Most App Store screenshot critique gets stuck on surface-level aesthetics — "this one has nice gradients", "this one's clean", "this looks premium". Those statements don't tell you what to change in your own carousel. Useful critique is structural: which decisions are making the screenshot convert, separately from which decisions are just taste.
We're going to walk through a generic-but-realistic example. Imagine a panel from a finance app — top-100 in the Finance category, App Store US — showing a portfolio dashboard. The panel reads as one continuous design with the previous and next panel in the carousel.
The headline sits in the top quarter of the panel, anchored to the left edge with a 60-pixel gutter. The caption itself reads: "Your money, all in one view." Six words, verb-first but in a slightly indirect way (the implied verb is "see"), specific noun ("money") rather than abstract ("finances").
Why top quarter: that's where the eye lands first in the carousel-thumbnail render. We see this consistently — eye-tracking studies from 2024-2026 ASO research (Phiture, AppTweak, Adapty roughly converge here) place the strongest attention at the upper-left third of the panel. Captions anchored there read; captions positioned mid-panel below the device often get skimmed over completely.
The iPhone is tilted ~12° on the Y axis (around the vertical axis, so the right edge tilts away from the viewer). Same tilt direction repeats across all 5 panels in the carousel — not alternating left/right.
2023 advice often said alternate. By 2026 the top-100 mostly doesn't. The reason: alternating reads as restless at thumbnail size. Same-direction tilt makes the row feel like one motion, like the carousel is rotating through views of the same product rather than fragments of different views.
Tilt angle: 10-15° is the sweet spot. Below 8° the tilt is too subtle to read at thumbnail size; above 18° the device looks like it's falling over.
Cool palette — deep blue gradient from #0a1530 top-left to #1e3f7e
bottom-right. Finance apps lean cool by category convention (banking signals trust, blue
signals trust, the convergence is not subtle). Health apps lean cooler still (teals,
soft greens). Entertainment leans warm (oranges, deep magenta). Productivity is mixed.
The color temperature isn't doing conversion work on its own — it's doing category-recognition work. A user scrolling the App Store thumbnails recognizes "this is a finance app" from the palette before they read the caption. That recognition speeds up the "is this for me?" decision.
The mistake worth avoiding: picking an off-category palette to "stand out". You stand out, but users misread the category at thumbnail size and bounce.
The gradient flows continuously across all five panels — the dark top-left of panel 1 fades through panel 3's mid-blue into panel 5's lighter accent. Each panel's device is positioned at roughly the same horizontal anchor (centered, slightly low). Caption position is identical across all panels.
This continuity is what separates "designed as a unit" carousels from "five individual screenshots". The user reads the carousel as one story; even though each panel features a different app feature, the visual scaffolding doesn't reset between them.
Things that vary widely across high-performing top-100 carousels:
Honest takeaway: when you study top-100 screenshots in your category, copy the four structural decisions and ignore the surface aesthetics. Specifically:
The first list is doing the conversion work. The second list is taste. Spending design time on the first list compounds; spending design time on the second list usually doesn't move the conversion number.
This sounds prescriptive, and it is. The point isn't that everyone should ship the same carousel; it's that the structural decisions are where the conversion lives and the taste decisions are where the brand lives. Keep them separate.