Negative keywords for B2B: a system, not a list

What are the four species of B2B search waste?

Read any B2B account's search terms for an hour and the waste sorts itself into four recurring species, each deserving its own themed list:

• Careers: jobs, salary, vacancy, internship, career queries. People hunting employment click ads with startling persistence; on a $5.87 median CPC for business services, each click is real money.
• Learners: what is, how does, tutorial, certification, course. Informational intent that will never book a demo this quarter.
• Wrong-pocket buyers: free, open source, cheap, template, DIY. Real intent, no budget; a paid click either way.
• Competitor support: competitor name + login, pricing, help, cancel. Users of a rival product looking for its support page. Bidding on competitor brand terms can be a deliberate play; paying for their password resets never is.

Why themed shared lists instead of ad-group negatives?

Because maintenance beats archaeology. Four shared lists named by species, applied account-wide, mean every weekly addition lands in one predictable place and every new campaign inherits protection on day one. Ad-group-level negatives still have a job, shaping which ad group wins an overlapping query, but waste blocking belongs at the account level. When we rebuilt the HR-software account in this teardown, all 412 additions went into three themed lists; the CPA fell from €212 to €96 across ten weeks.

How does the weekly cadence work?

1. Open the search-terms report, last 7 days, sorted by cost.
2. Read every query above roughly 0.5% of weekly spend; decide keep, negative, or watch.
3. Add negatives to the matching themed list, at the narrowest match type that kills the waste.
4. Log the week's additions and their 7-day cost in the change log, so the recovered euros stay visible.
5. Monthly, sanity-check the blocked-terms impact of any wide negatives added.

The cadence matters more than any single session, because broad match keeps exploring: Google's matching reaches further every year, and a list frozen at launch decays like an untended garden. Under an hour a week holds the line on accounts up to roughly €50,000 in monthly spend.

Which match-type traps catch B2B accounts?

Negative keywords use stricter, older matching than positive keywords, and the asymmetry causes both kinds of failure. Waste slips through because a negative does not expand: blocking "free crm" does nothing against "crm at no cost", so themes need their variants spelled out, plurals and common misspellings included. Reach dies because a negative is too wide: one phrase negative on a category word can quietly block thousands of buying queries. The narrowest effective negative is the rule; the blocked-terms check is the safety net.

How do you build the first lists in one afternoon?

New accounts and neglected ones start the same way: seed the four themed lists before reading a single report. Careers gets the obvious dozen (jobs, salary, vacancy, hiring, internship, career, recruitment and their plurals); learners gets the question words and "certification", "course", "tutorial"; wrong-pocket gets "free", "open source", "template", "cheap"; competitor-support gets each rival's name paired with "login", "pricing", "support", and "cancel". Thirty minutes of seeding prevents the first month's most predictable waste.

Then spend the rest of the afternoon on the account's own history: 90 days of search terms, sorted by cost, read with the four families in mind. The seeded lists catch the generic waste; the history read catches your niche's specific vocabulary, which no template ships. An afternoon total, and the weekly hour keeps it compounding from there.

Do negatives work in Performance Max now?

Finally, yes. Campaign-level negative keywords rolled out to Performance Max through 2025, and the 2026 updates extended support at account level, with caps tighter than search campaigns allow. Spend the limited slots on your most expensive waste families rather than duplicating the full search library, and combine them with brand exclusions, still the bigger lever for the overlap problem described in the Performance Max guide. The November 2025 rollout of PMax's channel report also means you can finally read PMax search terms with Search-grade granularity, so the weekly review now covers every campaign type in one sitting.

What results should the system produce?

Two visible outcomes within a quarter. First, the waste share of spend drops: from a 21%-class leak toward low single digits, which on the 2026 cross-industry median CPL of $66.69 means the same budget buys measurably more leads. Second, CPA stabilizes, because Smart Bidding stops learning from junk clicks. The industrial account in our lead-cost teardown added 267 German-language negatives in its first month as part of the rebuild that took CPL from €126 to €78. For the full context of where negatives fit among the other fixes, start with the self-audit walkthrough.