Conditional policy

A mapping from observable states to different actions or target levels instead of one fixed reference for all conditions.

Marketing for your course is a switch you only know how to leave on. When demand surges, your support inbox drowns and quality slips; when you pause ads out of guilt, revenue craters and panic flips them back on. One fixed target, "grow enrollment," is producing this whiplash, because the right action genuinely depends on conditions that keep changing.

A conditional policy is the reference shape for that dependence: a mapping from observable states to different actions or target levels, instead of one fixed reference for all conditions. Use it when conditions change materially, you can observe them, and feedback arrives before irreversible harm. Build it as a small table: when delivery quality and support capacity are green, scale promotion; when support load enters amber, slow demand and improve delivery; when the truth-and-quality floor fails, stop promotion entirely. Add a persistence rule so one bad day doesn't trigger a mode switch. What changes: the exhausting judgment call "should we push growth right now?" becomes a lookup, and the policy makes the tradeoff you'd want made when you're too busy to think. The warnings: policies become opaque, twitchy, or gameable, so use few states, reliable signals, guardrails, and keep a human override for anomalies the table never imagined.