Learning vs. active

A brand-new flow hasn't earned trust yet. Learning is the false-positive defense that keeps it quiet until it has.

Why a learning period exists

The moment a flow is set up, FlowProofs has no baseline for it yet: no sense of the normal volume, no confirmed matching pattern, nothing to compare a "missing" record against. Alerting immediately would mean alerting on noise: a slightly slow first test record, an edge case in the first few real ones. Learning exists so a flow can accumulate a real pattern before anything is allowed to page you about it.

The three states

• Learning: collecting and matching pings for a baseline, but alerts never fire, no matter what happens.
• Active: fully monitored. The same matching runs, but now confirmed gaps and completeness flags can actually trigger alerts.
• Paused: monitoring suspended entirely, on purpose (for example while you know a client's system is being rebuilt). No matching, no alerts, until unpaused.

Graduating from learning to active

This happens automatically, the moment either of two things is true: the flow has seen at least 10 matched pairs, or 24 hours have passed since both sides first connected, whichever comes first. Enough volume, or enough time, to trust the baseline is honest, whichever arrives sooner.