The short answer
- Two or three cards for a narrative carousel. Longer sets are catalogues, where the swipe is browsing.
- Card one must be incomplete. Every pattern that keeps running puts a question or a gap on the first card.
- Same ratio on every card — 1080 × 1080 — or the set reads as a mistake.
- Check the run dates. An ad live for 200 days is a pattern; one live for two is a test.
Six patterns worth stealing
| Pattern | What it does | Cards |
|---|---|---|
| Product tour | One feature per card, in the order a new user meets them | 3–5 |
| Before / after | Card one the mess, card two the fix — the swipe is the payoff | 2 |
| Listicle | “Five things we learned”, one per card, editorial rather than salesy | 3–5 |
| Proof wall | A named customer quote per card, for claims that strain belief | 2–4 |
| Catalogue | One product per card with price; usually dynamic, not hand-built | 5+ |
| Single image, cut | One wide image split across cards so the swipe reveals it | 2–3 |
Live examples
These are running now. Both use the two-card reveal, which is the shape that dominates in practice:
Browse every carousel in the Sprites ads library → Each one appears card by card with the advertiser named and its run dates attached.
Writing card one
- Ask something the reader cannot answer. “Facebook ads are a competitive game” works because the next card is the escape.
- Do not finish the thought. A complete card one is a static ad with extra steps.
- Keep the product off card one in narrative patterns — it belongs where the payoff is.
Next
- Facebook ad sizes — the ratios to export, and the safe zones.
- Analyse competitor ads — turning patterns you spot into your own.
- Meta ads library — live Meta creatives by brand and industry, with run dates.
- Sprites — the AI marketing agent that turns a pattern into a launched carousel test, with every change approved first.