Data Discovery: Finding Distant Relatives You Didn't Know About

Opening Scene

A genealogical database search occasionally turns up a third cousin nobody in the immediate family had ever heard of, living two towns over, holding a photograph that fills in a gap the whole family had assumed was permanently lost. Nobody was looking for that specific person; they were just searching, and the connection surfaced on its own.

In Plain English

Data discovery is the search and exploration experience a catalog provides, letting people find relevant datasets they didn’t already know existed, based on keywords, tags, popularity, or similarity to data they already use. It’s distinct from simply looking up a known table by name; discovery is what happens when someone doesn’t know exactly what they’re looking for yet, and a well-designed catalog surfaces it anyway, the way a good search engine surfaces an answer to a question you couldn’t have phrased precisely.

The Old Way

Before catalogs offered genuine discovery capability:

Finding relevant existing data required already knowing it existed and who to ask, which meant most useful data simply never got found by people outside the team that built it.
Duplicate datasets proliferated across an organization because nobody could search broadly enough to notice something similar already existed.
New analysts had no way to explore what data was available beyond whatever their onboarding buddy happened to mention.

Real discovery — search that surfaces the unexpected, not just the already-known — is what a mature catalog adds on top of simple lookup.

What’s Changing (and Why AI Is the Reason)

Catalog search is moving from exact keyword matching toward semantic search, capable of surfacing relevant datasets even when the query doesn’t match the exact table or column name.
This shift complements the self-service ambitions covered later in this series’ article on self-service analytics, since discovery is the first step in any genuine self-service workflow.
Large language models embedded directly into catalog search now let people describe what they need in plain language and get relevant dataset recommendations back, dramatically lowering the skill floor required to find useful data.

The Metaphor, Fully Extended

Finding an Unexpected Relative	Data Discovery Concept
A genealogy search surfacing a cousin nobody knew to look for	A catalog search surfacing a dataset nobody knew existed
Not knowing exactly who or what you’re searching for	Not knowing exactly which table or field you need
A distant connection that fills a real gap in the family story	An unexpected dataset that fills a real gap in an analysis
A search engine built to surface the unexpected, not just the known	A catalog built to surface the unexpected, not just the known

For Beginners: What to Actually Do

Use the catalog’s search function to explore broadly before assuming the data you need doesn’t exist.
Try searching by concept or keyword, not just exact table names, especially in catalogs with semantic search.
Browse popular or highly used datasets occasionally; useful data outside your immediate team often surfaces this way.

For Practitioners and Leaders: The Deeper Layer

Invest in search quality and semantic capability as a first-class catalog feature, not an afterthought bolted onto simple keyword lookup.
Track discovery-driven data reuse as a success metric, since it directly reduces duplicate dataset creation.
Curate and tag high-value datasets deliberately, since good tagging materially improves discoverability beyond what search alone can achieve.

Quick Recap

Data discovery lets people find relevant datasets they didn’t already know existed, not just look up known tables by name.
Without genuine discovery, useful data stays effectively invisible outside the team that built it, driving wasteful duplication.
Catalog search is moving toward semantic, natural-language capability rather than exact keyword matching.
Embedded AI search lowers the skill floor required to find relevant data using plain-language queries.

Where This Fits in the Series

Article 10 covered agreeing on shared vocabulary across the organization. This article covers finding data using that vocabulary, including data nobody knew to look for. Article 12 turns to a more formal use of the same underlying lineage and metadata: proving, for compliance purposes, exactly where a record came from.