/projects//google_ads/queries/ ).
1. The Broad Match Expansion Dilemma
Google's machine learning bidding models push advertisers aggressively toward broad match keywords. While broad match expands reach, it also matches against conversational, informational, and completely irrelevant queries that burn through ad budgets without producing conversions.
LIVE PLATFORM SCREEN & TELEMETRY
SEO Runway Google Ads Search Queries & Negative Keyword Builder — Active Workspace
Below is the exact production screen with live parameters and dummy data illustrating the diagnostic findings and remediation buttons.
2. Isolating Waste with SEO Runway Search Query Mining
- Open Screen (Google Ads → Search Queries):
- Apply the filter:
Conversions = 0andCost > $35. - Inspect the search terms table sorted descending by Cost. Identify irrelevant terms containing words like 'free', 'diy', 'definition', or competitor employee logins.
- Select the offending rows and click 'Add as Negative Keyword'. Choose whether to apply at the Campaign or Account Negative List level.
3. Measuring Immediate CPA Reduction
Within 14 days of pruning irrelevant search terms, accounts typically see a 25% to 40% reduction in Cost Per Acquisition (CPA) and an immediate improvement in overall ad relevance.
Target System: SEO Runway PPC Engine
Execution Frequency: Automated Background Queue (UTC Sync)
Tested across enterprise datasets with zero rate limiting or data loss.
4. Why Negative Keywords Need an Engine, Not a One-Off Clean-Up
Wasted spend in Google Ads rarely comes from one bad keyword. It comes from thousands of irrelevant search terms, each costing a little, accumulating every day as broad match and automated campaigns expand into new queries. A one-time negative keyword clean-up helps for a few weeks, then waste creeps back. A negative keyword "mining engine" is a repeatable process: it regularly pulls search term data, finds patterns of waste, turns them into negatives at the right level and match type, and measures the savings.
The engine matters more than ever because of match type changes. Broad match now relies heavily on Google's understanding of intent, and phrase match covers more variations than it once did. Performance Max campaigns surface search terms only partially. Without a systematic mining process, a growing share of spend goes to queries nobody chose deliberately.
5. The Data You Need
- Search terms report for every Search and Shopping campaign: search term, matched keyword, match type, campaign, ad group, impressions, clicks, cost, conversions and conversion value.
- A sensible date range: the last 30 days for active accounts, 90 days for lower-volume accounts so patterns are visible.
- Performance Max search term insights (search categories and the available search term data) to spot irrelevant themes.
- Your existing negative lists, so you don't duplicate or contradict them.
- Business context: products and services you don't offer, locations you don't serve, and customer types you don't want, such as job seekers, students or DIY users.
6. Step 1: Find Waste with N-Gram Analysis
Reviewing search terms one by one doesn't scale. N-gram analysis breaks every search term into single words (1-grams), two-word phrases (2-grams) and three-word phrases (3-grams), then sums cost, clicks and conversions for each across all terms. A word like "free" might appear in 400 different search terms that each cost a few pounds; individually they look insignificant, but the 1-gram total reveals that "free" cost £1,800 last month with zero conversions.
Sort n-grams by cost with zero or very low conversions, and by cost per conversion far above your target. The top of each list is where the biggest, most systematic waste sits.
7. Step 2: Classify Before Adding
Not every low-converting n-gram should become a negative. Classify each candidate:
- Clearly irrelevant: wrong product, wrong service, wrong intent ("jobs", "salary", "course", "template", "DIY" for a service business). These become negatives.
- Wrong location: towns or regions you don't serve. Negative them, or fix location targeting.
- Low intent but relevant: research terms ("what is", "how does") that rarely convert directly. Consider moving them to a separate, lower-bid campaign rather than blocking them.
- Too little data: terms with only a few clicks. Leave them until there is enough evidence.
- Competitor terms: decide deliberately whether to bid on them, rather than blocking or allowing them by accident.
8. Step 3: Choose the Right Level and Match Type
Level
- Shared negative lists (account level): for terms that should never trigger any campaign, such as "jobs", "careers", "free" (for paid-only products) and adult or unrelated terms. Apply the list to every relevant campaign.
- Campaign-level negatives: for terms irrelevant to one campaign but valid for another, for example blocking "emergency" in a routine-service campaign so an emergency campaign captures it.
- Ad group-level negatives: for routing queries to the right ad group within a campaign.
Match type
- Negative broad blocks searches containing all the negative's words in any order. It is good for single words that are always irrelevant.
- Negative phrase blocks searches containing the exact phrase in order. It is useful for multi-word ideas such as "how to".
- Negative exact blocks only that exact query, which is precise for routing between campaigns.
Negative keywords don't automatically match close variants such as plurals and misspellings, so add the variations yourself.
