Headless Business Intelligence — Base Case Capital
Summary
BI tools bundle two things that should be separate: (1) defining business metrics and (2) visualizing them in dashboards. By coupling metric definitions to a visualization layer, companies lock their most important business logic inside a single consumption surface. Metrics can’t be accessed by sales tools, product systems, or ML pipelines without re-implementation — creating inconsistency and engineering bottlenecks.
The “headless BI” concept proposes decoupling: define metrics once in a shared semantic layer, then consume them anywhere — dashboards, CRMs, product UIs, alerting systems, ML features. This is the same headless pattern from CMS (content vs. presentation) applied to analytics.
The mental model: metrics are an API, not a dashboard.
Relevance
Central to 01-projects/data-marketplace/index — a headless metrics layer is essentially what a data product marketplace would serve. If metrics are an API, they become tradeable, composable, embeddable.
Connects to 06-reference/2026-04-03-data-products-taxonomy — Weber’s “product metrics” category is exactly what headless BI operationalizes. The metrics layer is the infrastructure that makes metrics a true data product.
For 01-projects/phdata/index, this is a positioning opportunity: help clients decouple their metric definitions from their BI tools, especially during Snowflake migrations or modernization work.
Relevant to 06-reference/concepts/skills-as-building-blocks — defining metrics as reusable, composable building blocks rather than one-off dashboard calculations.
Also see 06-reference/2026-04-03-selling-data-science — headless BI helps solve the “selling” problem by making metrics visible in the tools executives already use, rather than requiring them to open a BI tool.
Open Questions
- Is dbt’s metrics layer the winning implementation of headless BI, or will a standalone product (Cube, Transform) win?
- How does the metrics layer interact with the feature store layer (see 06-reference/2026-04-03-feature-stores-hierarchy)? Are ML features just a special case of metrics?
- For the newsletter (01-projects/newsletter/index), is “metrics as an API” a compelling framing for a post?