Skip to content
trendoviz

Coverage and limitations

What we observe, what we can miss, and how far our numbers can be trusted.

Methodology 2.1 · updated 2026-11-21

What we observe

  • Trendoviz observes configured, authorized sources (RSS feeds and GitHub releases today), not the entire world.
  • Absence from Trendoviz does not prove that something is unimportant.
  • Attention means coverage in the feeds we collect, not total worldwide discussion; we do not collect social networks.
  • Our sources are mostly in English and favour some regions, platforms and organisations over others.
  • RSS and GitHub coverage favours open-source and digitally mature projects; closed products and offline institutions are under-represented.
  • We do not yet collect a vulnerability enrichment feed (CISA KEV, EPSS) or regulators' registers directly: those indicators appear only when a source we collect states them.

What our assessments mean

  • Evidence confidence describes the evidence we retrieved, not absolute truth.
  • Importance is not popularity: a critical trend can have almost no coverage.
  • Trend strength (the Trendoviz Score) is not real-world impact: it measures how broad, persistent and fast-growing the observed activity is.
  • Low observed substance may reflect missing coverage rather than an absence of real work.
  • Early classifications may change as evidence develops. Provisional trends and scores are labeled; history and corrections are kept.

Validation status

Weights and thresholds are explicit, versioned design choices. They are not scientifically validated: we have not yet collected enough labeled history to calibrate them. A validation framework measures grouping precision and recall, claim and citation accuracy, confidence calibration, lifecycle stability, ranking sensitivity to weights and time to publication on labeled fixtures; results are published here when labeled data exist. Today the rule stage of grouping is evaluated on a small labeled fixture set, which is not a measure of production accuracy.

Questions you can answer from each page

  • Why was a signal grouped here? The grouping method and reason are stored with each signal and shown to editors; the trend page lists the evidence.
  • Why is a source independent or not? The trend page shows signals, feeds, publishers, ownership groups and distinct underlying evidence.
  • Which claims are facts, source statements, hypotheses, tests or inferences? Each claim carries its label and citations.
  • Why this importance, score or stage? Trend pages list the importance dimensions and qualifying rule, the score components, unavailable metrics and lifecycle history with rule versions.
  • What changed and what triggered an alert? The classification history and each alert say so.
  • Could payment change any of this? No: see No paid placements.