How to read a figure
four kinds of number- OBSERVED
Reported by a named public source and shown with the period it covers. If the source could not be reached, the figure is hidden rather than estimated.
- MODEL OUTPUT
Computed by GlobalMatrix from observed inputs using a published formula. Deterministic and re-derivable, but it is our arithmetic, not a measurement.
- SCENARIO
A synthetic, internally-consistent scenario set. Nothing in it is a claim about the world, and it is labelled wherever it appears.
- DATA UNAVAILABLE
No verified reading exists for this field right now. The gap is shown rather than filled with an estimate, a carry-forward, or an interpolation.
Not connected
stated rather than hidden- IMFUNAVAILABLE
The IMF's open-data edge rejects requests that do not carry a contact URL, and we will not send a domain we do not control. Balance-of-payments and fiscal positions are therefore not measured, and no figure on this site is drawn from them.
- Equity and commodity pricesUNAVAILABLE
No licensed price feed is connected. The ECB publishes official euro reference rates and the euro area yield curve instead; those are central-bank reference rates, not market quotes.
- Company filingsUNAVAILABLE
No structured filing feed is connected, so no company-level claim is made anywhere on the site.
- Bilateral trade corridorsUNAVAILABLE
UN Comtrade's public preview tier reports reporter totals, not a full corridor matrix. Only totals are shown.
- EIA, IEA and OPEC energy statisticsUNAVAILABLE
Each publishes under its own access terms — a key, a licence or a document format this build cannot consume without one. Energy figures on this site are model output from the corpus, never a reported barrel price.
- Google Trends search interestUNAVAILABLE
There is no officially published Trends API to read from. It is an interest signal rather than a statistic in any case, so it is deliberately absent rather than approximated from media coverage.
- Policy lifecycle registryUNAVAILABLE
Policy activity is surfaced through the event corpus rather than a dated legislative tracker.
Each source is polled on the schedule its publisher actually keeps — weekly for the annual datasets, every 30 minutes for GDELT — and a source that fails is backed off instead of retried in a loop. Asking now is a request, not a command: the server refuses it if a fetch is already running or the source is still in backoff.