Business Problem & Solution

The coupling that makes the ecosystem feel like one product — every app reading one booking, one guest, one review — is also its biggest risk: a silent data-quality regression in a shared entity is everyone's bug at once. A missing geo on properties breaks search; stale reviews skew the KPIs; a broken payment reference corrupts FinOps.

Data Quality & Lineage Monitor keeps the shared layer trustworthy. It runs declarative DQ rules across the entities every app reads, trends the results, and draws the coupling between entities, apps and events so a failure can be traced to where it started rather than where it surfaced.

It proves that data quality is a measurable, dimensioned discipline: each rule names an entity, a field, a dimension (completeness, validity, uniqueness, freshness, consistency), an expression, and a pass threshold, and one rule deliberately degrades over the window so the failing-trend path is real. Lineage is shown as a concrete entity↔app↔event graph, not prose.

It simplifies execution. Checks are computed deterministically over the synthetic dataset rather than run against a live warehouse, and the AI-assist root-cause step is a simulated, metered hypothesis. The rules, scores and lineage cost nothing — only the explanation is metered.

Synthetic checks only — eight rules over a 14-day window across shared-core entities. No row-level access and no PII (the monitor reads metadata and scores, not records). Figures are representative and labelled; the explanation is a hypothesis.