$48 for 483 clicks and zero signups: how to tell you're buying accidental taps
I launched my first Google Ads campaign two weeks after launching my product
(a personal finance web app). Day two, the dashboard looked incredible:
417 clicks, 8.3% CTR, nine cents a click. I've read enough marketing
threads to know 8% CTR is "your ad is amazing" territory.
My database had one user in it. Me.
Here's the forensic trail, because the tells generalize to any campaign,
and I'd have saved $48 if someone had shown me this list a week earlier.
The tells
1. The device mix was wrong. 39% of my clicks came from tablets.
Tablets are low-single-digit percent of web traffic almost everywhere.
There is exactly one place tablets dominate: kids' games and free
utility apps, where a banner sits under a thumb that was aiming at
"continue."
2. Desktop got 46 impressions and zero clicks. I provide mortgage and
debt calculators. People research those at a desk. A campaign that
can't find a single desktop click isn't finding researchers, it's
finding screens.
3. "TV screens" appeared as a device row. Search ads do not run on
televisions. If your "search" campaign has a TV row, some of your money
is buying video or display inventory, full stop.
4. The metrics vocabulary changed. My summary panel reported CPM
and "co-viewed" metrics. Search campaigns price per click. The moment
your dashboard starts speaking in impressions-per-thousand, you're in
the display world, whatever the campaign is called.
5. The clicks were impossibly cheap. $0.09 in a finance vertical
where real CPCs run $0.50–$5.00. Cheap clicks feel like winning. Ask
who sells a click for nine cents, and who's clicking it.
What actually happened
I'd built a Performance Max campaign (Google's wizard steers you there
hard) with Maximize Conversions bidding — and I had zero recorded
conversions, because my conversion tracking uploads offline on a weekly
schedule and nothing had converted yet.
So I'd told an optimizer "maximize a thing you have never seen." Given
a daily budget and no signal, it did the only thing it could: bought
the cheapest inventory on earth, which is banner slots inside mobile
games. Every number above follows from that one decision.
That's the part I want to underline: PMax wasn't broken. It did
exactly what I asked. Maximize Conversions with no conversion history
is not a strategy, it's a blank check.
The rebuild
I paused the campaign and rebuilt as boring, hand-controlled Search:
- "Create a campaign without guidance" (skips the wizard's PMax funnel)
- Search only: Display Network and search partners unchecked (the
Display checkbox is how "Search" campaigns leak into game banners) - Maximize Clicks, not conversions, with a $1.50 CPC cap
- Phrase-match keywords only, four tight ad groups, each landing on the
exact matching page - Every AI-expansion toggle off (AI Max, text customization, final URL
expansion)
Same product, same budget, three days later: 2.1% CTR at $0.59 a
click, 85% mobile / 14% desktop / 2% tablet — a human-shaped device
mix. My best keyword ("biweekly mortgage calculator") clicks at 9.4%,
and now that's a real number, because it's real people typing a real
query.
One more landmine from the rebuild: when I exported my keyword report,
Google's setup flow had added ~35 broad-match keywords, things like bare "minimum payment" and "minimum credit score
for credit card" (a completely different, much more expensive
searcher). They'd eaten 72% of one ad group's impressions at 0.8% CTR.
After saving any campaign, open the Keywords tab and delete everything
you didn't type.
The honest part
Signups from the rebuilt campaign so far: also zero. About a
hundred genuine search clicks, no conversions yet. At typical
visitor-to-signup rates that's still within noise, but the question has
moved somewhere more uncomfortable: it's no longer "can I buy the right
traffic" — I can — it's "does my site convert it." That's the next
thing I'm instrumenting.
If you're about to run your first campaign: check your device mix on
day one. It's the fastest lie detector Google gives you.
