Amazon PPC Campaign Structure: The Setup That Actually Scales

Most Amazon accounts I open look the same: campaigns named "Campaign - 3 - final - NEW," every match type mixed together, and the same keyword competing against itself in four places. The account "grew organically," which means nobody can tell which dollar is doing what.

Structure isn't cosmetic. It's what makes every other optimization possible. Here's the setup I run.

What a structureless account looks like in an audit

Before I restructure an account, I map it. The pattern I see most often: a seller has been running ads for two or three years. Someone pressed "launch campaign" whenever a new product went live. Someone else harvested occasionally but never negated. Now there are 15 broad campaigns with overlapping keywords, three auto campaigns pointed at the same ASIN, and exact keywords that haven't had a bid touch in eight months.

I work across six brands and have touched accounts running 500+ campaigns. The $300–400K/year accounts are often the worst — not because they're large, but because structure calcified over years and nobody wanted to break it. The seller is afraid to touch anything because the account "works." Structureless doesn't mean efficient, though. It means hidden waste and no levers to pull when you need to move numbers fast.

Each match type has one job

Auto campaigns — discovery. Amazon matches your listing to search terms you'd never think of. Auto is a research engine you pay for, so treat it like one: modest budget, harvested regularly, never your main spend.

Broad match — expansion. Broad (and phrase) catches variations around your known keywords. This is where I spend a lot of my attention — broad plus search-phrase analysis is my specialty, and it's where the next tier of winners comes from.

Exact match — profit. Proven terms only, precise bids, most of the budget. Exact is where you actually control your economics, because you know exactly which search term you're buying.

A concrete structure for one product

Here's how I'd set up campaigns for a single mid-range product — say, a silicone spatula set:

• SP | Auto | Spatula Set | Discovery — modest daily budget ($5–15 depending on price point), pulls raw search term data
• SP | Broad | Spatula Set | Expansion — feeds off terms harvested from auto, catches variations
• SP | Exact | Spatula Set | Profit — proven converters only, highest budget allocation
• SP | Exact | Spatula Set | Brand — your own brand terms, defensive bid

Four campaigns. That is the core. You can layer in a competitor targeting campaign if margin allows, but these four cover the full loop. Nothing is mixed. Every search term has exactly one home.

Naming conventions matter at scale

The campaign names above follow a consistent pattern: [Ad type] | [Match type] | [Product] | [Purpose].

Ad type: SP (Sponsored Products), SB (Sponsored Brands), SD (Sponsored Display). Match type: Auto, Broad, Phrase, Exact. Purpose: Discovery, Expansion, Profit, Brand, Competitor.

This isn't an aesthetic preference. When you are diagnosing which campaign is burning budget at 11 pm before a major sales event, you need to know at a glance what a campaign does. When you filter in Seller Central, consistent naming means your filters work. When you hand off to a VA or bring on another manager, the naming explains the structure without a briefing doc.

At 400–500 campaigns, the person who skipped naming conventions three years ago has created an account no one can operate clearly. I have audited accounts where every campaign is named "[Brand] - Keyword Targeting - 1" through "[Brand] - Keyword Targeting - 34." That account is invisible from the inside.

The flow: harvest up, negative down

The structure works as a funnel:

1. A search term gets sales in auto or broad (my threshold: real orders, and at least 10 clicks of data).
2. Add it as exact in the exact campaign, with its own bid.
3. Negative-exact it in the source campaign.

Step 3 is the one everyone skips, and it's what stops self-competition. Without it, your auto, broad, and exact campaigns bid against each other for the same customer, and your reporting turns to soup. With it, every search term has exactly one home.

Monthly harvesting routine

Harvesting once and forgetting about it is how structure degrades. I run the same checklist every month:

1. Pull the search term report for all auto and broad campaigns — last 30 days minimum
2. Filter for 10+ clicks with at least one order — these are candidates to move up to exact
3. Filter for 10+ clicks and zero orders — these are candidates to negative in the source campaign
4. For each winner: add as exact keyword in the exact campaign, negative-exact in the source
5. For each wasted term: add as negative-exact in the campaign it appeared in
6. Verify no exact keyword appears in more than one exact campaign
7. Review existing exact bids — terms consistently hitting your ACoS target can absorb small bid increases; terms running well above target get a cut

This takes 60–90 minutes per account once a month. Skip it for two months and self-competition quietly creeps back in.

Bidding inside the structure

• Dynamic bids, down only. Amazon can lower bids on auctions unlikely to convert but can never inflate them. "Up and down" hands Amazon permission to overspend on your behalf.
• New exact keywords get a strong opening bid — I typically start around double Amazon's suggestion for the first days to win placement and gather data fast, then settle to what the conversion data supports.
• Judge nothing under 10 clicks. Under that, you're pattern-matching on randomness.

Budget split across auto, broad, and exact

New sellers often ask where the money should go. My working split for a stable, proven product:

• Auto: 10–15% of total PPC budget. It is a research cost, not a primary revenue driver.
• Broad: 25–35%. Active monthly harvesting keeps this lane feeding the exact list with fresh winners.
• Exact: 50–65%. This is where your proven terms live and where the majority of sales should close.

These ratios shift with product lifecycle. A new launch needs heavier auto and broad spend — you are buying data. A mature product with a full exact list should concentrate budget there. The target ACoS for exact campaigns should sit around 70% of your break-even ACoS — tight enough to be profitable, loose enough to stay competitive on proven terms.

On single-keyword campaigns

The short answer: yes, but only for your top performers.

If a single search term drives 15–20% of your revenue, it earns its own campaign. You want isolated budget control, independent bidding, and clean data with no other terms introducing noise. One keyword in, one conversion path.

For everything else, group terms by theme and intent inside your standard exact campaigns. Running 25 single-keyword campaigns for a mid-range product creates a maintenance burden that outweighs the marginal bid control you gain. Save the SKC structure for the terms that genuinely move the account.

Why this matters more in the Rufus era

Amazon's AI shopping assistant (Rufus) and its evolving ranking systems lean harder on relevance signals — which listing genuinely satisfies which query. A clean structure concentrates your sales history keyword-by-keyword instead of smearing it across duplicate campaigns. Structured accounts are teaching Amazon exactly what they're relevant for; messy ones are whispering.

The rebuild, in practice

You usually don't need to torch the account. The sequence I run: map every live keyword and its home → kill duplicates → establish the auto/broad/exact lanes → set up harvesting as a recurring routine, not a one-off cleanup. Sales history stays, structure improves around it.

If your account grew organically and you suspect it's a mess — it probably is, and it's costing you every month. The free audit will show you exactly where: structure, wasted spend, and the three highest-leverage fixes. Or start with the numbers: check your ACoS against break-even.