This is the structure I deploy on almost every retainer. It is unglamorous on purpose — the point is that it still works when spend is ten times higher.
The skeleton
- One research campaign per product group — broad and phrase, low bids, its only job is finding new terms.
- One exact campaign per hero SKU — this is where budget lives.
- One product-attribute campaign — catches the long tail cheaply.
- A negative sculpting layer — research terms promoted to exact get negated in research.
The bid ladder
SKU-level target ACOS sets the ceiling. Bids start below break-even on new terms, step up once conversion is proven, and step down the moment a term exceeds its target for two consecutive weeks. No term gets more budget before it earns it.
Three mistakes that break it
- Harvesting too slowly. Terms sitting in research for months are just funding Amazon.
- Harvesting too fast. Two conversions is not a proven term.
- Skipping the negative layer. Without it, your exact campaign competes with your own research campaign and you bid against yourself.
