Methodology

PandemicRadar aggregates five families of public, freely-available signals into a single composite score between 0 and 100 for the world and for twelve tracked countries. The score answers one question: “Right now, how loud is the early-warning chatter that could precede new lockdown-grade restrictions?”

The composite

Risk =  0.30 · Epi
      + 0.20 · Media
      + 0.15 · Search
      + 0.20 · Policy
      + 0.15 · Market

Each family is normalised against its own 90-day baseline (z-score, then sigmoid-mapped to 0–100). This means we measure anomaly, not absolute level: a 30% spike in search interest matters even when nothing is technically wrong, because it usually precedes the moment when something is wrong.

Families

Risk bands

• LOW (0–24) — baseline noise, no actionable signal.
• ELEVATED (25–49) — something is moving in one or two families. Worth watching.
• HIGH (50–74) — multiple families converging. New restrictions are plausible in the next 4–8 weeks.
• CRITICAL (75–100) — broad-based alignment of signals. Lockdown-grade measures are imminent or already in force somewhere.

What we learned from 2020

• Search interest (Google Trends, Wikipedia) spiked 3–6 weeks before formal restrictions in most countries.
• Media tone (GDELT) preceded official advisories by 2–4 weeks.
• The WHO is, structurally, a lagging indicator. We weight it, but we do not wait for it.
• Wastewater surveillance is the single hardest-to-game signal. It is on the v1.0 roadmap.
• Equity markets in PPE and travel moved days to weeks ahead of restrictions.

What this is not