Question
PUBLIC DATA · OBSERVABILITY
When does a quiet pipeline stop telling the truth?
A real ECB series, an explicit consumer contract, and a monitor that keeps structural failures separate from unusual—but potentially valid—data.
Evidence
Official ECB SDMX data, fixed identity fields, TARGET-day freshness, and synthetic mutation tests.
Finding
Awaiting the first complete scheduled observation.
Boundary
A review signal is not evidence that an exchange rate—or its publisher—is wrong.
V1 EVIDENCE MAP
Stable software. Unequal evidence.
The counters remain separate so synthetic validation cannot be mistaken for time-earned observation.
- Product version
- 1.0.0
- Prospective real runs
- 2
- Retrospective prefixes
- 7,010
- Synthetic fault cases
- 9 / 9 detected
The retrospective replay uses one current historical vintage. It cannot reveal revisions that happened between past publications. The available prospective runs do not yet constitute longitudinal evidence.
LATEST RECORDED RUN
One status. Two meanings.
Contract failures describe broken expectations. Statistical signals request human inspection. They never share a label.
ContractLoading
Statistical reviewLoading
- Latest observation
- —
- Rows inspected
- —
- Checked
- —
Loading the machine-readable report…
INTERACTIVE VALIDATION
Inject a known failure.
Choose one controlled mutation. The lab shows which layer should react and what can legitimately be concluded.
EXPECTED CLASSIFICATION
PASS · CLEARThe response satisfies the hard contract and raises no statistical review signal.
Permitted claimThis controlled response satisfies protocol version 1.0.0.
EPISTEMIC BOUNDARY
Observability is not omniscience.
This system can identify a response that violates a consumer contract. It cannot inspect the ECB's unpublished upstream inputs, establish economic correctness, forecast exchange rates, or infer that a statistically unusual movement is erroneous.