Data Governance · Break-in-series Register
Every long-running data series keeps its name while the regime underneath it changes: a threshold moves, a population is redrawn, an instrument is replaced. Those joins are invisible in the data itself. They live in methodology notes, change tickets and the memory of people who have since left, which is why a chart can look continuous while measuring two different things. This register writes them down and makes them computable.
An event records what changed and why; an impact records what that change did to one specific series, in which direction, and whether a correction exists. Comparability is then derived from impacts rather than events, which is what lets the register answer the question a model owner or an auditor actually has: for this particular use, how far back does the data support the claim, and which change decides it.
Panel 01 · The breaks
Panel 02 · Reconciliation
Assessment states are explicit, so not found is distinguishable from not assessed. Select a card to filter the register to that category.
Panel 03 · The register
An event says what happened. Its impacts say what it did, to which series, in which direction. Search covers titles, identifiers, series, mechanisms, remediation, evidence and adjudication notes.
Panel 04 · Comparability
Rows are series families, columns are decades. Cells show the worst raw verdict among impacts affecting that family from that decade forward, plus any contaminating impact already in force.
Panel 05 · Taxonomy
This is the controlled vocabulary the register is written in, and it is the reference to reach for when writing or reviewing an entry. Every term carries its definition, and every list is read from the loaded file rather than compiled into the application, so a register with its own layers or mechanisms brings its own definitions with it and updates the filters, lanes, legends and validation without a code change. The same definitions appear on hover wherever a term is used elsewhere on this page.
Panel 06 · Your data
series with your own series records, and uses with the models, metrics and reports that consume them. A use lists series identifiers.events record per change, then one impacts record for each series that change actually affected. An impact carries the verdict, direction and bridge state; the event carries only history.Appendix A · Clean windows
For every downstream use, the raw window is the span the originally published data supports with no intervention. The adjusted window is what becomes available once bridges that exist are applied. Every boundary names the impact that set it; select one to read the record behind it.
| Downstream use | Scope | Raw window | Adjusted window | Impacts |
|---|
A window is computed, not recorded. Nothing in the register states these dates; they fall out of the impacts that touch each use.