Draft · not for publication · Imran to edit before launch
Experiment 001 · 07 / Building
Unpinned, AI-drafted headlines cut CPL from $71 to $54
I stopped hand-writing and pinning Google Ads headlines on a lead-generation account. Costs rose for two days, then fell for five weeks.
[ TESTED ] ·
Hypothesis
Pinned, hand-written headlines were stopping Google from finding better combinations. More unpinned variants, written from the account's own data, should lower cost per lead.
Method
One lead-generation Google Search account. Claude was given the landing page, the ideal customer profile, notes on what a good lead looks like and six months of search term reports, and asked for 15 headline variants per ad group. They were loaded with light edits. Every pin was removed except legal text, and Google's asset combinations were left alone.
Result
Week one, CPA rose 10–12% for the first 48 hours. Week two, CTR was up 18%. By week five, CPL was $54, against a $71 baseline that had held for about four months.
Interval
One account, one five-week test, no holdout group.
Decision
Unpin everything except legal text. Write more assets rather than better ones.
What did not work
The first two days. Cost per acquisition went up 10–12% while the system relearned, which is exactly when most people would roll the change back.
For months I wrote 14 headlines per ad group by hand, like a good little PPC manager. Pinned the brand. Pinned the offer. Kept the keywords where I wanted them.
Cost per lead sat at $71 for about four months.
So was the copy the problem? Not really. The problem was control. Pinning told the system what I believed converted, based on how Google Ads worked a few years ago. It stopped the algorithm testing anything I hadn’t already decided.
The change was simple: more variants, written from the account’s own search terms and lead quality, and no pins except where the law needs them.
What would I do differently? Run it with a holdout from the start, so the drop in CPL can’t be put down to the season. One account is a good signal, not proof.
Most accounts I audit are still pinning three headlines out of habit. That habit is expensive.