ADFIXER
Platform / Amazon
Keyword funnels

Discovery to conversion, fully automated.

Search terms move through a four-stage funnel. Converters get promoted, waste gets negated, and match types tighten as confidence grows.

Funnel stages
4

Funnel stages

Conversion window
30d

Conversion window

Automated movement
24/7

Automated movement

Keyword funnelLive account
DiscoveryAuto + Broad
1,840 terms
ResearchPhrase
412 terms
PerformanceExact
96 terms
RankingExact / top
24 terms
promote ↑negate ↓match type tightens left to right
Four stages

Every keyword gets a path to prove itself.

Discovery stays open so the account keeps learning. Everything downstream of it is earned, which is why the spend concentrates without you writing a rule for each term.

Stage progressionAuto to exact
DiscoveryAuto + Broad
1,840 terms
ResearchPhrase
412 terms
PerformanceExact
96 terms
RankingExact / top
24 terms
promote ↑negate ↓match type tightens left to right
Discovery
Auto + Broad
Surface every search term that could convert, on a fixed learning budget.
Research
Phrase
Isolate the converters and decide which terms have earned a graduation.
Performance
Exact
Tighten targeting around proven converters and price them by margin.
Ranking
Exact, top of search
Push the strongest performers with placement modifiers.
Automatic optimization

Four moves, running continuously.

Promotion, negation, match-type changes, and budget allocation all run from conversion data. No weekly report housekeeping.

Promote

Converters graduate

A term that clears its order and confidence threshold moves to the next stage with a tighter match type.

Block

Waste is negated

Spend without orders past the threshold is negated upstream, so the same query cannot be bought twice.

Refine

Match types tighten

Broad becomes phrase, phrase becomes exact, and exact earns placement modifiers as evidence accrues.

Allocate

Budget follows proof

Discovery keeps a fixed learning budget while the rest concentrates on terms that have already converted.

The evidence

Terms graduate on proof, not opinion.

Each move carries the numbers that justified it. When a term is promoted you can see the orders behind it, and when one is blocked you can see exactly what it cost before it was cut.

  • Explicit thresholds for orders, clicks, and confidence at every stage
  • Negatives applied upstream so campaigns never bid against each other
  • Every graduation and block written to the same decision ledger
Stage 1 to 2warehouse label printer14 orders / 340 clicks, CVR 4.1%Promoted
Stage 2 to 3thermal printer shipping9 orders, ACoS 21% under targetPromoted
Stage 3 to 4zebra label printer 4x6Top-of-search modifier +32%Ranking
Negatedfree label template$362 spend, 0 orders in 30 daysBlocked
Heldlabel printer bluetoothSample too small, 60 clicksLearning
Against the manual routine

What the funnel replaces.

Most accounts run some version of this by hand, in a search-term report, once a week, when someone has time.

 Manual search-term reviewKeyword funnels
CadenceWeekly, when someone gets to itContinuous, on every data refresh
Promotion ruleAnalyst judgement, rarely written downExplicit order and confidence thresholds
NegationAdded locally, often duplicatedApplied upstream across the structure
Match-type changesBatch edits in bulk uploadsTightened per term as evidence accrues
Audit trailSpreadsheet history, if anyEvery move logged with its reason
4
Funnel stages
30d
Conversion window
70%
Spend on proven converters
24/7
Automated movement