The ad you are scaling is older than you think
We took fifteen Australian DTC brands and looked at what is actually live in Meta's ad library. Not what they say they run. What is running.
How old is the oldest ad still running?
| Age of the longest-running live ad | Brands | Share of those advertising |
|---|---|---|
| Over 2 years | 1 | 8% |
| 1 to 2 years | 1 | 8% |
| 6 to 12 months | 3 | 23% |
| 3 to 6 months | 5 | 38% |
| Under 3 months | 3 | 23% |
The median brand's longest-running ad had been live 162 days - about five and a half months. The oldest had been running since May 2024.
It is not that they cannot make enough ads
The obvious story would be that these brands are starved of creative. The data says otherwise: the median brand in this cohort had 711 ads live on the day we looked, and one had 7,949. Volume is not the constraint.
What is happening is narrower and more expensive. Most of that volume is catalogue and dynamic product advertising - the machinery that runs itself. The hero creative, the ad actually carrying the account, is a small handful of assets, and those sit untouched for months while everything around them churns.
Seven of the thirteen longest-running ads were static images. The single oldest ad in the whole cohort, live for two years and four months, is a static.
Why the hero ad never gets replaced
We cannot reshoot every time we want to test something.
So the refresh waits for the next shoot, the next shoot waits for budget, and the ad already running keeps running by default.
It is still working, so why touch it?
Because it stops. In the accounts we audited, one ad absorbed nearly a quarter of the entire budget while delivering a worse cost per result than a smaller cut of the same idea.
We do not know what to make next.
Every new test changes the cast, the location and the edit at once, so nothing is learned. That is what the format benchmark is for.
How this was put together
What we did and did not do
The cohort was drawn from a commercial Shopify store database, filtered to Australian single-brand stores turning over more than US$250,000 a month with a Facebook page on record, then read across apparel, activewear, swimwear, lingerie, beauty, underwear and menswear. Multi-brand retailers and the Australian sites of global brands were excluded, because neither behaves like a DTC advertiser.
For each brand we took the longest-running ad currently live in Meta's ad library. Two of the fifteen had nothing live and are excluded from every percentage above rather than counted as zero.
No brand is named here. These are public ad-library records, but a page that points at named companies for running an old ad is a page that costs us the conversation with them. The distribution is the finding; the names are not.
The cohort skews toward fashion, because Australian DTC does. Read it as a fashion-weighted DTC sample rather than a statement about every category.
Questions, answered.
Is a long-running ad automatically a bad ad?
No. An ad that has been live for a year may well be the most efficient thing in the account. But it is being shown to an audience that has seen it repeatedly, and the cost of that shows up gradually as rising frequency and drifting cost per result rather than as a single bad week.
How do I check my own?
Meta's ad library is public. Search your brand, sort by what is still active, and look at the launch dates. If your best performer launched more than a quarter ago and nothing is briefed to replace it, that is the finding.
Why only 15 brands?
Because each one was checked by hand against the library rather than scraped in bulk, and because a small cohort you can describe precisely is worth more than a large one you cannot. The method above is reproducible if you want to widen it.
Does this mean I should replace ads more often?
It means you should have a replacement ready before you need one. Replacing a winner on a schedule is not the point; being able to replace it the week it stops working is.
Find out how old your best ad is
Send us the brand and we will pull the same view of your account, then build the replacements the format benchmark says should beat it.