Ask why an account has 600 campaigns and the answer is almost never a decision. It is sediment: each agency, each launch and each experiment added a layer, and nobody removed one.
The cost is not aesthetic. Past a certain campaign count you stop being able to answer basic questions — which terms are profitable, where budget is being wasted, whether a placement is working — without exporting everything to a spreadsheet. Optimisation quietly stops.
Three axes, and only three
A structure that stays readable separates campaigns along three dimensions and resists adding a fourth.
- Intent: discovery (broad, phrase, auto) versus harvest (exact match on proven terms) versus defence (branded).
- Margin tier: group SKUs by contribution margin band, because each band can tolerate a different cost per sale.
- Placement type: Sponsored Products, Sponsored Brands and Sponsored Display kept in separate portfolios so budgets do not silently compete.
Everything else — season, launch, promotion — belongs in a naming convention or a portfolio, not in a new campaign type.
Why discovery and harvest must be separate
Discovery campaigns exist to find search terms. Harvest campaigns exist to own the ones that convert. If both live in the same campaign, they compete for the same budget and you can no longer tell whether a term is performing because it is good or because it happened to get the impressions.
The practical protocol: when a term in a discovery campaign passes an agreed threshold of clicks and conversions, promote it into an exact-match harvest campaign with its own budget, then negate it in the discovery campaign so the two stop bidding against each other.
Naming conventions are not bureaucracy
A campaign name should be parseable by a spreadsheet. Something like SP | Harvest | Tier-A | Hydration | EXACT lets you pivot spend by type, intent, margin tier and product line without touching a single campaign manually.
The rule that matters more than the format: the convention is decided once and never changed. Two conventions in one account is worse than a bad convention applied consistently.
Migrating without losing rank
The risk in a restructure is real. New campaigns have no placement history, and Amazon's systems take time to build it. Rebuild everything on a Monday and you can expect a difficult fortnight.
A safer sequence:
- Build the new structure alongside the old one, at low budget, and let it accumulate history for two weeks.
- Migrate in waves by product line, starting with the line that carries the least revenue.
- Hold legacy campaigns live at reduced budget until the replacement matches or beats them over a full read window.
- Only then pause the legacy campaign — pause, not archive, for at least a month.
Do not start a migration in the eight weeks before a peak trading period. The read windows will not fit and you will be making decisions on noise during your most important month.
How many campaigns is the right number?
There is no target. A useful test instead: can someone who does not work on the account daily open it and correctly explain where the money is going within ten minutes? If not, the structure is too complicated regardless of the count.