Campaign Structure

Structure is leverage

A messy account is not just ugly, it is expensive. When products, match types and goals are tangled into a few sprawling campaigns, you lose the ability to control budget and bids at the level where the decision actually matters — and you lose the ability to tell what caused a number to move.

Structure is not admin. It is the thing that determines which levers exist. A bid you cannot set separately is a bid you cannot optimise, and a result you cannot isolate is a result you cannot learn from.

The hierarchy, and what lives where

Most structural mistakes come from not knowing which setting lives at which level. Amazon’s campaign structure gives you four, and each one owns different controls:

One question per campaign

The single rule that keeps an account legible: every campaign should answer one clear question. “How is my hero product doing on exact-match head terms?” is a campaign. “How is everything doing?” is a mess.

The test is diagnostic, not aesthetic. When ACoS moves three points, a well-structured account tells you which product, which match type and which intent moved — because each of those lives in its own container. A tangled account tells you only that something changed.

Which products belong in the account at all

Before deciding how to arrange campaigns, decide what goes in them. Advertising every ASIN you sell is a structural decision disguised as a default, and it is usually the wrong one — attention and budget both get thinner with every product added.

The candidates that reliably earn a place are the ones where advertising changes the outcome rather than accompanying it: products being launched, products with a deal running, products carrying excess inventory you need to move, and the small set that actually carries your revenue.

The ones to think twice about are those where ads cannot fix the underlying problem — a product with poor reviews, thin margin, or unreliable stock. Advertising a listing that does not convert does not just waste the spend, it writes the wrong evidence into your record.

Single-product campaigns, and when they are overkill

The cleanest default is one product, or one tight variation family, per campaign. Single-product campaigns give you product-level control of budget and bids, keep reporting legible, and stop a hero product quietly eating the budget of everything else.

But treat it as a default, not a law. At 400 SKUs, one campaign each is an account nobody can maintain, and most of those campaigns will never gather enough data to justify a decision. The honest rule is to spend structure where money and attention actually go.

Ad groups: the level most people misuse

An ad group is a bid and a set of targets applied to a set of products. That is the whole concept, and it is why what you put in one matters so much.

Amazon’s guidance is to keep ASINs in a campaign to a single category, and to group products by price range and conversion behaviour rather than by convenience — mixing products that perform differently forces one bid to serve two economics. Amazon reports ad groups containing similar products saw around 6% more units sold than those with less similar products.

• Same category, similar price, similar conversion rate. Those three make a coherent ad group.
• A target you add applies to every product in the group. If that is wrong for any of them, split the group.
• You can add products in bulk — up to 1,000 to a single ad group at once — which makes it tempting to over-fill them. Resist.

Discovery and performance are different jobs

The most valuable split in an Amazon account is not by product or match type, it is by job. Discovery campaigns exist to find out what converts. Performance campaigns exist to exploit what you already know converts. They want opposite bidding logic and opposite definitions of success, so they should never share a campaign.

Three ad types, three jobs

Structure is usually discussed as if Sponsored Products were the whole account. It is the foundation, but the other two formats do jobs it cannot, and mixing their reporting is how people conclude that brand advertising “does not work”.

Sponsored Products captures existing demand — someone is already searching, you compete for the click. Sponsored Brands occupies the top of the page with your logo and range, which is a visibility and brand job rather than a last-click one. Sponsored Display follows shoppers onto detail pages and off Amazon, which is retargeting.

Judge each on its own terms. Holding a brand-awareness format to a last-click ACoS will always make it look like a failure, because you are measuring the thing it is not for.

The keyword lifecycle

Structure is not static. Terms move through it, and the movement is the point:

• Auto and broad campaigns discover the search terms shoppers actually use.
• Terms that convert repeatably get promoted into exact-match campaigns where you set the bid deliberately.
• The discovery campaign then negates the harvested term, so the two campaigns stop bidding against each other for the same impression.

Negatives: which level, and why it matters

Negative keywords can sit at campaign level or ad group level, and the choice is a structural one rather than a detail.

