Carousel Holds the Longest Median Run on Meta at 56 Days, Against UGC's 44 and Static Image's 30. When Did You Last Build One?
Scope: a longevity-ranked sample of Meta ads from every industry and country the eCommerceBenchmark Ad Library tracks, 35,246 ads, read at the start of November 2026, with each ad counted from the day it was first seen. No split by country or industry.
Open the account and look at what has been on air the longest.
Across every industry and country the eCommerceBenchmark Ad Library tracks, the format that holds its place longest is the carousel. Its median run is 56 days. UGC comes next at 44 days. Studio video's median is 35 days. Static image sits at 30 [eCommerceBenchmark Ad Library, 2026]. A long run only means a brand kept paying to keep an ad on air; it says nothing about what the ad sold.
The order is read from 35,246 Meta ads at the start of November 2026, each ad counted from the day it was first seen. The sample is ranked toward long runs, so the day counts sit high by design. Treat them as an order, not as lifespans.
Here is that order, the second cut that agrees with it, and the one column that changes how you read the whole table.
- Durability is not fashion. Carousel's median run is 56 days. UGC's is 44. If next month's brief treats the carousel as a legacy unit and creator video as the thing that lasts, the order says otherwise.
- A short median can mean a young sample, not a short life. Founder-to-camera's median is 32 days. Yet 70.3% of its sampled ads were on air when the sample was read.
- Two cuts, one order. A second read of the same sample, split by delivery, puts carousel ahead of video and static at 46 days. So the ranking is not one slice's quirk.
Carousel holds the longest median run at 56 days
Carousel holds the longest median run of any format we can rank, 56 days. It rests on 9,868 sampled ads [eCommerceBenchmark Ad Library, 2026], the biggest cohort in the list.
A creative format is defined as the content type an ad is delivered as on Meta, such as carousel, static image, studio video, UGC or founder-to-camera footage. UGC comes second at 44 days. Animation is third at 39. Studio video, the polished in-house build most teams brief by default, sits fifth at 35 days. Static image is near the bottom at 30.
| Format | Median days on air | Share on air | Ads sampled |
|---|---|---|---|
| Carousel | 56 | 39.9% | 9,868 |
| UGC | 44 | 32% | 6,224 |
| Animation | 39 | 35.6% | 694 |
| Studio video | 35 | 42.1% | 6,163 |
| Lifestyle aspirational | 34 | 32.6% | 4,850 |
| Founder-to-camera | 32 | 70.3% | 316 |
| Static image | 30 | 57.8% | 780 |
| Product image + text | 29 | 39.1% | 4,241 |
| Influencer endorsement | 29 | 60.9% | 1,475 |
One honest warning. One format in this sample carries a higher day count than carousel, and it is not in the table: meme and trend rests on 20 sampled ads. That is too few to rank, so it is named as thin and left out, along with the three other formats below the floor.
So the order holds when you cut the sample by what the ad is made of. Does it hold when you cut it by how the ad is delivered?
The delivery cut puts carousel first too, at 46 days against static's 34
A second read of the same Meta sample, split only by how the ad is delivered, puts carousel first at 46 days. Video follows at 39 days. Static sits at 34 [eCommerceBenchmark Ad Library, 2026].
The cohorts here are sized differently from the ones above. Carousel holds 10,928 ads. Video holds 16,337. Static holds 7,981. Two reads of one sample, taken at different moments, so each day count belongs beside its own sample size and never added to the list above.
That matters because a different way of slicing the same ads could easily have flipped the order. It did not. Carousel came first both times (time on air, not sales).
One column in the first table has been doing quiet work all along. Look at the share on air.
A short median can mean a young sample: founder-to-camera at 32 days
Founder-to-camera's median run is 32 days, near the bottom of the ranked list. Its share on air is the highest of any ranked format, 70.3% [eCommerceBenchmark Ad Library, 2026].
Share on air is defined as the portion of a format's sampled ads that were live on the day the sample was read. That share sits on 316 sampled ads, a cohort that grew by 284 ads since the previous read. A format the library has only just started filing cannot show long runs yet, because most of its ads have not finished running.
Static image shows the same shape. Its median is 30 days, with 57.8% of its sampled ads on air at the read. Its sample grew by 517 ads to 780. Influencer endorsement's median is 29 days, and its share on air is 60.9%.
