Data Quality Service  ·  View 15 of 20  ·  Runtime

Assertion Lifecycle

The loop that keeps rules credible: priced, backfilled, on probation, enforcing, and demoted when they cry wolf.

Editable source SVG draw.io All views
Authored DSL in Git Priced dry-run Backfilled 30-day lookback Probation advise only Enforcing quarantine or block Demoted FP > 5% Assertion versioned bounded plan re-judged owner review FP under 5% cries wolf reworked Assertion Lifecycle — The Loop That Keeps Rules Credible Application we own Journey / task Opportunity Risk / gap Demotion is automatic and retirement is not: a rule that cries wolf is silenced by the platform, but only a person may decide a rule is no longer worth having — and its verdicts survive either way. v 1.0 · owner Data Platform Architecture · date 2026-09

Why it is a loop

  • Demotion is automatic; retirement is not. The platform silences a rule above a 5% false-positive rate without waiting for anyone, but only a person may decide a rule is no longer worth having (ADR-12).
  • A retired assertion keeps every verdict it produced. Retirement stops future evaluation; it does not rewrite history (ADR-01).
  • Pricing sits between authoring and history so that an expensive rule is refused, or negotiated, before it has consumed anything (ADR-08).

Assumptions

  • 30-day probation, ≤ 5% false-positive rate over a trailing 30 days as the demotion threshold. Stated assumptions.
  • Distributional assertions never leave probation automatically — they require an explicit promotion after their false-positive rate has been measured.

Risks

  • The loop can stall in 'Demoted': a rule that was demoted and never reworked is a coverage gap wearing the appearance of coverage, which is why demoted assertions are debt items (ADR-16).
◀  Retroactive Evaluation All views Deployment Architecture  ▶
Solution Architecture v1.0 · Databricks Lakehouse on AWS · Data Platform Architecture · 2026-09 · 20 views · 16 architecture decision records15-assertion-lifecycle