Exclude four searches that never sell from Search: coffee subscription
Searches for exactly these stop showing this campaign's ads. Nothing else changes.
If you approve, each month
About $15,000 more in sales
Where $14,940 comes from: the $5,695 they took would have brought about $13,943 in sales on the campaign's other searches, at what the next dollar there brings. That is over 28 days; a month is about $14,940.
About $9,000 kept after the goods at your 60% margin.
Spent in 28 days
Why, in Steve's words
These four searches took $5,695 in 28 days on 4,704 clicks with no orders, where the campaign's rate would have brought about 284. None of them is someone looking for a roasted coffee subscription: job seekers, people after free samples, home cold brew and pod buyers.
Why it may be wrong. It comes from the last 28 days of Google's search terms report, and the weeks ahead may differ.
How Steve decided
- What he saw
"free coffee samples" $2,265; "how to make cold brew at home" $1,618; "nespresso pods subscription" $1,294; "coffee shop jobs" $518 (spent in 28 days).
Last 28 days, from Google's search terms report. $5,695 on 4,704 clicks, no orders.
- What he concluded
These searches spent money and brought no orders, so excluding them leaves it for the searches that do.
- How sure
The numbers back this: at the campaign's own rate, each of these searches would bring no order on its spend by chance less than 1% of the time.
About $14,940 more in sales a month, one figure with no range around it.
Orders from the last 2 days can still arrive, so the newest figures are not final.
- What would change his mind
If one of these searches had brought 1 order, there would be no case to exclude it.
A tracking check that fails on these networks: the counts would be wrong, and Steve holds every decision on them.
- After you approve
Steve skips the first 2 days while the network settles, gives the first read on Nov 19 and the full 14 days on Nov 26, under Results.
It worked if the cost per order on this campaign on Google is more than 5% lower than in the 14 days before, with at least 20 orders on each side. Within 5% is no clear change.
Your goal is ROAS, and Results reads every change on the cost per order, which has enough results to read in two weeks. If orders get cheaper while ROAS falls, it is not counted as having worked.
Steve checks his own estimate against what came: $14,940 more in sales a month.