Pick Metabase if you want open source analytics you can self-host, where people who don't write SQL can start from a visual query builder and developers can embed charts in a product. Pick Looker if you're on Google Cloud and need one governed set of metric definitions, written in LookML, shared by dashboards and AI agents alike.
Who's going to be asking the questions? If it's the sales lead, the ops manager and the founder, poking at numbers between meetings, you need something they can open and use today. If it's a whole company that keeps arguing about what "active customer" means, you need one agreed definition sitting under every chart. Those two needs point at two very different tools.
Where my advice lands
For a small or mid-sized team without dedicated data modellers, Metabase is the quicker start. It's open source, you can run it yourself or use the managed Cloud option, and a non-technical person can build a question from the visual query builder without touching SQL. For a larger organisation already on Google Cloud, where agreeing metrics matters more than speed of setup, Looker makes more sense, because its whole pitch rests on business logic defined once in LookML.
Neither is better across the board. One is built to get people answering questions fast; the other is built to make sure everyone's answers agree.
How the two compare, line by line
| Area | Metabase | Looker |
|---|---|---|
| Licence and deployment | Open source, self-hosted or managed Cloud | Managed Google Cloud product with Cloud Console and IAM single sign-on |
| Modelling | Visual query builder, full SQL editor; Data Studio curates a semantic layer | Business logic defined once in LookML above the warehouse |
| AI features | Metabot, an MCP server and a CLI; inspect the query behind answers | Gemini-powered Conversational Analytics, Dashboard Agents and its own MCP server |
| Embedding | Modular Embedding SDK and white-label analytics | Embedded analytics with iframes, SDKs, APIs and a Conversational Analytics API |
| Data sources | 20+ databases and warehouses, plus CSV upload | Sits above a cloud warehouse, native BigQuery; ad hoc CSVs blend with LookML models |
| Learning curve | Pitched as easy for everyone, no SQL required | LookML, API and admin docs; a modelling layer to learn |
| Getting started | Open source | Free trial stated |
Running it yourself versus renting it from Google
This is the difference that shapes everything else. The open source tool can be installed on your own machine, and its home page even offers a Docker or Java command to get it running locally. That means a curious developer can have it up before lunch, and you keep your data on your own servers if that matters to you.
The other is a managed Google Cloud product. You get it through the Google Cloud Console, with IAM single sign-on and one account alongside your other Google Cloud services. If your warehouse is already BigQuery, that's a natural home. If you're not on Google Cloud at all, it's a bigger step.
There's a second half to this. The Google tool's governed-metrics promise depends on LookML. Someone has to write and maintain that layer before the dashboards and AI agents above it can do their job. The open source option lets a non-technical user start from the query builder on day one, and you can add structure later through its Data Studio. If you don't have anyone who'd own a modelling layer, that's worth being honest about.
Both now lead with AI agents
If you last looked at these two as simple dashboard tools, that picture's out of date. Both home pages now lead with AI and MCP servers.
The open source side offers Metabot, an MCP server and a CLI, and its site makes a point of letting you inspect the query behind every answer. That's useful if you've ever been handed an AI answer and wondered where it came from. Google's tool offers Gemini-powered Conversational Analytics, Dashboard Agents and its own MCP server. The vendor says its semantic layer grounds both human analysts and AI agents in one source of truth, which is the LookML argument again, applied to AI.
Putting charts inside your own product
If you build software and want customers to see analytics inside it, both can do the job. The open source tool has a Modular Embedding SDK and white-label analytics. Google's embedded analytics product covers iframe embedding, SDKs, APIs and a Conversational Analytics API, so you could put a chat-style question box in your own app.
On data sources, the self-hostable option lists 20+ databases and warehouses plus CSV upload. The Google option sits above a cloud warehouse, with native BigQuery integration, and lets you blend ad hoc CSVs with LookML models.
Three other BI tools to consider
If you'd like a wider view, our guide to open source BI tools covers more self-hosted options, and there's a separate page on tools to use instead of Metabase if the open source route appeals but this one doesn't.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Apache Superset
Best for: Engineering-led teams wanting an open source exploration and visualisation platform they run and customise, backed by an active community on Slack and GitHub.
Power BI
Best for: Microsoft's BI product inside Fabric, with Copilot to explore and explain data, a natural pick if your organisation already works in Microsoft tools.
Redash
Best for: An open source tool centred on an online SQL editor for querying sources, building dashboards and sharing them with SQL-comfortable colleagues.
What to try in your first week
- Write down three questions your team asks every week, and see how long each tool takes to answer them.
- Check where your data lives. If it's BigQuery, the Google Cloud option gets more attractive.
- Ask who would own a modelling layer. No owner usually means a lighter start is wiser.
- If you embed analytics for customers, try each embedding route against a real page in your product.
Do you make a BI or analytics tool? Tool makers can ask to be considered for a featured place on this page.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
Frequently asked questions
Is Metabase open source?
Yes. You can self-host it, and its home page gives a Docker or Java command to run it locally. There's also a managed Cloud option if you'd rather not run it yourself.
Do I need to learn LookML to use Looker?
Someone on the team does. Business logic is defined in LookML, and the governed-metrics pitch depends on that layer being built and maintained.
Does Looker only work with BigQuery?
No, it sits above a cloud data warehouse, but it has native BigQuery integration and is sold as a Google Cloud product, so it fits most naturally there.
Which is easier for people who don't write SQL?
The open source tool is pitched as easy for everyone with no SQL needed, starting from a visual query builder. On the Google side, users can explore once the LookML model exists.
Can both be embedded in my own app?
Yes. One has a Modular Embedding SDK and white-label analytics; the other has an embedded analytics product with iframes, SDKs, APIs and a Conversational Analytics API.
