Why most of our ads lose, and why that's fine

Last year we shipped 1,140 new ads across our clients. 168 of them became winners. That's about one in seven, and it's the number we're proudest of.

Losing is the job

An agency with a 60% win rate isn't testing anything. It's making small changes to ads that already work, which feels safe until those ads wear out, and they always do. Winning ads lose steam after six to ten weeks as the same people see them again. If nothing new is ready, cost per purchase climbs and nobody can say why.

So we test ideas that might fail: a new angle, a different kind of person on camera, a claim nobody in the category makes. Most of them lose. The ones that win become the account's next three months.

What a losing week costs

Each test gets a fixed budget, usually $500 to $1,000 over four days. A client on our Growth plan tests about four ads a week. If three lose, that's roughly $2,000 spent learning what doesn't work, against a monthly budget near $100k. That's the price of finding the next winner, and it's capped by design.

The expensive losses are the ones nobody switches off. An ad that's "fine" for six weeks costs far more than four honest test losses, because it quietly takes budget from something better.

How to read a week of losses

Look at why they lost, not how many. Our scorecard puts one line next to every switched-off ad: "hook too slow," "discount brought one-time buyers," "nobody believed the claim." Three losses for three different reasons means the account is learning. Three for the same reason means the next brief needs to change.

What we tell clients in month one

Expect most new ads to lose. Expect the first real winner around day 19. And expect us to switch things off quickly, including ads we liked.