Data Sources and Methodology (draft)

Before publishing

• Not built. No source is integrated yet; the pipeline is Planned (see Data Pipeline). Publish only the sources that are actually live, each with its real update frequency.
• Confirm the licence and attribution wording for every listed source (see Minimum Viable Data Set). OpenStreetMap (ODbL), OpenSky and ACLED have commercial conditions.
• Never list Google Maps, Yandex or Baidu as sources unless a commercial licence allowing redistribution of derived data is signed.

Demostatics collects information from public and licensed sources, processes it into consistent statistics, and publishes the results in the Database and Reports. This page explains where the numbers come from, so you can judge how much weight to give them.

Principles

• Every value has a source. Each figure records where it came from and when it was collected.
• Originals are kept. Raw data is stored as received before any processing, so results can be re-checked and re-computed.
• Machine-processed, human-checked. Automated systems, including machine-learning and language models, classify and structure information. Editorial staff review what is published as reports.
• Limits are stated. Coverage gaps, delays and known weaknesses are listed below rather than hidden.

Sources

Area	Example sources	Typical update frequency	Attribution
Economic and trade indicators	World Bank, IMF, UN Comtrade, FAO, EIA, Our World in Data	Monthly to yearly, as published	TODO per publisher
Exchange rates	Central bank reference rates (e.g. TCMB, ECB)	Daily	TODO
News and events	GDELT	About every 15 minutes	TODO
Conflict and political events	ACLED (commercial licence required)	Weekly	TODO
Environment and weather	Copernicus Sentinel, USGS Landsat, ERA5 / NOAA	Hours to days	TODO
Infrastructure and transport	OpenStreetMap, OpenSky Network, AIS feeds	Minutes to days	© OpenStreetMap contributors (ODbL) — TODO others

TODO: replace “Example sources” with only the sources that are live, and keep this table in sync with the internal source registry.

How data becomes a published figure

1. Collection — data is fetched on a schedule and stored unchanged.
2. Processing — it is cleaned, de-duplicated, classified and linked to places and categories.
3. Structuring — values are organised along five dimensions: time, capital, workforce, material–energy, and information.
4. Publication — values appear in the Database; editors write and publish Reports.

Geography and categories

Places follow a single hierarchy — region, subregion, country, state/province, city — based on TODO: name the geographic dataset and its licence. Categories and subcategories are defined by Demostatics; TODO: link the public category list.

Known limits

• Delays. A figure is only as fresh as its source; official statistics can lag by months.
• Coverage. Some countries and cities publish little data; empty filters mean no data, not zero.
• Revisions. Sources revise their figures; Demostatics updates values when they do.
• Automated classification can be wrong. Report errors to the team via the contact page.

Corrections

TODO: how users report an error, and how corrections are published (e.g. a correction note on the report and an entry in the Changelog).