HyperBasis treats data quality as a product feature: coverage is audited, staleness is flagged, anomalies are quarantined, and the results are published.
The completeness audit measures each asset's rows inside a rolling 90-day window against the 25,920-point standard (288 five-minute buckets per day). An asset below 95% coverage is flagged INCOMPLETE:
Live deviations carry the timestamp of the TradFi quote that produced them. A quote older than 5 minutes flags the row TRADFI_STALE. The flag is displayed with the timestamp everywhere the price is shown, and stale rows never drive real-time signal logic.
Rows beyond the historical deviation threshold (extreme oracle errors, liquidity flashes) are marked anomaly=1. Statistics consumers exclude them; the audit trail records the marking. Users can request real-only history to exclude flagged bars from analysis.
Where minute-level history is still accumulating toward the 90-day standard, HyperBasis uses a daily close-price z-score as a transitional layer, always labeled with its granularity. The minute statistics replace it automatically as coverage matures.
The aggregate coverage number (tradable universe average) is published live on the HyperBasis homepage and updates automatically as data accumulates — the website never shows a hardcoded projection.