One number for how loudly the world's press is talking about India's geopolitics. Updated every morning. Every number backed by the articles behind it.
IGRM measures press salience: the share of global news coverage matching five registered channels of India-relevant geopolitics — the Pakistan border, the China border, Gulf & energy security, US & trade policy, and shipping chokepoints — each ranked against its own trailing two years. It is not a measure of risk, it makes no forecasts, and it is not investment advice. What it offers instead is discipline: every construction choice registered before use, every claim checkable, every miss recorded in public.
Composite today –
Read it in ninety seconds
- The headline number is the composite: the mean of five channel scores, each a 0–100 percentile of today's coverage share against that channel's own last two years. 70 means louder than 70% of days in that window. That's all it means.
- The five channels decompose it. A quiet composite can hide one loud channel; the channel rows and their sparklines show which one.
- The receipts are the evidence: for each channel, the exact registered query and the matched articles — most of them drawn from the very sample the day's score was computed over, tier-sorted by source credibility.
- The validation record is how you know it measures anything: detection of a pre-registered episode list the query terms never name, placebo channels that must stay quiet, robustness under alternative dictionaries, and the frozen 2026-06-30 through 2026-08-07 sampling-band window.
Choose your door
Policy practitioner
Today's picture and what moved: the front page, then receipts for the loud channel, then the maps.
Business owner
Pick your sectors on the exposure page and see which channels touch them, today's readings with bands, and the articles behind each — salience, not risk; computation shown; profile in the URL, no accounts.
Researcher
Constructions and reuse: methodology, codebook with a five-line quick-start, and the frozen API contract. For backtests, the point-in-time vintages panel reconstructs what the series showed on each publication morning. Data CC BY 4.0.
Why this is an instrument, not a dashboard
- Registered dictionaries. The query terms are versioned with per-term rationales and an append-only changelog; they never chase events after the fact.
- Pre-registered validation. The episode list was frozen before the first test; hit and miss both publish. Current record: on the validation page, misses included.
- Public reliability record. The morning contract is scored from commit timestamps, not self-reporting; the status page shows it live beside every data source's freshness, and misses stay listed forever.
- A standing invitation to break it. Six ways to falsify the instrument, the tools to try each one, and the promise that serious submissions publish here unedited — with a permanent corrections entry if you succeed.
- A corrections ledger. Every error caught, dated, append-only.
- Frozen machine interface. Every endpoint frozen with promised fields, no key, no rate limit, open CORS.
What this cannot tell you
Whether markets will move, whether a crisis will escalate, or what a government intends. Salience is attention, and attention is not risk: the attention-pricing gap study documents exactly where press attention and market pricing diverge, as a finding, not a signal.