Evidence becomes decisions, decisions become governed actions, and outcomes calibrate what Nodes learns next · 4 waiting on a human ruling · Illustrative — synthetic demo data
The pattern matches the engagement-and-comp signature that preceded last year's regretted departures, and the intervention window is about six weeks.
Off-curve hires in the first 90 days are the strongest early signal of both slower productivity and later attrition.
Validated 90-day accuracy decayed this quarter; the signals that made the persona predictive are weakening.
Internal fills avoid external search cost and are the strongest retention lever for exactly this population.
Nothing executes until a person approves it. Two of the waiting decisions are inside their intervention window.
Outcome windows are how Nodes learns: realized value calibrates the next recommendation, so the windows must close before value is claimed.
This is the loop closing. A belief (the keyword screen protected quality) was disproven by evidence, ruled on by a human, and the outcome confirmed the ruling.
Every claim above draws only on sources a human approved · next action: Enable Slack reads to close the engagement-signal gap in the retention pattern.