Power BI Alternative Open Source: Superset, Metabase

Have Apache Superset or Grafana operated? The expertise pages for Apache Superset and Grafana describe hosting and operation in Germany. Book an initial consultation

Anyone looking to replace Power BI will not find a single open source product with the same scope, but three tools with different emphasis: Metabase for self-service analysis, Apache Superset for SQL-centric BI with centrally maintained datasets, and Grafana for dashboards over metrics, logs and databases. This article sorts the three by use case, license, operation and data sources, and names what cannot be carried over from Power BI.

All statements come from LICENSE files, vendor documentation and pricing pages as of 29 October 2026.

Table of contents

1. Short answer
2. Decision rules by use case
3. Licenses in detail
4. Data sources and modeling
5. Operation and architecture
6. Embedding and tenant separation
7. What does not carry over 1:1 from Power BI
8. Status and methodology
9. Our approach at WZ-IT
10. Further guides
11. Sources

Short answer

Decision rules by use case

Business teams build analyses themselves, without SQL. The Metabase query builder assembles a question step by step from filters, summaries and joins and shows the first ten rows as a preview. Anyone who knows SQL switches to the native editor. Converting a query builder question to SQL is, according to the documentation, a one-way street. For this case Metabase is the obvious candidate.

Analyst team with SQL skills and centrally maintained metrics. Superset works with datasets that carry metrics such as SUM(Cost), and with SQL Lab as the editor. SQL Lab queries must be saved as a dataset before charts are built on them. This suits teams that define metrics centrally and build charts on top.

Operations dashboards for servers, applications and sensors. Grafana is built for time series, logs and traces, with variables for interactive dashboards, transformations and alerting. Data from PostgreSQL, MySQL or Microsoft SQL Server can be added as well. Grafana is the suitable starting point for ops views. Business reporting with modeling belongs rather in Metabase or Superset.

Embedded analytics in your own product. Superset ships an embedded SDK (@superset-ui/embedded-sdk) that integrates dashboards via guest token. In Metabase, multi-tenant embedding and white-label are part of Pro and Enterprise. The license consequences differ considerably (see below).

Business and operations in the same organization. Both worlds can be combined: Grafana for operational data, Metabase or Superset for business data, both against the same PostgreSQL database. The tools have different jobs and do not exclude each other.

Print-ready reports with fixed page layout. Paginated reports (.rdl) are a separate format in Power BI for documents such as PDF and Word. The documentation of Metabase, Superset and Grafana that was read does not describe a counterpart. This use case has to be planned separately.

Licenses in detail

The network clause of the AGPL (section 13) applies when a modified version is operated: all users who interact with it over a network must then be offered the source code. Running Metabase or Grafana unmodified for internal use does not trigger this obligation. Anyone who makes their own changes or builds the software into a product clarifies the license question beforehand. This is orientation, not legal advice.

The license scope is not uniform inside a repository. For Metabase this applies to the enterprise directory, for Superset to bundled third-party components, for Grafana to individual directories under Apache-2.0 (LICENSING.md). Before using a tool in a product, check the LICENSE files of the exact version in use.

Data sources and modeling

Metabase lists 18 databases with officially maintained drivers: Athena, BigQuery, ClickHouse, Databricks, Druid, MongoDB, MariaDB, MySQL, Oracle, PostgreSQL, Presto, Redshift, Snowflake, SparkSQL, SQL Server, SQLite, Starburst and Vertica. Community drivers exist for others. The pricing page states "20+ supported data sources" instead.

Superset does not ship database drivers. For each engine, a Python DB-API driver and a SQLAlchemy dialect must be installed in the container or pip setup. This is part of operation and has to be planned for every new database type.

Grafana ships core data sources by category: metrics (Prometheus, InfluxDB, Graphite, CloudWatch and others), logs (Loki, Elasticsearch), traces (Tempo, Jaeger, Zipkin), profiles (Pyroscope) and SQL databases (MSSQL, MySQL, PostgreSQL).

