Which Meta ads aspect ratio to produce: why 4:5 beats 1:1 in the mobile feed, where 9:16 is non-negotiable, and a budget-based rule for how many ratios each concept earns.
The concept is approved, the copy is locked, and the designer asks one last question: which sizes do you need? For years my default answer was square. 1080 by 1080 ran everywhere without complaint, so it became the master format for every batch I briefed, including the months I was spending around €150k per month on paid social.
That answer was wrong, and it was wrong on every mobile impression. A square asset in the mobile feed hands back a strip of screen that a 4:5 version would have filled. So is 1:1 dead? No. The better question is which ratios deserve production budget for a given concept, and which placements you can safely leave to Meta's own adaptation. Here is how I decide now.
The takeaways
No. Square still has placements where it is the correct format: Facebook right column, search results, Marketplace and the Messenger inbox all run square or near-square assets according to Meta's placement specs. If your account spends meaningfully there, a clean 1:1 belongs in the set.
What 1:1 has lost is its job as the default master for the feed. On a phone, feed width is fixed, so height is the only dimension you can win. A 1:1 asset at 1080 px wide shows 1080 px of height. A 4:5 asset at the same width shows 1350 px. That is 25% more screen for the same auction price, no targeting change, no bid change. Square made sense when desktop feed and mobile feed mattered equally. Today, with most paid social delivery on mobile, briefing 1:1 as the lead format means paying full price for four fifths of the available canvas.
Because the feed scrolls vertically and Meta's specs cap feed assets at a 4:5 ratio. Anything taller gets masked down to 4:5, so 4:5 is the tallest format the feed will fully display. The tallest permitted asset claims the most vertical space, keeps the user's thumb busy longer, and pushes the next piece of content (often a competitor's ad) further off screen.
There is a second, quieter benefit: composition room. The extra 270 px of height on a 1080-wide canvas is space for a hook line above the visual or proof below it, without crowding. In my own batches, the 4:5 recompositions were rarely just stretched squares. The good ones used the extra fifth for the one element the square version had to drop. If you check creatives with an attention model or just squint at them, that breathing room is usually where the hook earns its first second.
Stories and Reels, both on Facebook and Instagram. These placements are full-screen vertical experiences, and Meta's specs call for 9:16 there. Feed-shaped assets technically still deliver: Meta fills the empty space with generated or blurred backgrounds. It looks exactly like what it is, a feed ad wearing the wrong clothes, and it signals "repurposed" to a viewer who lives in that format all day.
9:16 also comes with its own composition rules. The top and bottom of the frame carry interface overlays in Stories and Reels (profile name, CTA, engagement icons), so Meta's guidelines advise keeping text and logos out of those zones. A 9:16 asset is therefore a recomposition job with its own safe area, and a designer needs to know that before export, because no automatic adaptation will move your caption out of the way of the CTA button.
Asset customization lets you assign a different image, video or crop to each placement group inside a single ad: one asset for feeds, another for Stories and Reels, another for right column. Engagement still aggregates on the one ad, so you keep social proof while serving the correct shape everywhere. This is the mechanism that makes a two-ratio or four-ratio production strategy practical without multiplying your ad count.
The alternative is letting Meta adapt a single asset on its own, through automatic cropping and its various enhancement options. Sometimes that is fine. Often it is how captions get cut mid-word, how a centered product drifts off-frame, and how the logo you placed in a corner disappears entirely. My rule: Meta may choose which asset to serve, never how to recompose one. Upload the shapes yourself and switch the automatic adjustments off, then audit the previews per placement before launch.
Tie it to the concept's budget tier, because every added ratio costs designer hours and those compound across a testing program.
For new, unproven concepts, two masters: 4:5 and 9:16. The 4:5 covers Facebook and Instagram feeds, which is where most delivery lands. Compose it center-weighted, and a 1:1 crop for the minor square placements survives without surgery. The 9:16 covers Stories and Reels natively. Two exports per concept is a cost a ten-concept test batch can carry.
For proven winners, the full set: 1:1 and 16:9 built natively, each recomposed for its placement rather than cropped from the 4:5. The 16:9 matters less than it used to, but in-stream video and some desktop and Audience Network slots still call for it, and a winner running everywhere should look deliberate everywhere. A concept spending five figures a month makes per-placement composition cheap insurance. This second pass also folds neatly into the refresh cycle, since a fatigued winner needs new executions anyway, and the refresh brief is the natural moment to fill in missing ratios.
The mistake to avoid sits at both extremes: producing four ratios for a concept that dies after €300 of spend, or scaling a winner for months on an auto-cropped square.
One thing running multi-format batches taught me: a 9:16 Reels asset and a 4:5 feed asset from the same concept do not score alike in their first days, and comparing their raw early numbers misleads. That observation is built into how Adscalr reads tests. Each format carries its own performance prior, so a new ad's early results get pulled toward what its format normally does before any verdict, and creatives are generated in all four native ratios (1:1, 4:5, 9:16, 16:9), each composed natively for its placement. The thinking behind that lives on the ad creation page.
You can run the same discipline manually: brief two masters per concept, reserve the full set for winners, and never let an auto-crop decide what your hook looks like. The ratios are arithmetic. The discipline is the part most accounts skip.
This is the thinking behind Adscalr.
See the product →