Read the two columns together. A low median beside a high share on air describes a format the market has only just begun filing, not one that falls over.
Here is what that opening sounds like. drsquatch.com runs a founder-to-camera ad that begins: “Hi, this is Jack, founder and CEO of Dr. Squatch. If you're seeing this video, you probably just placed your first order online with Dr. Squatch.” That ad comes from the library's longest-running list, a separate slice of Meta ads, where it has been on air 1,023 days. It is not part of the 32-day median above.
Which leaves one habit worth naming.
Three format tests to run before your next rotation
- A carousel against your longest-serving single asset. Rebuild one control as a carousel and run it beside the original: static image's median run is 30 days, against carousel's 56.
- A founder-to-camera cut of your best creator script. Same script, founder on camera, against the creator version: founder-to-camera's sample holds 316 ads, and most of them were on air at the read.
- One studio video rebuilt as its own frames. Pull three stills from a studio video into a carousel and run both: studio video's median is 35 days.
The eCommerceBenchmark Ad Library sorts your ads into the same hooks, angles, concepts and offers used in this report and shows your mix next to the market, so you can see which shapes you have never shipped.
Audit my creatives, freePick one and write it down before you launch. Then look at the habit these tests are built against.
The mistake: rotating a format out because its median looks short
The practice is simple: read one column of a durability table and plan the month off it. Static image shows a 30-day median, so the single asset gets cut from the slate, and the format with the longest median never makes the brief at all. The second column says something else about that same row: 57.8% of static image's sampled ads were on air when the sample was read [eCommerceBenchmark Ad Library, 2026].
When a whole field does it at once, every account churns the same short-lived single assets month after month while the format holding a 56-day median goes unmade. A shopper scrolling a feed meets the same monthly reset in every one of them, and nothing in any account has been tested against the format that holds its place longest.
Your test card
Here is the first test from the box, written the way a buyer would write it before launch. Fill in the blanks with your own numbers, and write them down first.
| Hypothesis | A carousel rebuild of my longest-serving single asset will hold its place in the account for at least as long as the original, without taking budget from it. |
| Change only this | The build: the same offer and the same opening words, delivered as a carousel instead of one still. |
| Keep the same | Audience, placement, budget, landing page, offer wording and the headline. |
| Funnel stage | Say first whether this ad is for prospecting (cold) or retargeting (warm). A deal in the first line is often right for warm audiences and wrong for cold ones, so test it where the ad will really run. |
| Setup | One control and one variant, same audience, same placement, even budget split. Do not edit either ad while the test runs. |
| How long and how much | Until each version has about 50 purchases, or two full weeks, whichever comes later. This is general testing practice (Meta's own learning-phase guideline is about 50 results per ad set per week), not a library finding. |
| Success metric | Cost per purchase, not clicks. |
| Stop rule | Pause a variant that has spent about 3 times your target cost per purchase with no purchase. If it is close, keep it running. |
| What counts as a win | The variant beats the control on the metric by an amount you would act on, and still does in a second week. |
Key takeaways
- Carousel holds the longest median run of the formats we can rank, 56 days [eCommerceBenchmark Ad Library, 2026], on the biggest sampled cohort in the list.
- UGC comes second at 44 days, so creator-style video is durable but not the most durable format here.
- Studio video's median run is 35 days, below both UGC and carousel.
- Static image's median is 30 days, near the bottom, on a sample that grew by 517 ads to 780.
- 70.3% of founder-to-camera's sampled ads were on air when the sample was read, the highest share of any ranked format.
- The delivery cut agrees on the order, putting carousel first at 46 days, ahead of video and static.
- 35,246 Meta ads were read for this ranking, a subset of the decomposed cohort, so the counts here are smaller than a whole-corpus figure.
Questions
Run the audit on your own account. Pull every ad you had on air last month and sort it into the same boxes this ranking uses: carousel, UGC, studio video, founder-to-camera, static image. Count what is in each box. Carousel holds a 56-day median here and static image 30, so if the carousel box is empty while the single-still box is full, you have been planning off the shorter half of the table. Build one carousel before your next budget cycle and judge it on your own numbers.