Metabase and Apache Superset Alternative for Zero-Server Local Files

Why Metabase and Superset are heavy for file-based and ad-hoc analytics

Metabase and Apache Superset are popular self-hosted BI portals for engineering teams with a production Postgres or ClickHouse database. When you simply want to explore a local CSV, Excel, Parquet, or .twbx file, their server architecture gets in the way:

1. You have to run a server and a database first. Self-hosting Metabase requires running a Java/JVM process or Docker container, and uploading CSVs requires connecting a supported writable database (such as PostgreSQL, MySQL, Snowflake, Redshift, or ClickHouse Cloud) and selecting a database and schema where Metabase's database user has permission to create tables. Self-hosting Apache Superset requires a multi-container Docker Compose stack (Python web server, Celery workers, Redis, Postgres metadata DB) or Kubernetes deployment.
2. Hosted cloud tiers aren't built for solo local files. Metabase Cloud Starter costs $100/mo monthly or $90/mo billed yearly ($1,080/year); first 5 users included; +$6 per user/month ($65 per user/year). Preset (managed Superset) has a free Starter tier (up to 5 users and 1 workspace; Professional is $20/user/month billed annually; $25/user/month billed monthly), but still requires connecting a database or uploading flat files into a hosted database.
3. Form-based chart builders vs calculation and shelf grammar. Both Metabase's Notebook editor and Superset's Explore form ask you to pick a pre-set chart type and fill in form dropdowns. Neither gives you Level of Detail (FIXED, INCLUDE, EXCLUDE) expressions, .twbx workbook support, or Tableau-style draggable shelves (Columns, Rows, Marks card).

How to run zero-server BI in your browser with xmetric

xmetric removes the server, the Docker stack, and the database import step for file-based analytics:

1. Open the xmetric app (/app/) in any browser. Nothing to docker compose up, no JVM memory flags, and no database credentials to configure.
2. Load your files. Import a .csv, .xlsx, or .parquet file, or open a .twb or .twbx workbook, directly into local browser memory. The free workbook inspector (/inspect/) also reads .pbix and .pbip files.
3. Build with shelves and calculations. Drag dimensions and measures across Columns, Rows, Color, and Size, and write Level of Detail (FIXED, INCLUDE, EXCLUDE) or table calculations without SQL boilerplate. To test a formula alone and see the equivalent SQL, use the formula playground (/play/).

Writing multi-level metrics without a SQL subquery view

In Metabase or Superset, comparing a row's metric against a cohort or parent-group aggregate usually forces you out of the visual builder and into the raw SQL editor (or a Superset Virtual Dataset). In xmetric, you write a single Level of Detail expression.

Customer revenue share within their acquisition channel (FIXED LOD)

xmetric calculated field (Share of Channel Revenue):

SUM([Revenue]) / SUM({ FIXED [Acquisition Channel] : SUM([Revenue]) })

Equivalent SQL query:

SELECT
  acquisition_channel,
  plan_tier,
  SUM(revenue) AS tier_revenue,
  SUM(revenue) / SUM(SUM(revenue)) OVER (PARTITION BY acquisition_channel) AS channel_revenue_share
FROM subscriptions
GROUP BY acquisition_channel, plan_tier;

Year-over-Year growth across any dimension on the sheet (Table Calculation)

xmetric calculated field (YoY Growth):

(SUM([Revenue]) - LOOKUP(SUM([Revenue]), -12)) / ABS(LOOKUP(SUM([Revenue]), -12))

Equivalent SQL query:

WITH monthly AS (
  SELECT
    DATE_TRUNC('month', order_date) AS order_month,
    SUM(revenue) AS monthly_revenue
  FROM orders
  GROUP BY 1
)
SELECT
  order_month,
  monthly_revenue,
  (monthly_revenue - LAG(monthly_revenue, 12) OVER (ORDER BY order_month))
    / NULLIF(ABS(LAG(monthly_revenue, 12) OVER (ORDER BY order_month)), 0) AS yoy_growth
FROM monthly;

Current scope and limits

• Local files today: The xmetric app imports .csv, .xlsx, and .parquet files and opens .twb and .twbx workbooks (not every dashboard will look identical to Tableau), with shelves, Show Me charts, calculated fields, LODs, and table calcs. Sign in to save a workbook online and share a view-only link.
• Live database connections: Direct connections to Postgres and BigQuery are coming.
• Agent commands: Every action in the app is a command. An MCP server, ChatGPT app, and Claude connector are in development.

Spec comparison: xmetric vs Metabase and Apache Superset / Preset