The short answer
Look for ads that combine time and repetition. On our board of 2,443 Meta ads with $50k+ in estimated spend, ads with 3 or more variants reach the top tier 35% to 45% of the time, against about 10% across the board. Days live tells you an ad survived. Variants tell you the brand is pouring more into it.
The four signals, ranked by how hard they are to fake
A brand can launch a bad ad in five minutes. What it can't do cheaply is keep that ad live for months, build ten versions of it, and keep feeding it budget. Those behaviors cost money, so they are the honest signals.
| Signal | What it shows | Where you see it | Strength |
|---|---|---|---|
| Days live | The ad survived the brand's own kill rules | Start date in the Ad Library | Strong, but slow |
| Variants | The brand is investing more creative into it | "Multiple versions" and repeated creatives | Strong and current |
| Engagement growth | Budget is flowing right now | Reactions and comments over time | Strong if tracked weekly |
| Estimated spend | Size relative to other ads | Third-party estimate | Good for ranking, not exact |
Here is what the combinations look like on the Ad Radar board: 2,443 active Meta ads with $50k+ in estimated spend, snapshot of October 9, 2026.
| Days live | Variants | Ads | Median estimated spend | Reach the top tier |
|---|---|---|---|---|
| 30 to 89 | 1 or 2 | 615 | $154K | 5.5% |
| 30 to 89 | 3 or more | 86 | $617K | 45% |
| 90 or more | 1 or 2 | 1,518 | $222K | 5.5% |
| 90 or more | 3 or more | 224 | $431K | 35% |
The top tier is the best 10% of the board by winner score, among ads live 21 days or more. Variant count is one of that score's inputs, so the top-tier column is partly by design. The spend column isn't: young ads with three or more variants carry a median estimated spend four times higher than young single-version ads.
Signal 1: days live
Start with the start date. Every ad in the Meta Ad Library shows when it started running, and nothing on the page is more useful. Our board's median winner has been live 120 days, and we break the full distribution down in how long winning Facebook ads run.
But days live has a trap. Old ads are proven, and they are also often coasting. Estimated spend per day drops steadily with age on our board:
$3,098
median estimated spend per day, ads live 30 to 59 days
$1,619
ads live 90 to 179 days
$563
ads live a year or more
An ad that has run for a year is a control the brand trusts. It is not necessarily where the brand is pushing budget today. Treat old ads as proof that a concept works and young, multiplied ads as proof of what the brand is betting on now.
If you have a cat older than seven, you'll want to hear this.
- Est. spend
- $1.9M
- Days live
- 308
- Format
- Video, 130s
Ten months live as a single version. A classic control: a callout to one specific owner, then an expert explanation. Proven, steady, not being multiplied.
Signal 2: variants
When an ad works, brands rarely leave it alone. They add new headlines, new first frames, new cuts, and run them under the same Library ID or as near-copies on new IDs. In the library this shows up as "This ad has multiple versions" or as the same creative repeated across cards.
On our board, 700 of 2,443 ads carry two or more variants. The jump is steep: median estimated spend goes from $177K at one variant to $1.1M at ten or more. A full breakdown is in variant count: the signal that a brand is scaling an ad.
This shade makes my green eyes pop.
- Est. spend
- $853K
- Days live
- 75
- Format
- Video, 61s
- Variants
- 9
Under three months live with nine versions, and the same video runs under at least two different headlines. That is what scaling looks like from the outside.
Signal 3: engagement growth
Reactions, comments and shares are public on most ads once you open the post. The absolute count means little, because a $5K ad with a viral comment section can out-react a $500K ad. The growth rate means a lot. An ad that adds hundreds of reactions a week is getting delivery, and delivery costs money.
You can track this by hand: note the reaction count on your shortlist every Monday and compare. It is tedious, which is why most people don't do it. Ad Radar tracks engagement over time for every ad on the board and uses how fast it grows as one input to the winner score, next to days live and variants.
Signal 4: estimated spend
Meta doesn't publish spend for commercial ads. Every spend figure you see in a tool is a model. Good models are still useful because they let you rank: an ad estimated at $2M is almost certainly bigger than one estimated at $80K, even if neither number is exact. Read how to estimate a competitor's Meta ad spend for the approaches and their limits.
Why "most recent" misleads
The Ad Library can't sort by how long an ad has run, and on an active Page the fresh tests outnumber the survivors. A brand testing 20 new creatives a week will fill your screen with ads that are three days old, most of which will be off by next Monday.
Three habits fix this:
- Skip anything under 30 days unless it already has several variants. Recency is direction, not proof.
- Hunt for the oldest start dates first. Scan the whole Page and list every ad that started more than 90 days ago. That is where the controls live.
- Look for repetition in the middle. Ads between 30 and 90 days old that appear in several versions are the ones being scaled right now.
Winners are also not always cold-traffic ads. Some of the most-multiplied ads on our board are plain reminders aimed at people who already know the brand:
Did you forget something?
- Est. spend
- $2.4M
- Days live
- 83
- Format
- Image
- Variants
- 8
A four-word cart reminder with eight versions. If you only study prospecting hooks, you miss how much budget goes to the bottom of the funnel.
A scorecard you can use today
Score every ad on your shortlist from 0 to 2 on each signal. Anything at 6 or above goes into your swipe file.
| Signal | 0 | 1 | 2 |
|---|---|---|---|
| Days live | Under 30 | 30 to 89 | 90 or more |
| Variants | 1 | 2 | 3 or more, or repeated across IDs |
| Engagement growth | Flat for two weeks | Growing slowly | Clearly faster than the brand's other ads |
| Estimated spend | Under $50K or unknown | $50K to $250K | Over $250K |
Then do three things with the top scorers:
- Transcribe the first sentence. That is the hook you will test.
- Note the landing page type, because the ad and the page win together.
- Write one line on why it works, in your words, before you save it.
The reverse-engineering template takes it from there, and the 30-minute weekly research process keeps the shortlist fresh. If you want the scoring done for you, the Ad Radar board only shows ads that already cleared $50k in estimated spend, sorted by winner score.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.