Probabilistic forecasting in production across 43 countries.

Early warning of food crises means combining many weak, noisy signals into something people can act on, with the uncertainty stated honestly rather than hidden. Getting it wrong in either direction costs lives or credibility.

CERES builds probabilistic forecasts on open data and publishes them reproducibly. The uncertainty stays in the result; the pipeline can be re-run and checked by anyone.

Forecast data runs in production on OCHA's Humanitarian Data Exchange, the UN's humanitarian data platform, across 43 countries.

The same discipline applies on a plant floor: combine weak indicators into one decision, keep the result auditable, and never present a guess as a number. If it is robust enough for humanitarian response, it is robust enough for your Monday morning meeting.