Dark data

Data an organization collects, generates, or stores that is never actively used, analyzed, or, most relevantly for this topic, monitored by anyone with clear accountability for it. Industry estimates of its share of total enterprise data vary widely by study and methodology; several studies over the past several years have suggested it commonly makes up a large share, in some cases a majority, of what organizations store, though any single percentage should be treated as illustrative rather than a fixed benchmark. A first-pass estate survey typically undercounts it because un-owned data is disproportionately data nobody is looking at.

Data that exists somewhere in an organization but is not on any current inventory, such as a forgotten export, an old shared-drive file, or a copy piped into an unapproved third-party tool. It cannot be scheduled for destruction until it is found, which is why an audit is a floor, not a ceiling.

A data source (an old warehouse table, a cached export, a spreadsheet a departed employee built) that feeds a pipeline without ever appearing in a formal source inventory or lineage map, because no one thought to register it. A chapter drafted only from named, registered sources will miss dark data by construction; the defensible response is a stated discovery method and date, not a claim that no dark data exists.