Purpose of this document:
To explain how to benchmark a campaign’s true cost-per-conversion and cascade through four buckets to pause keywords that are either converting too expensively, spending past their fair shot without converting, or showing impressions without earning clicks.
Why this procedure exists
The common rule “pause keywords with 12+ clicks and zero conversions” has two weaknesses:
- A fixed click count doesn’t adapt to CPC range. In a $50 CPC niche, 12 clicks is $600 of runway. In a $5 CPC niche, 12 clicks is $60 and the rule is too cautious.
- It only catches non-converters. Keywords that convert but at 2–3× the campaign’s true CPA quietly drag the average up and eat budget that would convert better elsewhere.
Benchmarking against the campaign’s actual Cost/Conv. fixes both. The benchmark adapts to the niche automatically, and the cascade catches expensive winners as well as non-converters. Done weekly, this is the highest-leverage 15–20 minutes of the optimization rhythm.
When this step doesn’t apply
Skip until the campaign has a stable benchmark. That usually means 30+ days of data and at least 5–10 conversions distributed across multiple keywords. With one converting keyword, the benchmark IS that keyword’s CPA, and the cascade can’t tell expensive winners from average performance. Wait until conversions spread across the keyword set before treating the benchmark as meaningful.
How it works
The procedure benchmarks against the campaign’s actual cost-per-conversion — the Cost/Conv. across every enabled keyword that has ever converted. That number is the campaign’s true CPA, not a guess, and it adapts automatically to the niche’s CPC range.
The cascade has four steps, each catching a different failure mode:
- Step 1 establishes the benchmark; nothing gets paused here.
- Step 2 pauses expensive winners — keywords converting at ≥ 1.5× the benchmark.
- Step 3 pauses non-converters past their fair shot — zero-conversion keywords whose spend exceeds the benchmark Cost/Conv.
- Step 4 pauses bad-relevance keywords — 20+ impressions with zero clicks.
Pause, never delete. Delete throws away historical data permanently; pause leaves it queryable for the previously-converting check on later cycles, and the status is reversible if volume drops.
The Process
-
Establish the benchmark. Open the campaign → Keywords in the left sidebar. Set filters:
- Date range: All time
- Keyword status: Enabled
- Conversions > 0
Look at the Total: Filtered Keywords row at the bottom of the table. The Cost / Conv. value on that row is the benchmark — the campaign’s true CPA. Keep it handy; Steps 2 and 3 reference it.
-
Pause expensive winners. Same filter as Step 1. Sort by Cost / Conv. (descending) and apply the rule from the top:
Cost / Conv. vs benchmark Historical conversions Decision ≥ 1.5× benchmark 5+ PAUSE — dragging campaign CPA up ≥ 1.5× benchmark 1–4 Flag, monitor 2–4 more cycles 1.0×–1.5× benchmark any Keep, monitor < 1.0× benchmark any Star, keep -
Pause non-converters past their fair shot. Change filters to:
- Date range: All time
- Keyword status: Enabled
- Conversions < 1
Sort by Cost (descending) and apply:
Total cost vs benchmark Cost/Conv. Decision > benchmark Cost/Conv. PAUSE — afforded a conversion at the campaign’s average rate without producing one ≤ benchmark Cost/Conv. Keep — hasn’t had its fair shot yet -
Pause keywords with impressions but no clicks. Change filters to:
- Date range: All time
- Keyword status: Enabled
- Clicks < 1
- Impr. > 20
Every keyword in the filtered list: PAUSE, no further analysis. 20+ impressions with zero clicks is a CTR / Quality Score signal that the ad isn’t compelling for whatever this keyword triggered — ad-copy mismatch, wrong ad group, low Quality Score, or off-target searcher intent.
-
Pause in Google Ads. For each flagged keyword: click the row → Status dropdown → Paused → Save. Or batch-pause by checking multiple rows → Edit → Change status → Pause. Never delete.
Edge cases
Edge case 1: previously-converting keyword in a cold streak
A keyword flagged by Step 2 with 15+ historical conversions may be in a temporary cold streak rather than structurally underperforming. Flag it in the optimization log, monitor for 1–2 more cycles before pausing. Common causes: seasonality (HVAC in October, lawn care in January), competitor surge in the geo, ad fatigue (refresh in the monthly review), landing page regression (run Step 7 of weekly).
Edge case 2: campaign on Smart Bidding
The cascade still applies. Smart Bidding doesn’t pause keywords for you — it bids algorithmically given the keywords you’ve left enabled. A bleeding keyword under Smart Bidding feeds noisy training data, so pausing is arguably more important, not less. Under Target CPA, an “expensive winner” is the algorithm doing what you told it to; pausing lets it find better-fit traffic.
Edge case 3: benchmark dominated by one keyword
If a single keyword accounts for 60%+ of campaign conversions, the benchmark may be skewed toward that keyword’s CPA. Use the previous month’s benchmark as a sanity check, or wait one cycle and re-establish before pausing aggressively.
Edge case 4: paused a keyword, volume dropped
If you pause a keyword and conversion volume drops noticeably the following week, un-pause it. The “Paused” status is reversible. This is also why you never delete — the historical data stays queryable.
Edge case 5: thresholds feel too aggressive for the first few cycles
Run the cascade conservatively for the first 1–2 cycles: 2× benchmark instead of 1.5× for Step 2, higher historical-conversion bar for the previously-converting exception. Once the procedure has demonstrated it doesn’t break volume, tighten to the defaults.