Freshness & Data Quality

Freshness & Data Quality

This page is the analytical view of corpus health: what does the indicator pool look like at this build? It is not operational freshness — Grafana owns "is the pipeline running" and "did the last run succeed." Here we answer "of the indicators we have, how many are still useful, and what shape is the data in?"

Indicator freshness by type

Each bar stacks freshness buckets for one indicator type. expired and no_expiration are read directly off the indicator's expiration timestamp; expiring_le_30d is anything still active but expiring within 30 days; active_gt_30d is everything else.

TLP distribution

Traffic Light Protocol marking across all pulses in the corpus.

Data quality snapshot

One row, snapshotted from the most recent successful run.

dropped = present in raw.indicators but no longer attached to its parent pulse on the latest sync. orphan = indicators whose pulse_id doesn't match any pulse we hold (should be zero — flagged by a dbt test). null_indicator_value likewise should be zero.

Brief input freshness

The Analyst Brief is generated from the brief_input mart. Its generated_at timestamp is the moment the brief was rendered, not necessarily the current build time.