Negative Keywords: A Weekly Routine That Cuts Wasted Ad Spend

Every search term report contains terms you never chose. Some earn orders. Many only spend money. The usual fix is to add a negative keyword after a fixed number of clicks with no sale, say ten, but that rule treats a product that earns $9 per order the same as one that earns $90. A click count has no connection to what a sale is worth. This article builds the rule from the profit one order earns, explains which negative match type to use, and sets out a short weekly routine you can repeat.

Where the wasted spend comes from

The search term report shows the actual words shoppers typed, or the product pages they were viewing, when your ads were shown and clicked. For each one it lists impressions, clicks, spend, orders and sales. You choose the keywords you bid on. You do not choose the terms they match to, and the broader the targeting, the wider the gap.

Waste usually falls into three kinds:

• Irrelevant terms: the shopper wanted something else, such as a replacement part for a product you sell whole.
• Relevant terms that do not convert: the shopper wanted something like your product, but your price, size or features did not fit.
• Terms that duplicate other campaigns: you already target the term somewhere else with a bid you set on purpose, but an automatic or broad campaign is also serving it, with a bid you did not choose.

Each calls for a different response. Treating all three as one pile is how sellers end up blocking terms that would have paid.

Small leaks add up. Thirty terms that each spend $18 without producing an order cost $540 over the window you are looking at, enough to change a product's profit for the month. Because the leak is spread across many rows, it rarely looks urgent in a single campaign view.

A rule built on what one order earns

Start with the profit one order earns before advertising. Call it P. In the examples below P is $9, from a $30 product with a 30% margin before ads. P is the unit of account for every test you run, because it is the most a single order can contribute toward paying for the clicks behind it.

Now consider a term that converts at exactly your break-even rate. It neither makes nor loses money. On average, it returns one order for every $9 you spend on it. If orders arrive at random, the chance of seeing none after $9 of spend is about 37%. After $18 it falls to about 14%, roughly 1 in 7. After $27 it is about 5%, or 1 in 20. These figures come from simple probability, not from Amazon data, and they assume random arrivals.

That gives the rule. A relevant term with no orders should be negated once its spend reaches about twice P, which is $18 here. For a core term very close to your product, wait until three times P, $27, and try a lower bid first. In both cases the odds of cutting a term that was actually paying its way are low. If the term really was exactly break-even, cutting it costs you nothing, because it was earning nothing.

Irrelevant terms are different. Negate them at once, whatever the spend, because no amount of testing will make the shopper want your product.

The rule also makes your testing budget explicit. Each unproven term gets about two orders' worth of profit to prove itself, and then it goes. You can see in advance what discovery costs.

When the problem is the bid, not the term

A term with orders but a poor ACoS is usually a bid problem. Work out the most you can afford per click: profit per order multiplied by conversion rate. At 10% conversion, that is $9 × 10%, or $0.90. A term costing $1.20 a click will lose money on average even if it converts at your average rate. Lower its bid, or move it into its own campaign. Do not block a term that sells.

Exact, phrase and product negatives

In Sponsored Products you can add negative keywords at the campaign level or at the ad group level, and negative product targets for ASINs you do not want to appear against. The two keyword match types behave differently.

Use exact for terms that are relevant but not paying, and phrase for repeated irrelevant wording. Be careful with phrase negatives on common words. A negative phrase on a single generic word can silence searches you want, so check that the phrase does not appear inside any term that has already produced orders.

Product targeting works the same way, with ASINs in place of words. The report lists the product pages your ads appeared on. Add an ASIN as a negative product target when it belongs to something a shopper would not substitute for yours, such as an accessory, a different size class, or a product at a price far from your own.

Level matters too. A negative at ad group level only affects that ad group, which suits a term that is wasting money in one place but is fine elsewhere. A campaign-level negative covers everything in the campaign, which suits an irrelevant term.

Duplicates are worth handling on purpose. If a term has earned its own exact-match keyword in a manual campaign, add it as a negative exact in the automatic or broad campaigns. The term then serves from the campaign where you set the bid directly.

A weekly routine

Fifteen to thirty minutes a week beats a long monthly clean-up, because wasted spend compounds daily.

The recent days are left out because Amazon can credit an order to a click for some time after the click. A term with no orders from yesterday may still convert. Use a window of two to four weeks: long enough that the spend threshold can be reached and short enough to reflect the present.

Keep a simple log of each negative, with the date, the match type and the spend at the time. Every quarter, review the log. A negative added when your price was higher, or your images weaker, may deserve another test.

Mistakes that cost sales

• Negating a term that produced orders because of one weak week.
• Using the same click count for every product, regardless of what an order earns.
• Adding phrase negatives on words that appear in good terms.
• Removing so much that discovery campaigns have nowhere left to find new terms.
• Forgetting that negatives only stop waste. They do not create demand.

The short version

• Work out the profit one order earns before ads for each product.
• Negate irrelevant terms at once, relevant zero-order terms at about twice that profit, and core terms at about three times.
• Lower the bid on terms that sell but cost too much per click.
• Use negative exact for single terms and negative phrase for repeated irrelevant wording.
• Run the review weekly, leave out the most recent days, and log each change.