TL;DR: your numbers already exist, but nobody looks at them because they live in four different systems. Metabase is an open-source business intelligence tool that connects read-only to those databases and makes them readable: you ask a question by clicking, you get a chart, you assemble the whole thing into a dashboard, you schedule the Monday morning email. The edition you install yourself is free, under the AGPL licence, with no cap on people, where Power BI Pro bills CAD $19.00 per person per month, unless your Microsoft 365 subscription already includes it. The two real limits: your data has to live in a queryable database already, and enterprise single sign-on and row-level permissions start at the Pro tier.
In this article
- Metabase in two sentences: a visualization layer wired to your databases, AGPL outside the commercial directory.
- Who this is for, and who it is not: three questions settle it before you read on, and the queryable database is what rules out the most organizations.
- Asking a question without writing code: the visual editor handles filtering and grouping, the SQL editor stays optional.
- What it changes in a working week: receivables, margin by client, pipeline and grant ratios stop being requests to someone else.
- Metabase versus Power BI: CAD $19.00 per person per month at Microsoft, nothing for the edition you host, unless E5 already includes it.
- Connecting it, in practice: roughly twenty databases read-only, and the real work is agreeing on definitions.
- What it does not solve: no warehouse, no tidying up, and no vendor support in the open source edition.
- The dashboard nobody opens: a question in the title, one person accountable, a scheduled send, a number reconciled in advance.
- What the open source edition includes, and where paying starts: unlimited people and the SQL editor included, but enterprise sign-on starts at the Pro tier.
- How we go about it: three questions, definitions before queries, and a month's wait before the fourth.
Somewhere in your organization sits the answer to "are we profitable this month?". The trouble is that it is scattered across the accounting software, the CRM and three spreadsheets. As soon as anyone wants an overall view, the same ritual starts: export, copy-paste, hand-build a chart. And because the ritual eats half a day of the same person's time, it happens monthly at first, then quarterly, then not at all.
Metabase in two sentences
Metabase is a visualization layer that connects to your databases and presents their contents in a readable form. Once connected, it becomes the mirror of your activity: sales for the month, projects underway, overdue invoices, hours worked. It all refreshes on its own, without anyone re-exporting anything.
The word "open" has a precise meaning here, and it is written in the project's own repository. All the code outside the enterprise directory is licensed under the AGPL, and the images you install from the public channel are released under that same licence. The commercial code lives in a separate directory, with its own licence. Put another way, the software you deploy is yours in the sense that nobody can take it back, and you do not have to ask permission again next year.
Who this is for, and who it is not
Three questions decide this for you, and they take a minute. If two of the three answers are no, the rest of this will serve you less than another article would, and it is better to know now than at the end of a project.
Does your data live in a database you can query? An ERP, an accounting package installed on a server, an in-house application: yes. Three cloud services that offer nothing beyond an "export to CSV" button: no, not directly. This is the condition that rules out the most organizations, and no visualization tool gets around it.
Can someone look after a server, in-house or on your behalf? The saving discussed below comes from self-hosting. If nobody can take that on and you would rather not hand it over, compare the cloud offers against each other instead: the open source edition's price is not your price.
Do you have three specific questions in mind? Not "we would like some dashboards", but three questions whose answer changes a decision this month. If they do not come to mind on their own, a tool is not the first thing you need.
If the first answer is no and your numbers live mostly in spreadsheets, Grist brings together spreadsheet, database and visualization and fits your situation better.
Asking a question without writing code
The real strength of Metabase is its visual question editor: you pick a source, you filter, you group, and you get a chart, without touching the technical language of databases. An executive director can ask for "sales by rep, this quarter, excluding returns" in four clicks. People who are more comfortable write advanced queries in the SQL editor, but that is never a requirement for everyone else.
Those questions get saved, renamed, filed into collections, and assembled into dashboards you share with the team. A dashboard can also go out by email or into a team conversation, at a fixed time. That is often where the change happens: nobody has to remember to go and look.
What it changes in a working week
A dashboard is worth nothing on its own. What counts is the question it answers without anyone asking again.
