Campaign (Amazon Advertising)

A campaign in Amazon Advertising is the top-level container that holds one or more ad groups and defines the rules under which they bid: daily budget, bidding strategy, targeting type (manual / auto), placement bid modifiers, start/end dates, and portfolio membership. Everything below — ad groups, keywords, products, bids, negatives — inherits the campaign's constraints.

Account performance is determined more by campaign structure than by any individual bid decision.

Why structure dominates

Three reasons:

Budget allocation is per-campaign. A poorly structured account groups high-CVR exact-match keywords in the same campaign as low-CVR broad-match discovery, so the daily budget gets eaten by discovery before the profit layer can spend.
Bidding strategy is per-campaign. Branded defence and competitor conquest need different bidding strategies — but they can only be set differently if they live in separate campaigns.
Reporting aggregates per-campaign. A campaign that mixes branded and generic keywords reports a single ACOS that masks both layers. You can't optimise what you can't see cleanly.

The canonical campaign architecture

For a single SKU in a competitive category, the minimum defensible structure:

#	Campaign	Targeting	Role
1	SP - SKU - Auto Discovery	Auto	Discovery via Amazon's algorithm
2	SP - SKU - Broad Discovery	Manual (broad)	Discovery on seed keywords
3	SP - SKU - Harvested Exact	Manual (exact)	Profit layer from harvest loop
4	SP - SKU - Branded Exact	Manual (exact)	Branded defence
5	SP - SKU - Competitor ASIN	Product	Competitor conquest
6	SP - SKU - Category PAT	Product	Refined category targeting
7	SB - SKU - Branded	Manual (exact)	Brand store traffic
8	SD - SKU - Remarketing	Audience	Close the funnel

Eight campaigns per SKU sounds heavy until you compare it to one campaign per SKU with everything mixed in.

Campaign settings that actually matter

Daily budget. Set against the campaign's role, not against a uniform number.
Start/end dates. Used for event campaigns. Always set an end date — otherwise event-time bid modifiers leak into normal trading.
Portfolios. Group campaigns by SKU family or business goal for monthly-budget caps.
Negative keyword/ASIN lists. Apply at campaign level for terms negated across all ad groups.

Campaign duplication: the safe-rollout pattern

When making structural changes to a winning campaign, duplicate first, modify the copy, run side-by-side for 14 days, then pause the loser. Direct edits to live winners cause more damage than any other single account behaviour — Amazon's relevance algorithm partially resets when bids, keywords, or targeting change, and a "small tweak" can flatten a hockey-stick campaign for 7–10 days.

Common mistakes

One campaign per SKU. Mixes economic roles, produces unreadable ACOS, makes negatives impossible.
Editing winning campaigns directly. Duplicate-and-test or live with the regression.
No campaign for the harvest layer. Most failing accounts harvest into the same campaign that did the discovery, eliminating the entire point of the loop.
No naming convention. Makes the account unreadable. See Campaign Naming Convention.