9. Step 4: Check for Conflicts
Before uploading, check every new negative against your positive keywords. A negative that blocks one of your own keywords silently stops it from serving. Google Ads flags some conflicts, but a quick check in a spreadsheet, or a script that compares lists, catches mistakes before they cost you traffic.
10. Step 5: Automate the Routine
A negative keyword engine runs on a schedule:
- Weekly: pull search terms from the last 7–30 days, run n-gram analysis, review the top 20 waste candidates, and add the clear ones.
- Daily for new campaigns: in the first two weeks after launch, or after switching to broad match, review search terms every day.
- Monthly: review shared lists for outdated entries, such as a service you now offer that is still blocked.
- Alerts: Google Ads scripts or third-party tools can flag search terms that exceed a cost threshold without converting.
Platforms that mine search terms automatically and suggest negatives with their cost and conversion data shorten the weekly review to minutes. Whatever tool you use, keep a human approval step, because automated suggestions sometimes flag terms that are low-converting for reasons unrelated to relevance, such as a broken landing page.
11. Step 6: Measure the Savings
Track the cost of the terms you blocked in the period before blocking, then watch overall cost per conversion and conversion volume afterwards. Report three numbers monthly: spend removed from irrelevant terms, change in cost per conversion, and change in conversion volume. If conversions fall after adding negatives, review the most recent additions for over-blocking.
12. Starter Negative Themes by Business Type
- All businesses: jobs, careers, hiring, salary, internship, course, training, degree, definition, meaning, wiki, PDF, template (unless you sell templates).
- Local services: towns and regions outside your service area, "DIY", "how to fix", "parts" (if you only sell the service), "cheap" (if you're premium).
- B2B software: "free download", "crack", "open source" (if paid only), "login", "support number" (existing customers), "alternative to [your brand]" (unless deliberate).
- E-commerce: "used", "second hand", "repair", "manual", "wholesale" (if retail), and competitor brand names you don't stock.
- Professional services: "salary", "qualification", "exam", "free advice", "template".
Always review a starter list against your actual offer. "Free" is a costly negative for a business that offers a free consultation, and "repair" is essential for a repair shop.
13. Performance Max and Negatives
Performance Max campaigns now support campaign-level negative keywords, alongside account-level negative keyword lists and brand exclusions. Use account-level lists for universal exclusions, and review PMax search term insights regularly for irrelevant categories. Because PMax reporting is less detailed than Search campaigns, combine it with conversion quality checks (lead quality, order value) to catch waste that search term data doesn't reveal.
14. Common Mistakes
- Single-word broad negatives that are too aggressive, for example blocking "cheap" for a value brand, or "near" which removes every "near me" search.
- Negatives that conflict with your own keywords, silently stopping them.
- Judging terms on too little data and blocking queries that would have converted.
- Adding negatives only at ad group level, so the same irrelevant term keeps appearing in other campaigns.
- Never reviewing old negatives after the business adds new products or locations.
- Ignoring conversion quality: a term may "convert" with junk leads that should still be excluded.
15. Negative Keyword Engine Checklist
- Search term data pulled for every Search and Shopping campaign on a schedule.
- N-gram analysis to find systematic waste.
- Candidates classified before adding.
- Shared lists for universal exclusions; campaign and ad group negatives for routing.
- Match types chosen deliberately, with variants added manually.
- Conflicts with positive keywords checked before upload.
- Weekly reviews, daily for new or broad match campaigns.
- Savings, cost per conversion and conversion volume reported monthly.
16. A Worked Example
A commercial cleaning company spends £8,000 a month on Google Ads with mostly broad match keywords. A 90-day n-gram analysis shows "jobs" and "vacancies" costing £640 together with no conversions, "domestic" and "house" costing £910 with conversions only for residential customers the company doesn't serve, and several towns outside its service area costing £520. "Cleaning products" and "supplies" cost another £430 from people looking to buy equipment. Altogether, around £2,500 over three months, or roughly 10% of spend, went to clearly irrelevant searches.
The team adds "jobs", "vacancies", "careers", "products" and "supplies" to a shared negative list, adds "domestic" and "house" as campaign negatives (with phrase match for "house cleaning" to avoid blocking "warehouse"), and tightens location targeting. Over the next month, cost per lead falls and lead volume stays steady, because the budget that went to irrelevant clicks now reaches relevant searches. The weekly review then takes about fifteen minutes, focused on new patterns rather than repeating the same clean-up.
A negative keyword engine is one of the highest-return routines in paid search. It takes minutes a week once set up, it compounds as irrelevant patterns are blocked for good, and every pound it saves is reinvested in searches that actually bring customers.
Want the complete procedural runbook?
Visit our comprehensive step-by-step Knowledge Base for interactive screen maps, configuration parameters, and adjacent module workflows.
Open Knowledge Base Hub →

Incredible breakdown. We implemented this exact workflow last week and already saw a noticeable shift in our telemetry. As a VP Marketing, CloudScale, this provides exactly the architectural depth we look for.