One placement, several cards
A carousel ad is a single paid social placement that holds a row of cards the viewer swipes sideways through, each card carrying its own image or video and usually its own headline and destination link. The practical consequence is an ordering problem rather than a creative one: card one has to earn the swipe, and every card after it has to earn the next swipe, so sequence decides how much of the ad anybody ever sees.
Card limits, supported file types and aspect handling differ between ad platforms and change over time, so confirm current specifications in the platform's own documentation before you commission assets. What stays stable is the framing discipline. Every card should share one crop, because mismatched shapes make the swipe feel broken even when each individual image is good.
The nearest confusable thing is a single UGC video cut into chapters. A video decides its own pacing; a carousel hands pacing to the viewer, who can stop at the first card or read all five. That difference is why the unit can act as a small catalogue, a numbered argument, or a two-card comparison, and why a video script rarely converts straight into cards.
Filling the cards with creator assets
Carousels are a good home for content you already own. Five creators each holding the product becomes a social proof unit. One creator's routine becomes a sequence. A set of stills shot on plain surfaces covers colourways and sizes without booking anything new, which is usually the cheapest version you will ever build.
Worked example. Suppose you hold twelve creator stills and four short clips from last quarter. Card one is the strongest single image, card two answers the objection your support inbox repeats, card three shows the range, card four puts the product in a hand, card five carries the offer. That is a full unit assembled from existing UGC content and an afternoon of cropping.
The common mistake is treating the cards as a gallery. Five unrelated assets stacked together is not a carousel, it is a folder. Each card needs a reason to follow the one before it, and if you cannot name that reason out loud, cut the card.
Reading swipe and click numbers honestly
These units mislead if you only read the top line. A strong overall click number can come almost entirely from the first card while cards three to five do nothing at all, so check how far people actually swipe before you decide anything is working. Cut the dead cards instead of adding more.
Compare the unit against your video creative on the same objective, not against other carousels. Plenty of accounts find the format earns its place lower in the funnel, where range and detail matter, and loses to a single vertical video at the top. That is a test for your own account rather than a rule, and the answer moves by category.
Keep the test clean. Hold the card order and swap only the first card, because it is the one most viewers ever see and the cheapest variable to change. Rotating all five at once tells you a carousel changed without telling you which card changed the result.
What to put in the brief
Write the card jobs before you write the shot list. Name what each position has to do, then ask for the assets that fit: how many stills, which crop, what has to be readable, and where the negative space sits so a headline can go there later without covering the product.
Ask for landscape and vertical framings of the key shots, plus unedited originals at full resolution with no filter or logo baked in. The same set then feeds an email header and a product page gallery. Settle usage rights across all placements in one agreement, because a still licensed for organic only is useless the day you want to spend behind it.
How it's used
Swipe depth on the new carousel falls off a cliff after card two, so I am cutting it to three cards and moving the objection answer into second place.
Shoot request: five stills, identical crop, product centred, one per colourway, and leave the top third empty so we can drop a headline in later.