart builders and SQL writers are served.

It is also used through managed offerings, for teams that want the platform without operating it.

Points to be aware of

It is software you run. Deployment, upgrades, authentication, scaling and caching are your responsibility, and a production installation is a real piece of infrastructure rather than an application to install and forget.

Chart configuration is also broad, which means a first chart takes some exploration. The documentation covers each chart type and its options.

Getting started

The documentation includes a quickstart that runs Superset with Docker Compose, then walks through connecting a database, creating a dataset, building a chart and assembling a dashboard. Separate sections cover installation, configuration, security and each supported database.