A direct, verified observation of the real world, used as the reference against which models and estimates are checked.
Ground truth is data that reflects what actually happened, observed as directly as possible rather than inferred. In machine learning it is the reference used to train and validate a model; in markets it is the real-world activity behind a reported number. The phrase carries a standard - it is the thing other estimates are measured against, so it has to be trustworthy on its own terms.
Observing ground truth where an event occurs - overhead imagery of activity, direct observation of the web, a sensor reading - is what makes a dataset worth trading or training on. The moment you rely on a proxy for a proxy, error compounds, so the discipline is to get as close to the original observation as possible and to preserve the evidence of how it was observed.
A live evaluation measures EdgeOrigin against your coverage requirements: decision-ready real-time data, delivery latency into your systems, and record-level provenance.