06-reference

data products taxonomy

2026-04-03·article·source: https://ericdataproduct.substack.com/p/data-products-arent-all-the-same·by Eric Weber

Data Products Aren't All the Same: 5 Critical Examples — Eric Weber

Summary

Weber taxonomizes data products into five distinct categories, each requiring different skills and contexts:

  1. Experimentation platform — experimental design, cohort allocation, logging, guardrails metrics, readouts
  2. ML platform — model building, serving, management, updating, monitoring
  3. Product metrics — fewer, more streamlined metrics treated as products (precursor to ML and experimentation)
  4. Product analytics — leveraging data to form hypotheses about user/customer behavior
  5. Data quality — foundational; every upstream product depends on reliable, documented data

The key insight: sequence your data product investments. No one can win in every domain at once. Data quality is the foundation; metrics come next; analytics and ML build on top.

Relevance

Directly relevant to [[01-projects/data-marketplace/index]] — the marketplace needs to decide which categories of data products to serve first. Weber's taxonomy is a lens for segmenting the offering.

The hierarchy (quality → metrics → analytics → ML → experimentation) maps to [[06-reference/2026-03-31-block-hierarchy-to-intelligence]] — you can't build intelligent systems without the foundational layers. Also connects to [[06-reference/2026-04-03-feature-stores-hierarchy]] — feature stores occupy the ML platform layer and have their own internal hierarchy.

For [[01-projects/phdata/index]], this taxonomy helps scope consulting engagements: are we building a data quality product or an ML platform? Different skills, different timelines, different stakeholders.

The "sequence your investments" advice aligns with [[SOUL.md]] principles around focused execution over breadth.

Open Questions