Is research data loss architectural or operational?

The empirical record on research data preservation documents a steady-state outcome: 73 to 93 percent of published research carries underlying data that cannot be produced on request, with that fraction stable across two decades, multiple disciplines, and successive funder regimes. The standard interpretation treats this as an operational failure — insufficient researcher discipline, inadequate data management plans, underfunded libraries, uneven training — and addresses it through procedural reform.

This question asks whether that interpretation is correct. Are the documented losses produced by behaviors that procedural reform can fix, or by structural properties of the storage architecture that procedural reform cannot reach?

The answer determines what counts as a credible response. If the cause is operational, the policy stack already in place — stronger plans, mandatory deposit, funder enforcement — is the right approach and needs only better execution. If the cause is architectural, no amount of policy reform changes an outcome that the underlying storage substrate determines.

Subsidiary questions within Q-0001:

• Do operational explanations close the documented compliance gap?
• Do procedural mandates produce different outcomes on different architectures?
• Is the survival of a dataset a property of researcher behavior or of how many independent copies exist across independent failure domains?