Accounts receivable. The list of overdue invoices, sorted by age bracket, stops being a request to accounting and becomes a screen you open. Whoever chases payments works from a current list instead of last week's export.
Margin by client. Cross-referencing hours worked against what was actually billed quickly exposes the engagements that cost more than they bring in. It is an uncomfortable conversation, and that is exactly why it never happens without numbers. If your hours already live in a system, that cross-reference is within reach: our article on timesheets in Odoo Community explains where they belong.
The sales pipeline. Opportunities by stage and by rep, with the age of each one, replace the Monday round-table. The round-table then serves to decide, not to collect information the system already holds.
The ratios a grant application demands. A funding application often asks for three numbers that no standard report produces as such. A question saved once produces them again next year in a single click.
Metabase versus Power BI
The comparison comes down to two axes that matter to a small business: what it costs per person, and where the numbers get processed.
| What is being compared | Metabase, open source edition | Power BI Pro | Power BI Premium Per User |
|---|---|---|---|
| Listed price | free | CAD $19.00 per person per month, paid yearly | CAD $32.60 per person per month, paid yearly |
| Number of people | unlimited | billed per person | billed per person |
| Software licence | AGPL outside the commercial directory | commercial subscription | commercial subscription |
| Where the numbers get processed | on the server you choose | in Microsoft's cloud | in Microsoft's cloud |
| Official vendor support | community forums | included with the subscription | included with the subscription |
Tax is added on top, and Microsoft notes that the listed price may differ from the checkout price. For a team of twelve, Power BI Pro comes to CAD $2,736 a year, every year, for the right to look at your own numbers.
The arithmetic does not always favour Metabase, though, and it is worth knowing that before choosing. If you would rather not host it yourself, the vendor sells its cloud service: the Starter tier covers five people for USD $1,080 a year, which is the same order of magnitude as Power BI for a small team. The saving comes from self-hosting, not from the product name.
Two clarifications make the comparison honest, and the first one can wipe out the saving entirely. Microsoft includes a Power BI Pro licence in every Microsoft 365 E5 subscription: if your organization is already on that plan, you are already paying for Power BI and the arithmetic above does not apply to you. Check what your subscription contains before comparing anything. The other clarification cuts the other way: Power BI does have a free tier, but Microsoft states plainly that those users cannot use the sharing or collaboration features with others, or publish content to other people's workspaces. A dashboard you cannot show anyone is not a dashboard: the bill starts at the exact moment the tool becomes useful to the team.
That leaves the question of what "free" actually costs, because open source software has to be installed somewhere. The vendor's published requirements are modest: one core and 1 GB of memory as the floor, then one core and 2 GB for every twenty people connected at the same time, plus one core and 2 GB for the small database that keeps your questions and dashboards. A team of twelve with four looking at once therefore lands around three cores and five gigabytes, application database included. What costs is not the machine: it is the time to install it, back it up, apply the updates and answer when it stops working on a Tuesday morning. So the real question is not "is it free", it is "who looks after it".
Connecting it, in practice
Metabase connects to roughly twenty databases through drivers maintained by the vendor: PostgreSQL, MySQL and MariaDB, SQL Server, Oracle, SQLite, and warehouses such as BigQuery, Snowflake or Redshift. If your ERP, your accounting software or your in-house application runs on one of those, the source already exists. You can also upload a CSV file for the data that has no home yet.
The important point is not technical, it is contractual: the connection uses a read-only account. Metabase watches your data, it never changes it, so a badly built dashboard cannot break anything in the source system. That is the difference between handing over a key and handing over a window.
The real work is not the installation, which takes a morning. It is agreeing on definitions. "An active client", "a sale", "a billable hour": as long as two people in the same organization answer differently, no dashboard will command agreement. That conversation happens before the tool, and the seven decisions to make before installing anything apply here too.
What it does not solve
Metabase visualizes data, it does not tidy it up for you. If your information is scattered and inconsistent, a dashboard will only make the mess more visible, and faster. Plenty of organizations discover at this stage that their problem was never the absence of charts.
