We read the picture on every live bath ad in America, then looked at what the busiest advertisers kept and what they killed. We expected a pattern. We did not find one, and here is the table that says so.
You paid a designer or an agency to build last month's bath ads. Half of them are already off.
We count every bath company advertising on Meta in America every week. On 14 September 2026 that was 216 companies running 4,389 live ads. This week we also read the picture on every one of those ads. That is 2,463 different pictures.
Half of them had been running 21 days or less.
So the question worth an hour of your week is simple. What lasts, and what do they switch off fast?
The oldest bath ad in America has been live 847 days. It belongs to a one-man shop. Nobody has looked at it in two years. That is not a winner. That is a forgotten card on file.
So we cut the list down to the companies that manage their ads. Twenty or more running at once, and a median age under two weeks. These 29 companies swap creative every few days.
One runs 350 ads with a median age of three days. Another runs 225 at twelve days.
When a company like that leaves an ad up for a month, it is doing something for them. When they pull one in four days, it was not.
Between them these 29 ran 833 pictures this week. Only 70 had been live 30 days or more.
Here is what we expected to find. Here is what we found instead.
| On the picture | The 70 kept | The 763 killed | Odds this gap is chance |
|---|---|---|---|
| A price or a discount | 16% | 25% | 1 in 9 |
| A before and after | 17% | 12% | 1 in 6 |
| The owner on camera | 3% | 7% | 1 in 3 |
| Someone talking to camera | 13% | 18% | 2 in 5 |
| A person in it at all | 46% | 41% | 1 in 2 |
| Video | 70% | 67% | 7 in 10 |
| A customer on camera | 14% | 13% | 7 in 10 |
| Readable with the sound off | 86% | 84% | 3 in 4 |
Base: 70 pictures kept past 30 days, 763 killed inside 30, by the same 29 companies, week of 14 September 2026. The last column is a Fisher exact test on each row.
Read the last column. Not one row is far enough from chance to call it a finding.
The closest is price, and it still comes up one time in nine by luck alone. We tested eight things. Testing eight things gives you roughly a one in three chance of seeing a result that good with nothing going on.
So we are not going to tell you that survivors are the plain ones. Seventy pictures is too few to carry that sentence, and we would rather say so than sell it to you.
Here is the same week asked a different way, with enough behind it to answer.
On 4 September we recorded 2,045 live bath ads across 861 companies. On 14 September we looked for the same 2,045.
1,529 were still running. 516 were gone. A quarter of the trade's creative switched off inside a fortnight.
We only counted companies we opened in both weeks. An ad we did not go back for is not a dead ad.
That is a fact about your money. If you pay for creative every month, a quarter of it has a two week life in this trade.
Of those 2,045 ads, 640 carried a price or a discount. Two weeks later 68% were still running. For ads with no price it was 78%.
That gap is ten points on 2,045 ads. It is real and it is not luck.
But it is not a fact about the words on your ad. Almost every company in this trade goes one way or the other. It writes prices into everything, or into nothing. Only 18 companies ran enough of both for us to compare a company with itself, and inside those 18 the gap disappears.
So what we measured is a habit, not an effect. Companies that discount are companies that rebuild creative fast. Which one causes which, this week cannot tell you.
We will take that over a rule that falls over the first time somebody checks it.
A viewer can follow 2,102 of the 2,463 pictures with no sound. We would like to tell you those last longer.
We cannot, yet. The picture on each ad only got read for the first time this week. To test it the way we tested price, we need the same pictures found again a fortnight from now.
That run happens on 28 September. The answer goes in that week's edition, whichever way it comes out.
A long-running ad is not proof of a sold job. Nobody outside a company can see its spend or its sales.
What we can see is what a company chose to keep. Among firms that swap creative every few days, that choice is the closest public signal to what worked. We publish the method and every row behind these numbers, and we correct them in public when we get one wrong.
Open your own ad account. Sort your live ads by how long each has been running.
Then ask two things. Which of your oldest ads are old because they work, and which are old because nobody looked? And how many of your live ads carry a price you no longer run?
Neither of those is a question about creative taste. Both are questions about what you are paying to rebuild.
Eleven of the 216 companies in your trade let a homeowner book a time straight from the ad. If you are one of the other 205, the ad that lasts is only doing half its job.
Lovro Lucic, on US hours from Zagreb. Your page, your number and your calendar stay in your time zone. Call +1 (650) 643-4497, or pick a time on the calendar.