A campaign-level negative blocks the term everywhere in that campaign. It is the right tool for a search that is simply wrong for the product — junk, a competitor brand you cannot serve, a use case you do not fit. An ad group negative blocks it only in that group, which is what you want when you are routing traffic rather than rejecting it: the term is fine, it just belongs to the exact-match group instead of the discovery one.

Match types as a structural decision

Match type is usually taught as a targeting setting. Structurally it is a budget decision, because match types have wildly different costs per unit of certainty.

Broad and phrase buy you reach and surprise; exact buys you control. If they share a campaign they also share a budget, and the loose match types will reliably spend it first — leaving your proven terms capped while your experiments run free. Separating them is the only way to guarantee your winners get funded.

Portfolios and the budget hierarchy

Budget is enforced at more than one level, and using only the campaign level is how accounts overspend a month while every individual campaign looks reasonable.

Portfolios group campaigns by brand, category, season or whatever matches how you actually think about the business, and they can carry their own budget cap. That gives you a ceiling on a whole product line without micro-managing thirty campaign budgets underneath it.

Placements: the modifier that quietly doubles your CPC

Amazon serves your ad in three broad placements — top of search, the rest of search, and product pages — and lets you raise your bid for each. The adjustment goes up to 900%, which means a placement modifier can multiply your effective bid several times over without you touching the bid itself.

Top of search usually converts best and costs most. That is a real trade, and it belongs in your structure: if a campaign exists to win top-of-search on your best terms, set the modifier there and nowhere else, so it never silently applies to your discovery spend.

Do this in SellerMateBulk Actions →Set Top of Search, Product Pages and Rest of Search modifiers across every campaign at once, instead of opening four hundred campaigns to change one number in each.

Naming conventions that scale

Boring, consistent names are a superpower at scale. A pattern like SP | ProductCode | MatchType | Goal lets you filter, bulk-edit and report in seconds, and it turns a 400-campaign account into something you can query.

The convention matters far less than the consistency. Pick one, write it down, and never break it — a naming scheme with exceptions is just decoration.

Where the money should sit

Structure only pays off if budget follows it. Amazon’s own advice on common campaign mistakes is to concentrate spend on high-priority products and hold experimentation to a modest slice — on the order of 10 to 15% of budget — rather than letting tests quietly compete with proven performers.

That ratio is also a structural test. If you cannot tell what share of spend is going to experiments, your discovery and performance campaigns are not properly separated.

How much structure is too much

Every guide tells you to segment. Almost none tell you when to stop, and over-segmentation is a real failure mode with a real cost.

Split finely enough and every campaign gets a trickle of impressions. Nothing reaches statistical significance, every number looks like noise, and you end up making decisions on three clicks. You have bought precision in control and paid for it in evidence.

How to restructure without losing what works

Every account eventually outgrows its structure, and the fix is genuinely risky: campaigns accumulate performance history, and a brand-new campaign starts without it. Rebuilding from scratch means paying to relearn things you already knew.

So restructure incrementally rather than in one weekend. Build the new campaign alongside the old one, move a slice of the terms, and let both run until the new one has enough history to compare. Then move the next slice.

• Never migrate everything at once — you lose the ability to tell whether the new structure or the new bids caused the change.
• Move your best terms last. They are the ones you cannot afford to have relearning.
• Keep the old campaign paused rather than deleted for a while, so its history stays readable.
• Expect a dip. A migrated term is a cold start even though you have known its numbers for a year.

Structural mistakes to avoid

Keep the structure from decaying

Structure decays. Products get added to the wrong ad group in a hurry, a harvested term never gets negated, a test campaign quietly becomes permanent. The account that was clean in January is not clean in June unless something keeps it that way.

Audit it on a schedule rather than when something breaks: campaigns with no spend, ad groups spanning price bands, terms live in two places at once, and budgets that no longer match priority.

Master Sifu — Decide the naming convention before you need it rather than after. Every filter, migration and bulk edit you will ever run depends on it — and a convention invented halfway through is worse than a mediocre one applied consistently.