How to structure Amazon PPC campaigns so you can actually read the data

Most Amazon advertising accounts we audit aren’t badly optimised. They’re badly organised. The bids have been adjusted hundreds of times, but nobody can say which change caused which result, because every campaign contains a mix of match types, product types and objectives.

You can’t optimise what you can’t read. So before touching a bid, we restructure.

Separate campaigns by job, not by product

The most common structure we see is one campaign per product, with every match type inside it. It feels tidy. It’s unreadable, because a single campaign is simultaneously trying to discover new keywords and to defend the terms you already know convert. Those are opposite jobs with opposite bidding logic.

Split them:

• Discovery — automatic campaigns and broad match. Job: find search terms you didn’t know about. Low bids, generous budget, reviewed for harvesting rather than for ACOS.
• Performance — exact match on terms you’ve proven convert. Job: win the impressions you know are profitable. Higher bids, tight ACOS target.
• Defence — your own brand terms. Job: stop competitors buying traffic that was already yours. Judged on impression share, not on ACOS.

Now every campaign has one job, and one number that tells you whether it’s doing it.

Group by how you’d bid, not by what the product is

Inside each campaign, ad groups should contain keywords you would bid similarly on. Two products with wildly different margins do not belong in one ad group, even if they’re in the same category — one bid will be too high for one and too low for the other, permanently.

A useful test: if you’d be comfortable applying a single bid change to every keyword in the group, the group is right. If you’d hesitate, split it.

Run a harvesting routine, not an occasional cleanup

Structure only pays off if search terms move through it. Weekly:

1. Pull the search term report from discovery campaigns.
2. Any term with enough orders to be meaningful moves to a performance campaign as exact match.
3. Any term with significant spend and no conversions becomes a negative in the discovery campaign.
4. Nothing sits in discovery indefinitely.

This is the entire engine. Discovery finds, performance exploits, negatives stop the bleeding. Done weekly it compounds; done monthly it leaks.

Give it enough time to mean something

Fourteen days is usually the earliest a structural change produces a readable signal, and for lower-volume products it’s longer. Judging a restructure after four days tells you about noise, not about the change.

What good looks like after 90 days

You should be able to answer, without opening a spreadsheet: which campaign is finding new terms, which is defending profitable ones, what your brand impression share is, and which terms you stopped paying for and why. If you can answer those, the account is readable — and a readable account is one you can improve every single month.

---

Managing Amazon advertising and want a second opinion on the structure? We audit accounts and give you a written read on what we’d change first.