A good keyword, a few bad terms.
A broad keyword finds sales and new terms worth harvesting. Two junk terms underneath it drag its ACoS down.
Rule engine · Negation
Stop the silent budget bleed.
Negation rules find search terms that click and never sell, or convert too inefficiently, and negate them. You approve, or it auto-applies.
Negative Keywords
Recommendation mode · 2 rules active · last run 2026-09-05 05:28
Last 30 days
Pending recommendations
38
waiting for approvalPending wasted spend
$612.40
clicks with no return · 30dTotal negative keywords
196
+34 this monthEst. future savings
$430 / mo
if all 38 are appliedRecommendationsKeywordsRules
Export
Search termSpend
cheap plastic bottleWater Bottles | Auto | Discovery
$17.92
Approve
water bottle for dogsWater Bottles | Auto | Discovery
$12.10
Approve
glass bottleWater Bottles | Broad | Research
$28.75
Approve
bottle openerWater Bottles | Auto | Discovery
$15.04
Approve
baby bottleWater Bottles | Auto | Discovery
$12.59
Approve
+ 33 more waiting for approval
The problem
Wasted clicks don't announce themselves.
They hide inside a keyword that's doing its job, and drag its numbers down with them.
You cut the bid, not the cause.
The aggregate looks like a failure, so the bid gets lowered. Now it finds fewer of the good terms too.
Negate the terms instead.
The keyword's real performance shows through. The bid stays where it should, and discovery keeps running.
water bottleBroad · Water Bottles · Discovery · last 30 days
Search termClicksOrdersSpendACoS
insulated water bottle849$109.2015.6%
water bottle with straw525$67.6017.6%
kids water bottle393$50.7023.4%
bottle openerNegated710$92.30no sales
baby bottleNegated600$78.00no sales
Keyword ACoS, all five terms 30.5% Target 25%
Looks like a failing keyword. The bid gets cut, and it finds fewer good terms.
Bid$1.10 → $0.88
Two search terms, thirteen points of ACoS. Not just broad match, either: auto campaigns, phrase and product targeting match to search terms the same way.
How it works
Define waste. Approve it, or let it run.
Three steps. Nothing is negated without a reason you can read.
Step 1
Choose your mode
Recommend, and nothing happens until you approve. Or auto-apply, and it just does it.
Step 2
Define waste with conditions
Clicks with no sales, or sales that cost too much. Stack the conditions that match your margins.
Step 3
Every flag explained
Which term, which campaign, how many clicks, how many orders, which rule. Nothing is negated silently.
Negation mode
RecommendAuto-applyRecommend
Flags land in a queue. Nothing changes until you approve.
Auto-apply
Negatives are added the moment a rule fires. Every one is logged.
Rule Builder
Active
Rule
Zero-sale termsNegative keyword rule
When
Clicks ≥10 AND Conversions =0 Add condition
Then
Recommend negativeasExact
312 search terms evaluated · 4 would be flagged
Every day · 05:28 · from the search-term reportSave rule
Activity Logs
NegationCSV
Rule: Zero-sale termsRule executed · 2026-09-05 05:28
NegationNK
312 search terms · 4 recommended
CompletedApplied negative: cheap plastic bottleApproved · 2026-09-05 08:12
YouManual
Exact · Water Bottles | Auto
CompletedApplied negative: bottle openerApproved · 2026-09-05 08:12
YouManual
Exact · Water Bottles | Auto
CompletedRule: Low CVRRule executed · 2026-09-05 04:00
NegationNK
312 search terms · 1 recommended
CompletedDismissed: glass bottleReviewed · 2026-09-05 08:14
YouManual
Kept · 1 order in 30d, watching
DismissedRule: Zero-sale termsRule executed · 2026-09-04 05:28
NegationNK
304 search terms · 0 recommended
CompletedOur own account
The two rules we run ourselves.
Both in recommend mode, with thresholds set per product against its own conversion rate. One way to use negation, not the only one.
Recommend
Clicked enough, never converted.
Enough clicks to prove it doesn't convert, and no orders at all. What counts as enough depends on that product's conversion rate. Too low and you kill terms that never had a fair run. Too high and you've paid for the answer twice.
Recommend
Converts, but not well enough.
It converts, but too inefficiently to be worth the spend. A rate needs more data to trust than an absence does, so the click floor sits higher than rule one, and the CVR floor is set against what that product normally converts at.
bottle openerWater Bottles | Auto | DiscoveryZero-sale terms
16 clicks · 0 orders · $15.04 · 30d
ApproveDismiss
Nothing to do with water bottles. Gone.
kids water bottle with strawWater Bottles | Auto | DiscoveryZero-sale terms
18 clicks · 0 orders · $23.40 · 30d
ApproveDismiss
That's our product. Not enough evidence yet. Keep testing.
The rule finds it. You decide.
A rule can tell you a term isn't converting. It can't tell you it's seasonal, or that it's clearly your product and you'd rather keep testing than close it off on evidence you don't trust yet.
That judgement is worth keeping, and it costs almost nothing. The recommendation arrives with its numbers, and approving is one click.
Let your rules run tonight.
Set up your first rule in minutes. Wake up to a log of exactly what it did.