Data Catalogue
A searchable inventory of data assets with their schemas, owners, lineage, quality signals and usage.
The catalogue answers the question that consumes an enormous amount of analyst time in any large estate: does data about X exist, where is it, can I trust it, and may I use it. Without one, the answer is found by asking colleagues, which scales poorly and produces different answers.
The determinant of success is automated harvesting. A catalogue populated by people is accurate briefly and misleading soon after. One that ingests schemas from the platform, lineage from query logs and pipeline definitions, freshness and quality from monitoring, and popularity from access logs, stays current without ongoing effort.
Usage signals turn out to be the most useful metadata and the least anticipated: knowing that a table is queried daily by forty people is a stronger trust signal than any documentation field, and knowing that a table has not been read in a year is what makes decommissioning possible.
Lineage is the second: column-level lineage answers impact analysis before a change and root cause after an incident, and both of those are otherwise multi-day exercises.
The failure mode is the catalogue as compliance artifact — bought, populated, mandated, unused. Adoption follows utility, and the test is whether analysts open it because it is faster than asking someone, which requires search that works and content that is current.