Weighted rotation vs auto-optimization — when to let the tracker decide

Weighted rotation vs auto-optimization — when to let the tracker decide

Your tracker can split traffic two ways. Manual weights or algorithmic. Pick by data volume, not preference.

Use weighted rotation (manual %) when:
— Daily conversions under ~50 — too thin for an algorithm to learn.
— You're testing a known hypothesis (60/40 split, controlled).
— You need predictable distribution for a brand or capped offer.

Use auto-optimization (AI/RON) when:
— Conversions per path exceed ~100/day — statistically feedable.
— You have 4+ paths and can't babysit weights hourly.
— The optimizer can act on revenue, not just CR.

Tradeoffs to record:
— Auto-opt exploits the current winner but stops exploring — it can lock onto a lucky early lead.
— Manual weights never overfit but they never improve on their own either.
— Hybrid: hold 10-15% on forced even rotation as a control arm even when auto-opt runs the rest.

Rule: thin data → manual. Thick data → auto with a control arm. Never auto-optimize on