Facebook carousel ads vs single image
Facebook carousel ads vs single image: where the 10x click claim comes from, when extra cards earn their place, and how to test both formats fairly.
Facebook carousel ads vs single image: where the 10x click claim comes from, when extra cards earn their place, and how to test both formats fairly.
The designer has room for one more creative this week, and the brief says "carousel". Someone read that carousels get ten times the clicks of a single image. Someone else read that Facebook says they cut cost per conversion by 30 to 50%.
I went looking. The 30 to 50% figure is repeated everywhere, credited to Facebook, and I could not find the original. The "ten times" figure at least has a trail. It ends in 2015.
Short answer: Facebook carousel ads beat a single image only when the extra cards do a job one frame cannot: walk the viewer through a sequence (problem, mechanism, proof, offer) or offer a real choice across a product range. Otherwise a carousel is a single image with homework attached, because its first card has to stop the scroll alone.
The takeaways
The claim that carousel ads get ten times the clicks of single images comes from a Digiday article from October 2015 citing the ad-tech firm Kinetic Social: roughly 1% click-through on carousel ads versus 0.1% on non-carousel sponsored posts. That comparison is ten years old, from a feed with far less video, and its baseline was ordinary sponsored posts.
There's a quieter problem. Every carousel card carries its own link, so every card is another chance to click. More clicks per impression is not the same as more buyers per euro. If many of those clicks are curious taps that bounce, the carousel's CTR looks heroic while its cost per purchase looks ordinary.
The first card of a carousel has to stop the scroll by itself, because the feed shows card one in full and the rest only to people who swipe. Meta publishes no swipe rate, so nobody outside Meta knows how many never swipe. It isn't a small share.
A first card that is a logo or a "swipe to see more" arrow asks for effort before giving a reason. Write card one as if it were a single image ad: one claim, legible at a glance, aimed at the viewer's problem. The other cards are a bonus for the people card one already stopped.
If you can't write a card one that works as a standalone static, a carousel won't rescue the idea. Everything about the one-glance job of a static applies to card one unchanged.
Extra carousel cards earn their place when they do one of two jobs a single frame cannot.
The first job is a sequence. Problem on card one, mechanism on card two, proof on card three, offer on the last. This suits a product that needs explaining before the price makes sense, and it is the closest a static format gets to a short video.
The second job is a choice. Colours, sizes, bundles or price points, where the viewer is better served by picking than by being told. Here the order matters less.
Anything else is decoration. Five angles of the same product, or four cards repeating one claim, is a single image with more production cost and the attention split five ways. If the cards form neither a sequence nor a choice, run the strongest one alone.
Per-card CTR in a carousel misleads because position collects most of the clicks. Card one is seen by everyone who sees the ad; card four only by people who swiped three times. So card one nearly always shows the most clicks, and a late card with a handful can still be the one that closes, because the few people who reach it are the most interested.
Cutting the "weak" late cards on click counts can quietly remove the proof or the offer, the part that did the persuading. The carousel gets worse, and the click report never shows why.
Read the ad's overall cost per result first. To judge a single card, compare it against how many people could have seen it, and treat small numbers as small. A late card with 12 clicks proves very little either way.
Meta's option to automatically show the best-performing cards first should be off for a sequence carousel and on for a product-range carousel. It lets delivery move whichever card it predicts will get the click to the front. For a range that helps: if the blue one sells, lead with the blue one.
For a sequence it breaks the argument. A viewer who opens on your proof card has no idea what is being proved, and an offer card shown first is just a price tag.
The reorder logic also chases clicks, the same position-biased signal described above, so it can't judge which card persuades. Check the option on every carousel before launch; Meta moves its settings often.
A fair carousel vs single image test keeps the angle constant and changes only the format. The single image carries the same claim as card one; the carousel adds the sequence or the choice. Give each its own ad set with the same budget and audience, because two formats inside one ad set don't get an even split. Delivery picks an early favourite and starves the other.
Judge both on the result the campaign optimises for, cost per purchase or cost per lead. CTR is exactly where a carousel gets free credit for taps on extra cards. And make the variation big enough to pay for the test: a carousel that only adds three near-identical cards tests nothing. With no public swipe-rate benchmark, your own test is the benchmark.
Adscalr doesn't generate carousels. Its 3 creative types are statics, UGC concepts and briefs, and motion storyboards, each composed for 4 native formats instead of cropped from one master.
It helps with the card that decides everything. Concepts come from competitor angles, real customer quotes and your own performance data, and a copywriting critic checks each against 6 direct-response principles, quoting the line that fails. If writing that first card is where your week goes, that's the ad-creation side of the product.
This is the thinking behind Adscalr.
See the product →