Amazon PPC Campaign Structure That Scales With Your Catalog

Key takeaways: Structure by purpose, not by convenience. One portfolio per product line, one campaign per job (discovery, control, defense, brand), one match type per campaign, and a small number of related ASINs per ad group. Keywords move from discovery to control when they prove they convert. Everything else is a rule you can write down.

A scalable Amazon PPC campaign structure has four layers, each with one job. Portfolios group products and hold budget caps. Campaigns hold one purpose and one match type. Ad groups hold a few closely related ASINs. Targets hold the bids. When each layer does only its job, you can read the account in ten minutes, move keywords between campaigns by rule, and add a product without redesigning anything.

Why structure matters more than bids

Most accounts we audit do not have a bidding problem. They have a structure problem that looks like a bidding problem. A broad-match keyword and an exact-match keyword for the same term sit in the same ad group, so Amazon picks whichever one it wants and the exact term never gets its own bid. Twelve ASINs share one ad group, so the search term report cannot tell you which product a click belonged to. A launch campaign and a mature campaign share a budget, so the launch starves the winner by lunchtime.

None of that is fixed by changing bids. It is fixed by giving each campaign one job.

The four layers

Portfolios

One portfolio per product line, brand or margin band; whichever split you would want a budget cap on. Portfolios are the level where you set a monthly spending limit, and that is the main reason to use them. If two product lines have different margins, they need different portfolios so one cannot drain the other.

Campaigns

One campaign has one purpose and one match type. We use four purposes, and the name tells you which.

What each match type is for, and how a term moves between them, is covered in Amazon Match Types: Broad, Phrase and Exact, Used Properly.

A naming convention makes this work in bulk files and reports. Something like SP | Kettle | Control | Exact tells anyone on the team the ad type, product, purpose and match type without opening the campaign.

Ad groups

Ad groups hold ASINs that share the same keywords and the same price band. Two colors of one product can share an ad group. A 500 ml bottle and a 1.5 L bottle should not, because the search terms, conversion rates and margins differ. The moment you need to know which ASIN a search term belongs to, the ad group has too many ASINs in it.

Targets

Each keyword or product target has its own bid. In discovery campaigns bids are modest and similar. In control campaigns each bid is set on that target's own data, which is only possible because the structure above it keeps the data clean.

How a keyword moves through the structure

The negative step is the part most accounts skip. Without it the discovery campaign keeps buying the same term at a bid that was never set for it, and the control campaign competes against your own auto campaign in the auction. The threshold for promotion is a rule you decide once: in most accounts two or three orders and an ACoS inside the target band over a rolling 30 days is enough. Full detail on how a term earns its own campaign is in B14 · How a search term earns its own campaign.

Single-keyword campaigns: when they help and when they do not

A single-keyword campaign gives one term its own budget, its own placement modifiers and a clean search term report. That is worth having for the five to fifteen keywords that drive most of a product's sales and rank. It is not worth having for two hundred long-tail terms, where the management cost outweighs the control and the data per campaign is too thin to act on.

Our rule: the head terms that you would defend a rank position on get their own campaign. Everything else that has proven itself lives in a shared exact-match control campaign, grouped by product.

Common structures that break as they grow

One campaign per product with all match types inside. Readable at ten products, unmanageable at forty, and the match types compete with each other.

Structuring by ACoS band. Moving keywords between "good", "medium" and "bad" campaigns as their ACoS changes means the campaign history never settles and the algorithm's learning resets each time. Structure by purpose and adjust bids instead.

Copying a template account. A supplement brand with 200 SKUs and a single-product hardware brand need different depth. The principle of one job per campaign holds; the count of campaigns does not.

No negatives between campaigns. Two campaigns bidding on the same term is the most common self-inflicted cost we find in audits.

B13How we structure your campaignsan account where every campaign has a purpose — discovery, control, defence or brand — and nothing overlaps.

B21How changes reach your accountchanges made in bulk, validated before upload, and logged so you can trace every edit.

Restructuring a live account

If the account already has history, do not rebuild it in a weekend. Existing campaigns hold conversion history that helps the auction, and pausing them all at once drops rank on every keyword at the same time. Build the new control campaigns alongside the old ones, move the proven terms across, add the negatives, and let the old campaigns wind down as their spend falls. Most restructures take four to six weeks to settle, and the weekly report should show spend moving from old to new without a drop in total sales.