Negative keywords for shops: lists that do not kill revenue by accident

Check the existing list first, then exclude new terms

The usual guide starts with brainstorming: create a list, enter words such as "gratis" (free), "gebraucht" (used) or marketplace names, done. For a shop I reverse the order. A superfluous click costs a few cents to a few euros. A wrong exclusion costs every bit of revenue the blocked search would have brought from today onwards, and nobody notices anywhere. So the starting question is what your existing exclusions block today, and only then what to add.

Google itself warns that too many negative keywords reduce your reach to potential customers. In my assessment, expensive mistakes tend to come from lists that were copied once from a template and never looked at again after range changes. On top of that, a keyword blocked by a negative continues to be shown as active in the account.

Why Shopping and Performance Max do not warn you

In Search campaigns there is at least one check: a negative that overlaps with one of your own keywords stops that keyword from serving, and such overlaps can be found. In Shopping and Performance Max this counterpart is missing. There are no keywords of their own against which an exclusion would show up as a conflict. An overly broad negative blocks product searches, and in my assessment you notice it only through missing revenue, with no message appearing.

Then there is the reach of the levels. According to Google, account-level exclusions apply automatically to all Search and Shopping inventory in Search, App, Shopping, Smart, Local and Performance Max campaigns. If you enter "gebraucht" (used) there because the Search campaign advertises new goods, you also block the Shopping ads for the refurbished devices in your feed. In Performance Max, exclusions apply to Search and Shopping inventory only and can be entered per campaign or assigned as an existing list. There, Google shows a forecast of how additional exclusions would affect performance. I would look at it, but not treat it as approval, because it knows nothing about margin.

Auditing the existing list against feed and revenue

The core of the audit is a comparison of two lists: on one side all active exclusions at account, list, campaign and ad group level, on the other what you sell and what you have sold with.

1. Against the feed. Lay every exclusion next to the product titles, stocked brands, conditions and promotion terms in your feed. If a negative hits a word that appears in titles, it is a candidate for removal or narrowing. Typical hits are "gebraucht" (used) or "refurbished" for retailers with B-stock, "sale" or "günstig" (cheap) for shops with an outlet section, and brand names that you suddenly stock yourself after a range extension.
2. Against past revenue terms. Export from the search terms report the terms with revenue over the last twelve months, including the period before a list was created or extended. Every term with revenue that a current exclusion would hit is a sign of blocked revenue.
3. Weight by contribution margin. Not every hit is equally expensive. An exclusion that hits a minor category with a thin margin is less urgent than one that touches your best-selling brand.
4. Clarify the level. For each critical hit you decide: remove it, narrow it to phrase or exact, or move it to a smaller level.

The analysis of the search terms themselves, meaning which terms lose money at all, belongs in the commercial view of the search terms report. This article is about the architecture in which the results end up.

New exclusions only with a calculation

A search term, or a group of similar terms, becomes an exclusion candidate when over a sufficiently long period it costs more than it brings in contribution margin:

Contribution margin after advertising = net revenue × contribution margin rate − ad spend

If the result is negative and the data base is large enough, you exclude. If it is positive, the term stays, even if it looks irrelevant at first glance.

A worked example with invented figures: suppose search queries containing "gebraucht" (used) cost €1,200 in a year and brought €3,000 of revenue. At a contribution margin rate of 30 percent, €900 remains, which is minus €300 after advertising. Excluding is worthwhile. If the rate is 45 percent, €1,350 remains, plus €150. The same term is then no longer a candidate but a small earner. Anyone who excludes on gut feeling does not see this difference.

This rule does not apply to terms that have nothing to do with your offer, such as job searches or instructions without purchase intent. You exclude those without a calculation as soon as they appear with noticeable cost. And it needs enough clicks: with a few dozen clicks, chance decides whether a purchase was among them.

Which level for which exclusion

The level determines how much a mistake costs. I would enter as little as possible at the very top.

Per account, Google provides for at most 20 lists of 5,000 exclusions each. In my assessment, anyone hitting that limit has a structure problem rather than too few lists.

For brands there are brand exclusions alongside negatives. In Performance Max they apply to Search, Shopping and YouTube search inventory, and you can still allow Shopping ads for excluded brands. For a retailer that sells other makers' brands, this is often the cleaner route compared with a negative on the brand name. Details are in the article on brand exclusions in Performance Max, and the order of exclusions is covered under negative keywords in Performance Max.

Standard lists: what is useful and what is dangerous

Common templates recommend terms such as "sale", "gebraucht" (used), "ebay", "Zalando" or "Otto" as general exclusions, with the note to adapt them. It is exactly this note that gets overlooked when copying. For a shop with an outlet, "sale" is a buying signal, for a retailer with B-stock "gebraucht" (used) is part of the range, and anyone who stocks products that are also searched for on marketplaces may lose buyers who compare prices by excluding the marketplace name.

As an account list, I consider suitable only terms with no buying intent whatsoever for your range: job ads, apprenticeships, free downloads, craft instructions, provided you do not sell craft supplies. Anything that describes a product feature, a condition, a price anchor or a retailer name belongs in a deliberately assigned list or at campaign level, once the calculation above has come out negative.

German word forms: too broad and too narrow at once

Exclusions work on words and, according to Google, do not match similar variants. A broad match negative "flowers" blocks "red flowers" but not "red flower". You have to add synonyms, singular and plural yourself. For German-language shops this has two consequences.

Too narrow: an exclusion "gebraucht" (used) does not catch "gebrauchte" or "gebrauchtes". If you want to exclude a word, you need each inflected form separately.

Too broad: broad match negatives block as soon as all terms appear in the search in any order. An exclusion "reparatur" (repair) therefore also blocks "reparatur set fahrrad" (bicycle repair kit), a buying search if you sell tools. Compounds, on the other hand, are not caught by any exclusion, because all of them work on words: "schuhe" (shoes) does not hit "laufschuhe" (running shoes). That helps with narrow exclusions and hurts if you rely on the exclusion.

My rule: for terms with possible buying intent I use phrase or exact, for clearly irrelevant topics broad with all forms. Exact match exclusions block only the exact search query without additional words, which makes them the tool for individual expensive outliers, not for topics. How broad match on your own keywords increases the number of search terms you then have to filter is covered under broad match in a shop.

Limits of this approach

The audit against the feed assumes that your product titles contain the words customers search with. If the feed is thin, you will not find conflicts that exist anyway; more on this under the feed as a model of the range. The contribution margin calculation needs margins per product or category and enough clicks. If either is missing, all you have left is the comparison against the feed and restraint at account level. Status of the Google details: November 2026; limits and options change regularly.