Metabase Self-Hosted vs Cloud: What Each Really Costs You

Break the total into recurring costs and time. Most teams undercount the second.

• Compute and storage. The server or container that runs the application, sized for your users and queries.
• The application database. Metabase stores its own settings, questions and dashboards, and you should follow the vendor's current production guidance for where that data lives. Budget for a managed database or its administration.
• Backups and recovery. Someone has to set them up, and test that a restore works.
• Upgrades. Regular version updates, plus testing that your dashboards still work afterward.
• Security. Access control, single sign-on if you need it, network rules, TLS certificates and patching the underlying system.
• Monitoring. Alerts for when the service is down or slow, and someone who responds.
• Licensing. Check the current terms for the edition you plan to run, since features and obligations differ between editions.
• Engineer time. Setup, ongoing care and the interruptions when it breaks. This is usually the largest line.

Estimate hours per month, then multiply by the loaded hourly cost of the person doing the work. Loaded cost includes taxes, benefits and overhead, and you can build it with the fully loaded cost worksheet.

In this example, say setup takes 20 hours once, and routine care takes 4 hours a month for upgrades, backup checks and user requests. Say an incident costs another 6 hours twice a year. That's about 5 hours a month on average. At a loaded rate of $80 an hour, in this example, that's $400 a month in time.

Add the infrastructure. In this example, hosting, a managed database and backups come to $150 a month. In this example, the self-hosted total is $550 a month, plus the one-time setup of 20 hours, or $1,600.

The hidden piece is opportunity cost. Those hours could go to work that directly supports customers or revenue, so ask whether analytics administration is the best use of your scarce technical time.

Get a written quote based on your actual number of users and any features you need, such as single sign-on, permissions or embedding. Pricing structures and tiers change, so read the vendor's current pricing page and ask what happens as you add users.

Look beyond the monthly price:

• What's included in support, backups, upgrades and uptime commitments.
• How users are counted. Ask whether viewers, editors and admins count differently.
• Data connectivity. A hosted service must reach your databases, so you may need to allow its network addresses, set up a tunnel, or expose a read replica. Involve whoever manages your database security.
• Data residency and compliance needs. If your rules require data to stay in a certain place or network, confirm the hosted option meets them.
• Exit terms. Ask how you can export your questions and dashboards if you leave.

Put both into one table with identical assumptions over 12 and 36 months. Use the same user count and features for both, and include one-time setup for self-hosting.

• Self-hosted, year one: hosting, database and backups, plus recurring engineer time, plus setup time.
• Hosted, year one: subscription, plus any one-time setup, plus the network work to connect your data.

Then run a sensitivity check. Ask how the answer changes if engineer time is double your estimate, if you double the number of users, or if the person who maintains it leaves. Self-hosting is fragile when one person holds all the knowledge, so write the setup down, as you would for any system a single person keeps running.

It's a good fit when you have someone whose job includes running infrastructure, when your data can't leave your network, when you want tight control over versions, or when you're running it for many users and the subscription would cost more than the operating time. It's a poor fit when no one owns it, when it's a side project for a founder, or when downtime would block decisions.

Also consider where you are in your journey. A team can begin on a hosted trial to prove the value, then move later, provided you can export your work. If you're still deciding whether Metabase is the right tool, compare it in BI tool comparison, BI tools for operations teams and BI tools for freelance IT services. If your source is accounting data, read connecting accounting data to a dashboard tool for a comparable data-path decision.