How carousel cards earn the swipe

The spec, exactly#

Two to ten cards, 1:1 square, 1080 by 1080 recommended, each card with its own destination URL and a headline truncated near 20 characters on Feed. Cards can mix image and video, and catalogue-fed carousels pull products automatically with prices and availability.

Meta own carousel specs list the per-placement rendering. Five cards is the practical storytelling maximum; later cards starve for swipes. The hub places carousel in the family.

Name card files for story order so future edits never scramble the sequence: numbers beat clever names at midnight. Keep a swipe map with each card's job, link and result; the next story starts from evidence. Retire tired openers on schedule.

Sequencing that sells#

Order cards as a story, not a shelf: payoff, proof, mechanism, offer, action. Each card must earn the next swipe with an open loop or a new fact; parallel cards with no order read as a menu and convert like one.

Write card headlines as a sentence across the swipe: read together they pitch, read alone each still makes sense. Our sequencing examples live in the deck lab, and Meta catalogue thinking applies where feeds assemble cards: catalogue ads guide.

Per-card links#

Every card gets its own deep link: product pages, steps, proof, pricing. Homepage dumps are the format's original sin, and quarterly link audits catch the rot as ranges change.

Map links to funnel stages, not just pages: awareness cards to guides, consideration cards to comparisons, action cards to checkout. Stage-matched links lift swipe-through because every tap continues the thought. Stage-matched routing aligns cards with temperature.

Tag each card distinctly so reporting shows which panel sells and which merely decorates. Attribution still applies per card.

UTM every card distinctly so analytics attributes revenue to the panel, not just the ad: card-level truth beats ad-level averages when sequencing the next story. Keep slugs human-readable; midnight debugging thanks you. Mismatch guide reconciles platform and analytics counts.

Catalogue-fed carousels#

Connect the product feed and Meta assembles cards from product sets by margin, stock or season, updating prices and availability automatically. Curation moves upstream: the set definition becomes the creative decision.

Keep manual carousels for hero stories and launches where narrative control beats automation. Product sets show the slicing.

Name sets for the campaign, not the catalogue: launch edits, margin heroes and seasonal picks read clearly at 2am. One feed can serve many stories once slicing is deliberate. Multi-country feeds multiply the discipline across languages and prices.

Testing carousels honestly#

Test carousel against a single-image control with identical budget and audience, and read swipe-through plus CPA rather than CTR alone. Test sequences against each other by changing card order, never five variables at once.

Give tests a full learning window before reading: early swipe data flatters novelty and punishes familiarity. Two weeks minimum, same budgets, same audiences, then decide. Learning windows protect tests from premature verdicts.

Retire tail cards that never get seen: a three-card carousel that converts beats a ten-card museum. Our A/B guide sets the read rules.

Test card order as its own variable once the control wins: promise-first against proof-first often moves swipe-through double digits. Document the winning arc per audience, because new buyers and warm browsers read in different orders. Audience splits shape the sequence.

When carousel loses#

Carousel loses where one promise suffices: single products, simple offers, urgent actions. It also loses on surfaces that show one card poorly, so check Stories and Reels rendering before judging.

Refresh sequences quarterly as ranges, prices and proof change. Ask our team for a carousel teardown when swipes stall.

Frequently asked questions#