ThreatWatch Analysis Quality
Keyword ranking kept rewarding familiar names. I started measuring novelty, freshness, and source health instead.
Analysis pipeline
Keyword matching was useful for the first briefing, but I saw it reward repeated coverage and familiar names. I began adding novelty by comparing new material with a recent window of processed reporting.
The result I want is simple: repeated summaries lose priority, while genuinely new evidence is easier to find. Source health and freshness also need to be visible so missing data is not mistaken for a quiet threat landscape.
Product rule
ThreatWatch should favour a smaller number of well-supported findings over a larger stream of loosely classified articles. Automated analysis can assist with grouping and triage, but evidence and deterministic checks remain the basis for publication.