Worked example
Two audited changes, both proven not to hurt the campaign.
A worked example from a live home water-treatment account. Every figure below was pulled directly from the account's own reporting on July 16, 2026 (data through July 15) — nothing is modeled or estimated. The client is anonymized and the Google account and campaign identifiers in the receipts are masked.
What happened
Twice this summer, a human approved a case the engine had written up: a real query, a real dollar figure, zero leads. Here is the full life of each change.
The engine flagged a specific wasted search.
A monitoring engine reads the account's Google Ads search-terms report every day and looks for money going to searches that never turn into a lead. A real query, a real dollar figure, zero leads to show for it. It does not change anything on its own — it writes up the case and puts it in front of a human.
A person approved it.
Not the engine — a named operator clicked approve, and that approval is timestamped in the record. If it isn't approved, it doesn't ship.
The change was made through an audited pipeline.
Google's own API executed it and handed back a receipt — a Google-issued resource ID. We store that receipt so we can point to it later.
We read the change back, then watched for collateral damage.
The moment the change was executed, the system re-read it from Google to confirm it actually took effect. Then we watched the campaign for two weeks to make sure blocking that one search didn't quietly starve the rest of it.
The point of this page is not the size of the two dollar figures — they are small and we show them exactly as they are. The point is that every step is logged, approved by a human, confirmed against Google, and monitored afterward for collateral damage.
The numbers
Two blocked searches, and what the campaign did next.
Blocked the search “how to get iron out of well water”
A do-it-yourself, how-to search. The person typing it wants instructions, not an installer. It had leaked into the “Well Water Treatment” ad group.
- Match type / scope
- Exact match, one campaign only — surgical: it blocks only this precise phrase.
- Approved by
- The account operator, June 10, 2026, 16:28
- Executed
- June 10, 2026, 16:28, via the Google Ads API. Read back and confirmed the same minute.
- Google receipt (masked)
customers/••••••••/campaignCriteria/••••••••••~••••••••••••
across its entire recorded history, last seen May 8, 2026
checked through July 15
Window: 14 calendar days on each side of the change, counting only days with recorded data — 12 days before, 14 days after.
| Per day | Before | After | Change |
|---|---|---|---|
| Impressions | 52.2 | 62.1 | +19% |
| Clicks | 4.83 | 3.64 | −25% |
| Cost | $14.14 | $14.03 | −1% (flat) |
Read: Clicks falling here is the goal, not a side effect: a negative keyword exists to stop paying for clicks that can never become customers. The clicks that disappeared were do-it-yourselfers who were never going to book an install — while impressions rose and daily spend held steady, so the campaign's reach didn't collapse. Same budget, aimed at fewer dead ends.
Blocked the search “best alkaline water machine for home”
A product-category search for a product this business does not sell — countertop alkaline machines, not whole-house or well-water systems.
- Match type / scope
- Exact match, one campaign only.
- Approved by
- The account operator, July 3, 2026, 16:50
- Executed
- July 7, 2026, 00:20, via the Google Ads API. Read back and confirmed the same minute.
- Google receipt (masked)
customers/••••••••/campaignCriteria/••••••••••~••••••••••••
across its entire recorded history, last seen June 25, 2026
checked through July 15
Window: 14 recorded days before, 8 recorded days after.
| Per day | Before | After | Change |
|---|---|---|---|
| Impressions | 54.2 | 65.8 | +21% |
| Clicks | 4.79 | 3.63 | −24% |
| Cost | $15.96 | $15.33 | −4% (flat) |
Read: Same picture as Action A — and again, the click drop is the point: those were shoppers for a product this business doesn't sell. Reach up, spend flat, wasted clicks gone. The over-block monitor completed its 7-day measuring window on July 15 and returned its verdict.
The system behind it
Approved, confirmed against Google, and reversible.
Approved by a human, logged with a Google receipt.
No change happens without a named person's timestamped approval, and each one stores the resource ID Google returns — a receipt we can point to later.
Read back after it's made.
The moment a change is executed, the system re-reads it from Google to confirm it actually took effect — not just that the request was sent.
Watched for over-blocking, and reversible.
For two weeks after, the engine compares the campaign's reach and spend before vs. after. If blocking a search quietly choked the campaign, it flags it and offers a one-command rollback. Both changes above are still reversible.
Methodology & caveats
What these numbers do — and do not — say.
These are exposure figures, not savings or return-on-investment. When a zero-lead search is blocked, the budget it had been drawing does not disappear — it reallocates to other searches in the same campaign. So “$2.97” and “$11.18” describe money that had gone to a search that never produced a lead; they are not a claim that the business is now $14.15 richer. The revenue effect of the reallocation is unknown and not asserted here.
Windows are stated on every number and count only days with recorded data. Google's search-terms and daily reports have gaps; we sum the days that exist rather than dividing by a flat calendar count, which is why “12 days before / 14 days after” is uneven.
Conversions are shown but never used to pass or fail a change. At this account's volume, lead counts per day are too sparse to be a reliable signal, so the over-block check is based on reach and spend, and lead figures are reported for context only.
Attribution lag. A click today can register as a lead days later, so the most recent days of any “after” window will tend to under-count leads until Google finishes attributing them.
Concurrent changes. These two negatives were not the only things happening in the campaign. During the same windows, several other routine negative keywords and an impression-share flag were also applied to this campaign. So the campaign-level before/after movement above cannot be pinned on any single change — it reflects everything happening at once. What is cleanly attributable to each blocked search is its own after-spend of $0.00: an exact-match block deterministically stops exactly that phrase, and both stopped.
Source: the account's live Google Ads database, queried read-only on 2026-07-16 (data through 2026-07-15).
Want a walkthrough of what this looks like on your account?
We'll show you the approval queue, the change log, and the over-block monitor running against a real account — the same machinery behind every figure above.