It is not a data warehouse either, nor a heavy transformation tool. Cross-referencing very different sources sometimes means preparing the ground upstream, and that preparation is a project in itself. If your data lives mostly in spreadsheets and you are looking to tidy it as much as to look at it, Grist fits that need better: it combines spreadsheet, database and visualization.
Finally, the open source edition comes with no vendor support. Answers come from the community forums, and official support is reserved for the paid tiers. For a tool that ends up carrying the numbers a board of directors looks at, that line deserves to be faced rather than discovered on a Friday.
The dashboard nobody opens
The most common failure mode is not technical. It is a beautiful fourteen-tile dashboard in front of which nobody can say which decision each tile informs. It gets consulted for two weeks, then the tab closes.
Three things prevent it, and none of them calls for technical skill. First, a dashboard carries a question in its title, not a theme: "which invoices do we chase this week?" survives better than "Accounts receivable". Second, it has one person accountable, the one who uses it to decide, and that person says what is missing. Third, it comes to people by email at a fixed time, instead of waiting behind a bookmark.
One last reflex saves a great deal of credibility: before showing a number to the team, reconcile it against the official report that carries authority, the one from the accounting software or the income statement. A dashboard that contradicts the accountant loses the argument every time, and it takes the trust in the other eleven tiles down with it. A 2% gap explained in advance is a footnote. The same gap discovered in a meeting is the end of the project.
What the open source edition includes, and where paying starts
The public grid has four tiers: the edition you install, Starter, Pro and Enterprise. The open source edition is more generous than its price suggests: unlimited people, unlimited charts and dashboards, visual editor, SQL editor, collection permissions, API keys, Google sign-in and basic LDAP, CSV uploads. Dashboards can even be embedded in another site, carrying a "Powered by Metabase" badge. Email delivery goes through your own mail server.
What starts at the Pro tier, billed USD $575 a month on monthly commitment or USD $6,210 a year, is mostly governance: enterprise single sign-on through SAML, JWT or OIDC, automatic account provisioning, row- and column-level permissions, embedding without the product's branding, official collections and verified questions. The Starter tier, at USD $100 a month or USD $1,080 a year for five people, is mainly the vendor's cloud offering.
The useful reading for a small business is simple: if everyone is allowed to see the same numbers, the open source edition is enough. If some people must only see their region, their department or their own files, the row-level permissions of the Pro tier become the real subject, and the decision turns on governance, not on charts.
How we go about it
A dashboard project that lasts starts small and gets judged on use, not on the number of screens delivered.
What we recommend when setting up a business intelligence layer for an organization:
- Pick the three questions that change a decision, and nothing more. Not twelve.
- Write the definitions before the queries, and have them validated by the people who will live with them.
- Connect the sources read-only, with a dedicated account that does nothing else.
- Host the tool where you want your numbers to stay, and write it down in your processing register.
- Schedule the weekly delivery of the three dashboards, so nobody has to think about it.
- Wait a month before adding a fourth.
The hosting choice is not a plumbing detail. A business intelligence layer ends up reading everything you have: sales, salaries, client files. Knowing where those queries run and under which jurisdiction is a matter of Law 25 compliance as much as technical preference, and it is also what separates a tool you install from a service you rent: what stays standing when a foreign vendor cuts off access is not a theoretical question.
Does your data have something to tell you? Let's talk about your first three questions.
Sources
- Metabase, the pricing grid: the four tiers, what the open source edition includes, and the features that start at the Pro tier.
- Metabase, the repository licence file: the AGPL outside the enterprise directory, and the commercial licence inside it.
- Metabase, the database connection documentation: the list of drivers maintained by the vendor.
- Microsoft, Power BI pricing in Canada: the per-person rates in Canadian dollars, tax extra.
- Microsoft, the Power BI licensing guide: every Microsoft 365 E5 subscription includes a Pro licence.
- Microsoft, Power BI service features by license type: what the free licence allows, and why it does not allow sharing.
- Metabase, how to run Metabase in production: the CPU and memory requirements, and the per-twenty-people rule.
- Metabase: the project's official site.