Modeling works differently in all three than in Power BI. There is no tabular in-memory model with relationships and DAX context. Metrics are created in SQL, in datasets and metrics (Superset), in questions with custom expressions (Metabase; per the docs these columns exist only in the respective question) or directly in the query (Grafana). Anyone building many reports on the same metrics should define them once in the database, for example as a SQL view.

Operation and architecture

Superset, according to its architecture page, consists of the Flask web server, a metadata database (recommended for production: a standalone, managed database; tested with PostgreSQL and MySQL), a cache such as Redis, and Celery workers plus the beat scheduler. Cache and workers are required for Alerts & Reports, caching, asynchronous queries and dashboard thumbnails.

Metabase runs as a single application with an application database. The embedded H2 database is not suited to production; the Docker Compose example in the docs uses PostgreSQL but is explicitly not meant for production. If the container is removed without an external database, the Metabase data is lost.

Grafana stores its configuration in SQLite by default. For production the docs recommend MySQL (8.0 and later) or PostgreSQL (12 and later); high availability requires one of them.

For all three: BI tools hold credentials for the connected data sources. A delayed update therefore also affects those systems. The Metabase SQL injection with CVSS 10.0 shows how serious this can be. Updates, backups of the metadata database and access restriction are part of operation.

Embedding and tenant separation

Superset: The embedded SDK integrates dashboards via embedDashboard(). The EMBEDDED_SUPERSET feature flag is required. The backend of the host system fetches a guest token from /api/v1/security/guest_token/ using a service account; by default it expires after five minutes (GUEST_TOKEN_JWT_EXP_SECONDS) and must be renewed. Allowed domains can be set per dashboard; an empty list allows any origin. Row filters (rls) can be passed in the token.

Metabase: According to the pricing page, "multi-tenant embedded analytics", white-label and row and column permissions are part of Pro. Pricing counts internal users and users of embedded analytics alike. The file app-embed.js is under its own license that prohibits removing the Metabase branding in the iframe unless a separate agreement exists.

Anyone offering analytics to customers inside their own product therefore compares not only features but also license model and user counting.

What does not carry over 1:1 from Power BI

What can be carried over are the data sources, the SQL logic behind the reports, the business metric definitions and the permission concept at business level. A sensible sequence: inventory the reports, sort by usage, write metrics out in SQL or views, and only then rebuild dashboards. Unused reports do not need to be migrated.

Status and methodology

As of: 29 October 2026. License statements come from the LICENSE files of the repositories (Superset LICENSE.txt on the main branch, Metabase LICENSE.txt, Grafana LICENSE and LICENSING.md). Feature and operation statements come from vendor documentation, prices from the Metabase pricing page (retrieved 29 October 2026, monthly and yearly prices in USD). For Power BI only Microsoft documentation pages were used. Not assessed: performance, range of chart types, Grafana embedding and Metabase schedules. Anyone who needs one of these points for the decision should check it on a test instance with their own data.

Our approach at WZ-IT

There are expertise pages for Apache Superset and Grafana describing hosting and operation in Germany. Where hosting, updates and backups of the chosen tool are needed, the services are described there. There is no Power BI consulting offer; the assessment in this article concerns choosing on the open source side.

Further guides

• Metabase emergency patch: SQL injection with CVSS 10.0
• ThingsBoard, Grafana and ChirpStack in the IoT stack
• Grafana expertise
• Apache Superset expertise

Sources

• Apache Superset, LICENSE.txt
• Metabase, LICENSE.txt
• Grafana, LICENSE
• GNU AGPL 3.0
• Metabase, pricing
• Metabase, embedding license
• Metabase Docs, connecting databases
• Metabase Docs, query builder
• Metabase Docs, Docker
• Superset Docs, architecture
• Superset Docs, databases
• Superset Docs, embedding
• Superset Docs, exploring data
• Grafana Docs, data sources
• Grafana Docs, dashboards
• Grafana Docs, alerting
• Grafana Docs, installation
• Microsoft Learn, DAX overview
• Microsoft Learn, dataflows
• Microsoft Learn, Power BI Report Server