How the World Cup of Coffee works

Metrics and weights

Each metric is min-max normalized to 0 to 100 across the nations that have it. The Coffee World Ranking is the weighted average of a nation's available metrics.

Judgment calls, in the open

• Missing data is excluded, never estimated. A nation missing a metric is scored on the metrics it does have, with weights renormalized. Nobody gets an invented number. Cabo Verde, Curacao, Iraq, Qatar, and Uzbekistan lack consumption data in both USDA and ICO sources; the dataset's metrics_present column shows exactly who is scored on what.
• Consumption years are mixed, openly. USDA tracks most nations currently (market year 2025) but reports the EU as one block, so EU member teams use the ICO's per-country workbooks, which are public only through 2019. Each dataset row states its year and source. National coffee habits move slowly, and we would rather use a real 2019 number than a fabricated 2025 one.
• England and Scotland share United Kingdom values. The statistical world reports the UK, not the home nations, so both teams carry UK figures on every metric.
• Production zeros are real zeros. A nation absent from both USDA's and the ICO's production tables has no commercially tracked coffee production.
• OSM cafe counts partly reflect mapping density. OpenStreetMap tagging is far more complete in Europe and Oceania than in South America, Africa, and parts of Asia. Brazil records fewer tagged cafes than Austria, which says more about mappers than about cafes. We keep the metric because it is the only open, uniform, per-country cafe count in existence, weight it moderately, and say this plainly. Treat it as mapped cafe density.
• Numbeo is crowd-sourced. It is the only free source with near-complete coverage of these 48 nations. We use the national cappuccino price in USD, adjusted by the World Bank price level ratio so the score reflects what a cappuccino costs relative to everything else in that country. Iran is excluded from this metric: its managed exchange rate makes the adjusted price an artifact rather than an affordability fact.
• A fifth metric was dropped, not faked. The plan included Google Trends interest. Google offers no official API and blocks the unofficial one, so the metric was dropped and the remaining four weights renormalize. We would rather lose a metric than invent one.
• Production is log-scaled. Brazil grows more coffee than the other 47 nations combined. On a linear scale every other producer would round to zero, so production is log-scaled before normalizing.
• Real results override coffee, and we keep score on ourselves. The bracket is the real one. Fixtures played before our latest update advance the real winner, and each matchup shows whether coffee called it right. Fixtures not yet played advance the higher Coffee World Ranking, labeled as simulated.