Methods index
Data, with its assumptions attached.
Four connected areas guide this publication. Notes make assumptions explicit and separate proposed checks from observed results.
Area 01
Models & Grain
Studies define table grain, keys, versioned records and the difference between an entity and an event. Join cardinality is checked before interpreting totals.
- Row identity
- Business keys
- Join cardinality
- Additive measures
Area 02
Data Contracts
Quality checks state their population, null policy, uniqueness key, expected reference coverage and failure behavior. A check becomes useful when its result changes a pipeline decision.
- Null semantics
- Uniqueness
- Reference coverage
- Quarantine rules
Area 03
Replayable Pipelines
Methods separate transport delivery from business identity. Notes examine ingestion manifests, transaction boundaries, conflict rules, partial failure and corrected input.
- Batch manifests
- Idempotent loading
- Transactions
- Late corrections
Area 04
Reconciliation & Semantics
Examples reconcile coverage and values separately, preserve missing-side records and attach time-zone and decimal policies to calculations. Differences remain explainable at the row level.
- Matched populations
- Signed deltas
- Time windows
- Decimal policies