Full-funnel Amazon advertising architecture

Most Amazon ad accounts are not architectures, they are archaeologies: campaigns layered by different hands at different times, optimised to a single blended number. A full-funnel architecture is different: every campaign has one job, a budget rationale and a metric that matches the job. This guide sets out that structure.

The funnel, translated to Amazon

Four jobs cover the channel. Capture the demand that already exists (shoppers searching for what you sell). Own your shelf (the branded search results and your product pages). Acquire new demand (shoppers in the category who do not yet know you). Retain what you won (repeat purchase and subscription). Each job maps to specific ad products, and each fails if measured by another job’s metric.

Layer 1: demand capture with Sponsored Products

Sponsored Products carries most budgets because it sits closest to purchase. Structure it so performance is legible: separate branded terms from generic terms from competitor terms, because their economics are incomparable. Run exact-match campaigns for terms you know convert, with research campaigns feeding them: mine the search term reports weekly, promote winners to exact, and add negatives relentlessly so research spend stays research spend.

Layer 2: owning the shelf with Sponsored Brands

Sponsored Brands banners do two jobs. On your branded terms they are defence: occupying the top of your own search results so a competitor cannot rent it, at cheap clicks and high conversion. On category terms they are preference-shaping: your range and story above the results, landing on your Brand Store rather than a single listing. Scale defence with the aggression you actually face, measured by impression share on your brand terms.

Layer 3: audiences with Sponsored Display and DSP

Above the search bar sit the audience tools. Sponsored Display re-engages shoppers who viewed your products or your competitors’, and reaches in-market category audiences. DSP extends the same first-party data programmatically across and beyond Amazon, with the minimums and management overhead that implies. The honest sequencing: exhaust search first. Audience spend earns its place when incremental search spend stops producing incremental customers, typically during launches, category pushes or once brand terms are saturated.

Retention: the quiet layer

Retention on Amazon is less an ad product than a system: Subscribe & Save enrolment on replenishable lines, remarketing to past buyers through Display where the maths supports it, and range architecture that gives a satisfied customer somewhere to go next. For healthcare brands with genuinely replenishable products, this layer often has the best economics in the account, because the margin funds a discount instead of a click auction.

Budgets and measurement per layer

The architecture only holds if each layer is judged by its own number:

• Demand capture: ACoS against a target set per term class, branded never blended with generic.
• Shelf defence: impression share on brand terms, not return on ad spend.
• Acquisition: cost per new-to-brand customer, judged against lifetime value, not first order.
• Retention: subscriber growth and repeat rate.
• The account overall: TACoS trending down as organic sales compound.

Budget flows from job priority: fund capture to its efficient ceiling, defence to the level threat requires, then push acquisition as far as your new-to-brand economics allow. Review the architecture quarterly; re-optimising a wrong structure weekly is how accounts calcify.

Go deeper

• Sponsored Products, Brands and Display: what is the difference?
• ACoS and TACoS: which number actually matters?
• What do new-to-brand metrics really measure?
• Why pay to advertise on your own brand name?