How registry change detection works

Re-fetch and compare

Most official registers offer no subscription or webhook mechanism - there's no way to be told the instant something changes. Change detection instead means periodically re-fetching an entity's current state and comparing it to what was last recorded, entirely on the monitoring system's own schedule.

Why a naive text diff isn't enough

Comparing two raw register responses character by character produces noise - formatting differences, field reordering, or a temporarily incomplete response can all look like a change when nothing meaningful actually moved. A useful comparison works at the level of specific, named facts (an ownership percentage, a registration status, a filing date), not raw text.

Classifying what changed, not just flagging that something did

A change in registered address and a change in beneficial ownership are both "a difference," but they matter very differently to a compliance reviewer. Useful change detection classifies what kind of change occurred, rather than surfacing an undifferentiated list of diffs someone then has to triage by hand. See how Keizu monitors for change for the actual categories used.

Handling ambiguous cases honestly

Not every difference is cleanly classifiable - a register temporarily returning incomplete data can look like a real change without being one. A system that forces every difference into a clean category will misclassify these; one that keeps a distinct "needs a second look" category is more honest about what it actually knows.