Apache Superset is the 2nth default for new builds — open-source, self-hosted, and flexible enough to cover most BI needs without a per-seat licensing conversation. Metabase, Evidence, Druid, and Grafana are mapped for the cases where Superset isn't the right shape.
Five tools that cover different points on the BI spectrum — from full-featured SQL exploration to code-first reporting to real-time OLAP. Superset is live; the rest land as content ships.
Open-source BI platform. SQL Lab for ad-hoc queries, 40+ chart types, dashboard authoring, embedded analytics via guest tokens. The 2nth default — preferred over Power BI and Tableau for new builds.
Self-service BI with a friendlier learning curve. Point-and-click questions, embeddable dashboards, and a community edition that covers most small-to-mid reporting needs without writing SQL.
Code-first BI with markdown reports. Write SQL in .md files, generate interactive pages, version control your dashboards alongside your dbt models. For teams that live in the IDE.
Real-time OLAP for sub-second queries on event streams. Not a BI tool in the traditional sense — more the engine underneath one. Pairs with Superset as the query backend for high-cardinality, time-series-heavy workloads.
Metrics, logs, traces visualization. Primarily an ops tool, not a BI tool — but the line blurs when your infrastructure telemetry is the dataset. Loki, Tempo, Mimir, and the observability stack.
The tools in this sub-hub differ in UI and target audience, but they all follow the same four-layer architecture. Understanding this pattern makes every individual tool easier to reason about.
Every modern BI deployment follows the same flow. Data lands in a warehouse or lakehouse (PostgreSQL, ClickHouse, DuckDB, BigQuery). A semantic layer defines the business logic — metrics, dimensions, relationships — so that every dashboard computes the same number for "revenue" instead of each analyst inventing their own SQL. The dashboard tool queries the semantic layer and renders charts. The output is consumed via browser, embedded iframe, scheduled email, or API.
# The four-layer BI architecture Source systems → ERP, CRM, logs, events, APIs ↓ Warehouse → PostgreSQL / ClickHouse / DuckDB / BigQuery ↓ (dbt transforms raw → staging → marts) Semantic layer → Superset datasets / Metabase models / cube.dev ↓ (metrics, dimensions, row-level security) Dashboard tool → Superset / Metabase / Evidence / Grafana ↓ Consumers → Browser / embedded iframe / email / Slack / API
The payoff: swap any layer without rebuilding the stack. Move from PostgreSQL to ClickHouse without touching your dashboards. Switch from Metabase to Superset without re-modeling your data. The architecture is the constant; the tools are the variables.
Analytics consumes what engineering and warehousing produce. Every ERP integration ends up feeding a dashboard; every ML model's output gets visualized somewhere.