Amazon Ads Keyword Targeting, Explained With a Worked Example

What keyword targeting actually is on Amazon

Keyword targeting is the mechanism that connects a shopper's search query to your ad. You attach keywords to a Sponsored Products or Sponsored Brands campaign, set a bid, and Amazon's auction decides whether your ad is eligible to show against that search — based on your bid and how relevant your listing is to the term, not bid alone.

It sits alongside two other targeting options: product targeting, where you bid on specific ASINs or categories instead of search terms, and automatic targeting, where Amazon selects the matching terms for you based on your listing content and past purchase behavior. Keyword targeting is the one you control directly — you choose the term, the match type, and the bid.

One operational detail that catches people out: once a campaign is published as automatic or manual, you can't switch it. If you launched auto and want manual keyword control, you build a new campaign rather than converting the old one. Plan the split before you launch, not after.

Match types: broad, phrase, exact — and what each one is actually for

Match type controls how closely a shopper's search has to resemble your keyword before your ad is allowed into the auction. Getting this wrong is the single most common way accounts either waste spend or starve themselves of volume.

• Broad match shows your ad on searches related to your keyword, including synonyms and loosely connected terms. It's the widest net — good for finding search terms you haven't thought of, expensive if left unmanaged.
• Phrase match requires your keyword to appear in the search as a phrase, with words allowed before or after it. It's a middle setting, narrower than broad but still open to some variation.
• Exact match only shows your ad on the keyword itself or its very close variants — plurals, misspellings, word order. It's where you put terms you've already proven convert, and where you exercise the most bid control.

The practical sequence most accounts use: launch broad or auto to discover which search terms actually convert, promote the winners to exact with a controlled bid, and use phrase to hold a middle tier while you decide. Skipping straight to exact on day one means you never find the terms outside what you already guessed.

Negative keywords: the setting most accounts get wrong

Negative targeting stops your ad from showing on a specific term or phrase. It costs nothing to add, and it's available on both automatic and manual campaigns — which is exactly why it gets skipped. It feels like housekeeping, not strategy.

Two negative match types matter. Negative exact blocks that one specific search term. Negative phrase blocks any search containing that phrase — the blunter, wider tool. The mistake we see most often, and one we've made ourselves early on a new account, is negating at phrase when exact would have done the job. That quietly cuts off adjacent search terms that were actually converting.

Negative keywords also do a second job people miss: they stop your own campaigns from bidding against each other. If a broad match keyword and an exact match keyword in a different campaign are both eligible for the same search term, you're competing with yourself in the auction. Negating that term out of the broad campaign, once you've promoted it to exact, is how you stop that.

A worked example: what a keyword's bid actually buys you

Every keyword bid is a claim on cost-per-click, and CPC benchmarks in this category get quoted without context constantly. A single blended figure that circulates widely in Amazon advertising commentary is $0.41 — and it's real, but it's an online-video number, not a representative cost across a full account.

Here's what an actual blended book looks like: across 30 of the advertiser accounts we manage, July 2026 ran 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost-per-click. That's a DSP book, not a sponsored-products keyword auction specifically, but it makes the same point that applies inside keyword targeting too — a single average CPC hides enormous variance between placements, match types, and categories. Treat any CPC number you read as a starting range to test against your own auction, not a target to hit.

The number that matters more than CPC in isolation is what it bought. That same book converted at a blended cost per acquisition of $5.49 across 57,137 attributed purchases, with 20.1% of those purchases coming from shoppers new to the brand. A keyword bid that looks expensive on CPC alone can still be cheap per acquisition if the match type is tight and the term carries real purchase intent — and a keyword that looks cheap on CPC can be a bad bid if it's pulling in browsers who never convert.

Common mistakes — and what to do when the fix doesn't work

The mistakes that actually move ACOS are boring. Broad match left unreviewed for weeks. Keywords judged and paused after eight clicks, before there's enough data to know anything. Negatives added at the wrong match type, cutting off terms that were converting. Auto and manual campaigns quietly bidding against each other on the same search term because nobody checked the search term report to see which one was winning.

When a keyword's ACOS spikes, don't pause the whole ad group. Pull the search term report, find the specific query burning spend, and negate that query at exact match in the campaign serving it. That isolates the problem instead of cutting good keywords along with the bad one.

When a bid increase doesn't move impression share, the problem usually isn't the bid. Amazon's auction weighs relevance alongside bid, so a low-relevance listing can raise its bid indefinitely and still not win the placement. Check the listing's title, images, and backend search terms against the keyword before spending more to chase a placement the algorithm doesn't think you deserve.

When you add a negative and nothing changes, check where you added it. A negative added at the ad-group level only blocks that ad group; if the same keyword exists in a different campaign, it's still live there and still spending.

Where this fits with reMKTR

Keyword targeting is a Sponsored Products and Sponsored Brands discipline — it's not what reMKTR runs day to day. We run Amazon DSP as a managed service, holding real DSP seats across 109 live advertisers, and we reconcile that spend against sponsored ads in Amazon Marketing Cloud so a brand can see whether display added incremental sales on top of what keywords were already going to capture, instead of both channels claiming credit for the same last-click sale. If your keyword targeting is already tuned and the open question is whether display is worth adding on top of it, that's a reconciliation question, not a keyword one — and it's the one we're built to answer.