run_lever_tick
Recompute daily outbound-optimization scores: refresh message/strategy weights, archive underperforming variants, and update channel affinity and targeting lifts to keep outreach effective.
Instructions
Run the project's daily outbound-optimization tick: recompute the message-variant draw weights pick_message_variant reads (Thompson sampling over graded reward; archives variants whose P(best) stays below the threshold at maturity, never below two active), the discovery-strategy draw weights over the active registry (same Thompson math; archives dominated strategies, never below two active), the per-industry channel affinity get_outbound_targets surfaces, and the measured targeting lifts (industry / size / country / discovery strategy / fresh signal) that re-score the get_outbound_targets ordering. After a sustained flat streak (every arm mature yet none likely best) it rotates out the weakest variant to free a slot for a fresh angle. Idempotent per UTC day — a repeat call returns that day's recorded decision without re-applying. Returns variant weights and archived variants (a stagnation rotation is marked as such), strategy weights and archived strategies, channelAffinity by industry bucket, targeting lifts, the futility vitals verdict (ok / insufficient / futile over recent mature email sends — futile means outreach is statistically drawing no replies and belongs in the cycle report; it clears on its own once recent sends draw replies again), and the live needsReplenishment / needsStrategyReplenishment flags (recomputed each call, not the frozen recorded value).
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| projectId | Yes | Project name or ID |