Metabase Tutorial: Let Your Non-Technical Colleagues Query Data Themselves

At 10:00 AM, a marketing manager at a Taipei e-commerce company asks in the group chat, "How many orders came through LINE last month?" The engineer sees the message, sighs, spends seven minutes opening the database to write an SQL query, and drops a screenshot back into the chat. At 3:00 PM, the same person asks, "What about the same period last year?"

This loop plays out every single day in SMEs across Taiwan. The problem isn't that engineers aren't willing to help—it's that this workflow shouldn't involve engineers in the first place. Metabase is built to solve exactly this.

What Is It?

Metabase is an open-source Business Intelligence (BI) tool. Hook it up to your company's database, and non-technical colleagues can use a point-and-click interface to build queries, save them as "Questions," and piece multiple questions together into a dashboard. With the AI-powered Q&A and SQL generation features added over the last couple of years, users can now even describe what they need in plain language.

Its biggest selling point is that its learning curve is the flattest among similar products. That might sound like marketing talk, but in practice, it makes a massive difference—the most common way a BI tool dies is not a lack of features, but that nobody uses it after deployment.

What Can It Do?

• Build queries using a point-and-click interface—no SQL required
• Save frequently used queries as "Questions" and re-run them anytime
• Combine questions into dashboards and schedule automated deliveries to Email or Slack
• Ask questions in natural language and auto-generate SQL (available across all plans, not locked behind a paywall)
• Set up alerts to notify teams when metrics spike or drop
• Embed dashboards directly into your own product for customers to view

How to Use It: From Zero to Your First Dashboard

Step 1: Installation

The fastest way is via Docker with a single command:

bash
docker run -d -p 3000:3000 --name metabase metabase/metabase

Once it's running, open http://localhost:3000 and follow the setup wizard. The open-source version is fully featured with unlimited users, and self-hosting is completely free.

Two things to watch out for in a production environment: First, it defaults to using an embedded H2 database to store Metabase's own configuration. Make sure to switch to an external PostgreSQL database for production, otherwise you risk running into trouble during upgrades or server migrations. Second, remember to set up backups.

Step 2: Connect Your Database

Add a new data source in the settings by filling in the host, port, database name, and credentials. It supports common options like MySQL, PostgreSQL, SQL Server, and BigQuery.

The key to this step is permissions. Be sure to create a read-only account for Metabase; don't take a shortcut and drop root access in. A BI tool only needs to read data—giving it write permissions offers zero benefits and introduces unnecessary risks.

Step 3: Build Your First "Question"

Click "New → Question," select a data table, and use the UI to set your criteria: filters (e.g., "Order date is last month"), aggregations (e.g., "Count"), and groupings (e.g., "Group by source").

This gives you the exact answer the marketing manager from our opening story wanted, and she can even change it to the same period last year herself—she just needs to adjust the date filter without bothering an engineer.

Once saved as a question, you can choose a chart type: bar chart, line chart, pie chart, or table.

Step 4: Combine into a Dashboard and Automate Delivery

Add related questions to a single dashboard—for example, "This Month's Revenue, Order Count, Revenue Share by Source, Top 10 Bestselling Products." Next, set up a "Subscription" to automatically email it to managers every Monday at 9:00 AM.

This automated delivery feature is severely underestimated. Most people won't actively open a dashboard, but they will definitely check their email. Pushing the numbers right in front of them is far more effective than waiting for them to look.

Advanced Tips

How to use AI Q&A. Ask questions in plain language like "Revenue by channel last month," and it will generate and execute the SQL. From hands-on experience: Accuracy is quite good for simple aggregation questions, but it easily trips up once multiple table joins or complex business logics are involved. My approach is to use it to generate a draft and then review the SQL myself, which is much faster than writing from scratch. You can use the official AI service (priced at $3.75 per million tokens, with a 1 million token starter bonus) or bring your own API key—the latter is more cost-effective for teams that already have API credits.

Build Data Models. If the exact same join logic keeps popping up everywhere, save it as a Model. Your colleagues can then run queries on a clean dataset without having to rebuild relationships from scratch every time. This is a crucial step to getting non-technical teammates to actually use the tool.

Make the Most of Field Descriptions. Add human-readable explanations to fields in the data settings—for example, labeling ord_st as "Order Status: 1=Pending Payment, 2=Shipped." Colleagues can only query data themselves if they actually understand what the fields mean. This sounds tedious, but it determines whether the system will actually get adopted.

Things to Watch Out For: Pricing Is the Biggest Trap

The self-hosted open-source version is free. Cloud-hosted paid plans are priced per seat:

• Starter: $100/month ($90 billed annually), includes 5 seats, with additional seats at $6/month each.
• Pro: $575/month ($517.50 billed annually), includes 10 seats, with additional seats at $12/month each.
• Enterprise: Custom pricing, starting at $20,000/year.

The biggest trap is that end-users viewing embedded dashboards also count as seats. If you plan to embed dashboards into a SaaS product for clients to view, the costs under this model can scale up significantly. Always run the numbers first. Many teams only realize this when they receive their first bill.

Additionally, the self-hosted version lacks SSO and granular row/column-level permission controls. If your data includes personal information, salaries, or cross-departmental sensitive figures, you'll need to figure out permissions on your own—likely by using database-level views as a workaround.

One more thing: Free doesn't mean zero cost. Self-hosting requires someone to look after the database, handle upgrades, and manage backups. These are all hidden labor costs.

When You Shouldn't Use It

Complex analysis still requires writing SQL. Point-and-click interfaces cannot handle multi-layered CTEs (Common Table Expressions) or window functions. If your analysis requires statistical modeling or machine learning, use Python instead of forcing a BI tool to do it.

For a more systematic way to string together your data workflows, check out our guide on AI Data Analysis Tools or explore the data analysis prompts in our Prompt Template Library.

TheAI Academy Review

I've seen too many companies buy expensive BI tools only to find that only the data team uses them while business units still run back to ask the engineers. Metabase breaks through this barrier not because it has the most features, but because it makes non-technical colleagues genuinely brave enough to click around themselves.

Review: Metabase's value doesn't lie in pretty charts; it lies in taking the burden of "pulling numbers" off engineers. However, don't let the words "open-source and free" fool you—per-seat pricing and maintenance labor are the real costs.

Practical advice for readers: Start by self-hosting the open-source version via Docker, spend a week turning "the ten most frequently asked questions in the company" into a single dashboard, and set it up to auto-email your managers every week. The ROI on this step is the highest of the entire deployment process. Once you genuinely need SSO or need to embed views for customers, evaluate the paid plans—and when you do, make sure to factor in the seat costs for external users. Converting USD pricing into local currency will make the final number much larger than it looks at first glance.

Sources

• Metabase Official Pricing Page (Plans, per-seat pricing rules, AI usage billing)

Compiled from public information; official sources take precedence. Actual pricing is subject to change.