Amazon negative keywords: the harvest-and-negate loop that stops your campaigns bidding against each other

Every Amazon PPC guide gives the same advice: run an auto campaign, find the search terms that convert, and move them into an exact-match campaign. That advice is right, but it stops halfway. The step most guides skip is the one that saves the most money: negating the winning term back in the campaign you found it in.

We saw it firsthand. In 65 days of search-term data from a supplement seller's account (Root'd, the brand behind Adtender), terms being served by two or more of the seller's own campaigns at the same time had taken about two-thirds of search spend.

What negative keywords actually do on Amazon

A negative keyword tells Amazon not to show a campaign's ads for a search. That's it. It doesn't lower a bid or pause anything else. It removes one campaign from one auction.

There are two kinds you'll use most:

• Negative exact blocks that exact search and its close variants (plurals, small misspellings). Use it to stop one campaign from serving a term another campaign now owns.
• Negative phrase blocks any search containing the phrase. Use it to cut whole themes that never convert for you, like "free" or a competitor's brand you don't want to pay for.

Reading the search term report

The search term report is where this all starts. You can download it from Sponsored ads reports in the Amazon Ads console, and Amazon keeps about 65 days of it. Three columns matter most:

• Customer search term: what the shopper actually typed, or an ASIN if your ad showed on a product page.
• Spend and orders: what you paid and what you got back.
• Campaign and ad group: who paid for it. This is the column people ignore, and it's the one that shows overlap.

Sort by search term, not by campaign. The moment you do, you'll see the same term listed under several of your campaigns.

Harvesting: when a term earns its own campaign

A term is ready to graduate to exact match when it has proved it converts, not just that it gets clicks. A reasonable starting rule for a small brand:

• At least 2 to 3 orders in the last 30 to 60 days, and
• An ACoS at or under your break-even for that product (more on break-even in our ACoS vs TACoS post).

Give it its own exact-match campaign, not just a keyword in the same ad group. Its own campaign means its own budget and its own bid, and clean reporting on what that one term earns.

The step everyone skips: negate it where it came from

Here's what happens if you stop at harvesting. The term now lives in your new exact campaign, and it's still live in the auto, broad or phrase campaign that found it. When a shopper types it, both campaigns are eligible. Amazon only shows one of your ads in a given auction, so your campaigns aren't literally raising each other's price. But you lose control in three ways:

1. The campaign with the most aggressive bid or placement setting wins, not the one you meant to own the term.
2. Your reporting splits the term's results across campaigns, so no single campaign has enough data to set the right bid.
3. Your auto campaign keeps spending on a term you've already learned everything about, instead of finding new ones.

On that account, the phrase "multivitamin powder for men" was being served by five of the seller's campaigns in the same 65 days. Most of the orders came through the exact-match campaign built for it. The others mostly collected clicks.

Negative exact vs negative phrase, and when each backfires

Use negative exact for harvesting. It's surgical: it only blocks the term you've moved.

Be careful with negative phrase. Negating "powder" in a vitamin powder campaign would block your best searches. Use phrase negatives for words that are wrong in every combination, like a size you don't sell or a use your product isn't for.

A common mistake is negating a term because it had a bad week. Give a term enough clicks to judge. For a product that converts at around 10%, 20 to 25 clicks with no orders is a reasonable point to cut it.

Cutting waste: the zero-order terms

The second half of the loop is cutting terms that never convert. In our 65-day window, 49 search terms had each spent more than $15 without a single order, about $1,479 in total. A few examples:

None of these were negated. The only negatives in most of those campaigns were a standard list of "deals" terms added when each campaign was created.

How often to run the loop

Weekly is enough for most brands spending under $20k a month. Daily is better once you have the volume, because a term that's wasting money wastes it every day you leave it. The point is to have a rhythm, not to do it once and forget it.

Automating it without losing control

This loop is a good candidate for automation because the rules are clear: graduate a term after X orders under your break-even ACoS, negate it in the source in the same step, and cut terms after Y clicks with no orders. What matters is that you can see every change and why it was made.

That's the loop Adtender runs: harvest, negate back, cut waste, and log the reason for each change with a one-click undo.

Questions