Doctor Ads Join the Lab
The Lab  /  Original Research

What 9.5 million search queries admit under examination.

Findings from the accounts behind $770M+ in ad spend overseen: 9.5 million search queries, read line by line. Names never published.

Join the Lab Read the findings ↓
The Corpus  /  At a GlanceAnonymized
1,950
campaigns
9.5M
search queries
116K
ads
quizedtechqrpdftrackdocsvisautildentarb
The corpus, at a glance
No account is identifiable.
Findings  /  What the numbers admit

Six things the platform will not tell you.

Mobile takes 72% of the budget. Desktop takes the conversions.

mobile · 72% CVR −28.6% ↓ desktop · 28% converts better ↑
Sep 2024 – Feb 2025 · the clear majority
What it means for youCheck your device split against margin, not against traffic.
What we do not claim: not every account; a minority lean the other way.

Smart Bidding keeps its promise one time in three.

tCPA campaigns within ±10% of target tROAS: 27 median shortfall −14%
1,950 campaigns · 7 verticals
What it means for youTargets are a starting bid, not a contract. Audit the gap monthly.
What we do not claim: not that automation is useless; that it is unsupervised.

1% of search terms carries 62% of the conversions.

top 1% of terms → 62% of conversions the other 99% → 38%
9.5M search terms
What it means for youGuard the head terms with your life. The tail is where budgets quietly die.
What we do not claim: shares vary account to account (28 to 83%).

Google's "Poor" ads out-click "Excellent" three to one.

9.73%rated "POOR" 2.93%rated "EXCELLENT" labels point backwards
116,000 ads · Sep 2024 – Feb 2025
What it means for youDo not rewrite ads to please a label.
What we do not claim: CTR is not profit; the point is the label, not the metric.

"Excellent" ad strength beats "Average" 55% of the time. A coin flip.

55% a coin flip "Excellent" won on CTR barely half the time
accounts with enough ads of both grades
What it means for youSpend the effort on offers and terms, not on chasing grades.
What we do not claim: an early signal, but a striking one.

The two levers almost nobody touches.

2.8% median budget inPerformance Max 0.7% campaigns withad scheduling
across the corpus
What it means for youThe crowd ignores whole control panels. Check yours.
What we do not claim: underused is not the same as underperforming.
Methodology

How these numbers are made.

raw accountsanonymizeaggregatecross-check ×3✓ names discarded
→ Code does the counting; conclusions survive re-runs.
→ Names are never published. Accounts live as codes.
→ Conversions are never compared across accounts; each business counts its own.
→ The sample is honest about itself: these are accounts that reached a consultant.
In Preparation

What the corpus is still admitting.

Preprint

The Promise Gap of automated bidding

How far Smart Bidding lands from its own targets, across 1,950 campaigns.

Preprint

The Zipf structure of wasted spend

Where the money goes to die in the long tail, measured across millions of queries.

Annual

The annual industry report

The full picture, once a year, from the accounts no single advertiser can see.

The Ad Profit Lab

Each new finding, in writing.

No schedule, no noise. When the data admits something worth your time, you will be the first to read it, published as a LinkedIn newsletter.

Join the Lab

A LinkedIn newsletter. One click to subscribe, leave anytime.

The next finding, the day it lands the chart, the number, the fix no schedule, no noise
Profit Forensics

Want this run on your account?

One account, one week, one signed verdict. The same method, on your numbers.

Request an examination$3,000 · fixed