Metabase Problems & Issues
Real user-reported problems, complaints, and limitations for Metabase. Based on analysis of 14 signals.
Overall score: 70/100
Issues by Category
Self-hosted Metabase crashes with OutOfMemoryError
Self-hosters report 'java.lang.OutOfMemoryError: Java heap space' crashes, sometimes recurring every 12-24 hours. Metabase runs on the JVM and needs 2 GB+ of RAM; the docs recommend raising the heap with -Xmx and capturing heap dumps.
Metabase upgrade failed: migration lock or H2 corruption
Upgrades can stop at 'Database has migration lock' or fail halfway, and the default H2 application database can corrupt or lose data. Metabase's docs say H2 is not for production and recommend migrating to PostgreSQL.
Metabase not syncing new tables or columns
A frequent forum and docs topic: new tables, columns or changed types don't show up until a sync runs, and JSON unfolding or large schemas make syncs slow. Filter dropdowns only sample the first 1,000 values per field.
Metabase 'Your question took too long' timeouts
Long queries end with 'Your question took too long', often at exactly 60 seconds because a load balancer or Nginx proxy cuts the request. Users report raising Metabase-side timeouts doesn't help until the proxy is fixed.
Why is Metabase Pro so expensive?
Starter is $100/month for 5 users, but SSO (SAML), row- and column-level permissions, custom domains and full embedding need Pro at $575/month for 10 users plus $12 per extra user. Enterprise starts at $20,000 per year, so the jump between tiers is steep for small teams.
Metabase Starter price went up from $85 to $100
Pricing trackers record Starter rising from $85 to $100 per month and the Pro per-user add-on from $10 to $12 in Q3 2025. Existing cloud customers on monthly billing pay the new rate.
Metabase usage fees for AI, transforms and storage
On Cloud, the managed AI service is free for the first 1M tokens then $3.75 per 1M, transforms are free for 1,000 runs then 1-2 cents per run, and built-in storage is billed per million rows. A development instance costs another $100/month on Pro.
Metabase dashboard filters and linked filters not working
Dropdowns that don't populate, linked filters that show every value, and filters broken in embeds or public links are among the most common forum threads. Causes include fields not set as Category, missing foreign keys and models used as filter sources.
Advanced Metabase dashboards still need SQL
The query builder covers simple questions, but Capterra reviewers say complex metrics, cross-database work and advanced dashboards push users back to SQL, and large libraries of questions become hard to organize.
Metabase exports and PDF reports are limited
Product Hunt and review-site users mention awkward exports, limits on emailed and PDF reports, and Metabase branding on exports and subscriptions on the free and Starter plans.
Why is Metabase so slow with large datasets?
Slow performance is the most-cited con on G2 and shows up in Capterra and SaaSworthy reviews too: dashboards and queries lag once tables get large or many cards run at once. Metabase's own docs say a 50-card dashboard is almost always slower than five 10-card dashboards.
Metabase charts are too limited to customize
Reviewers on G2, Capterra and Software Advice say visualization options and formatting control sit a step below Tableau or Power BI. Teams that want pixel-level dashboard design or unusual chart types hit the ceiling quickly.
Metabase open source has no row-level security or SAML
The free edition lacks row- and column-level permissions, SAML, SCIM and white-labeling, and exports and scheduled emails carry Metabase branding. Teams that need per-customer data isolation discover they must move to Pro.
Metabase support is slow on Starter and Pro
Open source users get community forum help only, and Starter and Pro come with a 3-day support SLA. A 1-day SLA and a dedicated success engineer are Enterprise-only, from $20,000 per year.
Regret Patterns
Common reasons users regret choosing Metabase.
Starting on H2 and losing setup
Teams try Metabase with the default H2 database or a Docker container without a volume, then lose questions and dashboards on restart or face a painful migration later.
Needing row-level security after launch
A team builds on open source or Starter, then needs per-customer data isolation or SSO and has to jump to Pro at $575/month.
One giant dashboard that nobody can load
Dashboards grow to dozens of cards on raw tables until every load takes minutes and users stop opening them.
Outgrowing the chart options
Analysts hit formatting and chart-type limits and start exporting to spreadsheets or a second BI tool.
Self-hosting turns into a maintenance job
Upgrades, JVM memory tuning, SMTP setup and migration locks take more engineering time than expected.
Cancelling Cloud deletes the instance
The only cancel button deletes the instance, so teams that wanted to keep their work must request a backup first.
When Metabase Breaks Down
Specific scenarios where Metabase fails or underperforms.
Dashboards with 30+ cards on raw tables
All cards query the database at once, queue up and time out; splitting dashboards and caching are the documented fix.
Queries longer than 60 seconds behind a proxy
Nginx or a load balancer cuts the request and users see 'Your question took too long'.
High-cardinality filter fields
Metabase samples only the first 1,000 values, so dropdowns are incomplete or switch to search boxes.
Docker without a persistent volume
Every restart shows the setup screen again because the H2 application database lived inside the container.
Self-hosting on a 1-2 GB server
Syncs and heavy queries trigger OutOfMemoryError and crashes.
Customer-facing embedding on Starter
Guest embeds show a 'Powered by Metabase' badge and lack tenant isolation; Pro is required.
Mixed time zones across database and server
Dates shift by hours or a full day in charts and filters until report time zone and JVM time zone line up.