xmetric vs Evidence and Observable Framework: Interactive Shelves vs BI-as-Code

Visual shelf exploration vs writing Markdown and JavaScript for every chart

Evidence (evidence.dev) and Observable Framework (observablehq.com/framework) are "BI-as-code" static site generators. To create a chart in Evidence or Observable Framework, a developer installs the Evidence CLI binary (or Node.js 18+ for Observable Framework), initializes a git repository, writes a SQL query or data loader, and codes {% bar_chart data="query" x="month" y="sales" /%} tags in Markdown (Evidence) or Plot.plot({...}) in JavaScript (Observable Framework).

Code-driven static reports work well for polished, version-controlled narrative pages, but they slow down ad-hoc data discovery:

1. You cannot pivot or change grain without editing code: If a reader viewing an Evidence or Observable report wants to split a revenue chart by product tier instead of region, or filter out an outlier cohort, they cannot drag a field onto a shelf. A developer has to edit the SQL query and component tag and re-run the build.
2. No Node.js or CLI required in xmetric: Non-engineers (or engineers in a hurry) can open the xmetric app in any browser tab, drop a .csv, .xlsx, .parquet or .twbx file, and try many visual cuts on shelves. Formulas can be tested in the formula playground.
3. Agent-native commands without static rebuilds: Every shelf action and calculation in xmetric is a command, so views change without waiting for a static-site bundler. An MCP server, ChatGPT app and Claude connector, so coding agents can drive those commands, are coming.

Comparing the same chart: xmetric shelves + LOD vs Evidence/Observable SQL

Suppose you want to plot monthly revenue colored by whether each customer is in their first purchase month or a repeat month.

In xmetric:

1. Create one calculated field Cohort Type (you can also test it in the formula playground):

IF DATETRUNC('month', [Order Date]) = DATETRUNC('month', { FIXED [Customer ID] : MIN([Order Date]) })
THEN "First Month"
ELSE "Repeat Month"
END

1. Drag MONTH(Order Date) to Columns, SUM(Revenue) to Rows, and Cohort Type to Color.

In Evidence / Observable Framework (SQL query + code component):

-- orders_by_cohort.sql
WITH customer_first AS (
  SELECT
    customer_id,
    DATE_TRUNC('month', MIN(order_date)) AS first_month
  FROM orders
  GROUP BY customer_id
)
SELECT
  DATE_TRUNC('month', o.order_date) AS order_month,
  CASE
    WHEN DATE_TRUNC('month', o.order_date) = c.first_month THEN 'First Month'
    ELSE 'Repeat Month'
  END AS cohort_type,
  SUM(o.revenue) AS revenue
FROM orders o
JOIN customer_first c ON o.customer_id = c.customer_id
GROUP BY 1, 2
ORDER BY 1;

Spec comparison: xmetric vs Evidence and Observable Framework

Pricing and limits below are current (as of October 2026).