An infographic that works as a poster usually fails on a phone, and the reason is not the colour or the typeface. It is that the format sets a reading budget: how many seconds and how many gestures the reader will spend before they scroll past. Design against the budget, not against your own patience.
The five formats, and what each one buys
| Format | Size | Reader effort | Best for |
|---|---|---|---|
| Square | 1080 × 1080 | One glance | A single figure, or a two-step comparison |
| Vertical | 1080 × 1350 (4:5) | One glance, more room | Three to five points, feed-native |
| Carousel | 1080 × 1350 per slide | One swipe per idea | A process, a listicle, a before/after |
| One-pager | A4 or US Letter | A deliberate read | A save, a print, an internal share |
| Loop | 3–6 seconds | Passive | A change over time, a build-up |
The pixel sizes are the platforms’ own; the “reader effort” column is a judgment about attention, not a measurement. Treat it as the design constraint and you will make fewer pages that only work on your monitor.
It also helps to hold the two halves together: the format sets how many gestures you get, and the layout sets what each gesture buys. A carousel gives roughly four swipes. A linear-progression spends one on each step; a comparison-matrix spends one on the table and leaves the other three with nothing to say.
Format is a layout input, not a crop
The most common wasted hour in this workflow is designing a tall infographic and then cropping it into a square. Cropping changes which element is the entry point, and the entry point is the entire design.
When we build an infographic with the infographic Skill, aspect ratio is declared before the layout is chosen, because the two constrain each other. A 21-layout catalogue looks like a menu of styles; it is actually a menu of information topologies — and each one carries differently in each aspect.
| Information shape | Layout | Carries in |
|---|---|---|
| A sequence or process | linear-progression |
Vertical, carousel, one-pager |
| A vs B | binary-comparison |
Square, vertical |
| Several factors at once | comparison-matrix |
One-pager, carousel |
| One idea with satellites | hub-spoke |
Square, vertical |
| Metrics and KPIs | dashboard |
Vertical, one-pager |
| A funnel or filtering | funnel |
Vertical |
| Density is the point | dense-modules |
One-pager, not a feed |
If the information is a sequence and the format is a square, the sequence gets squeezed into a corner and the reader never follows it. Change the layout or change the format; do not shrink the type.
How much text each format can hold
The budget is easier to hold in your head as words on the image, not seconds. These are the ceilings we plan for in our own drafts — working judgment from the plates we have made, not platform limits:
| Format | On-image words we plan for | What the rest becomes |
|---|---|---|
| Square | 10–20 | Caption, first comment, or nothing |
| Vertical | 25–40 | One supporting line under the chart |
| Carousel slide | 20–35 | One sentence of post copy per slide |
| One-pager | 150–300 | Footnotes and sources inside the page |
| Loop | 0–8 | A caption that names the change |
Two notes on reading that table. The safe margin — 8–10% of the canvas on the bottom and right — is a harder constraint than the word count, because platform controls do not move for your layout. And the number counts words on the image, including axis labels and legend entries; a legend you meant to delete is still three words of budget.
One brief, five formats
If you have one claim and five places to publish it, what changes between versions is narrower than it looks:
- The entry point changes. A square opens on the number; a carousel opens on the claim as text and earns the number on slide two; a one-pager opens on the number and the source.
- The module count changes. Four supporting blocks fit a one-pager. A vertical fits one. A square fits none.
- The type scale does not change. Keep the same 20% step between levels and the family looks like one system even when the layouts differ.
- The claim and the literal numbers do not change. If a figure shifts between formats, one of the versions is wrong.
What you are not doing is cropping. Reopening the layout with the same brief takes minutes in a generator; stretching a finished plate takes longer and loses the entry point.
Square is a billboard, not a page
A square gets roughly two seconds — long enough for one number and its label, or two bars against each other. It is not enough for a four-part framework, and reducing the type size to fit one is how infographics become unreadable at thumbnail scale.
Two rules make a square work:
- One element owns the page. In our composition rules the primary element takes 55–65% of the visual weight. In a square that reads as one big shape, one label, and a short line of support.
- Nothing below the safe margin. Platforms overlay their own controls on the bottom and right edges. Keep 8–10% of the canvas clear and your caption will not sit under a “like” button.
If the point needs three paragraphs of context, the square is the wrong format. Put the context in the post text and let the square carry the number.
Vertical buys a headline and a body
1080 × 1350 keeps the feed’s shape while adding about a quarter more height. That extra room is usually one more block: a headline, a chart, and a one-line conclusion. Vertical is the default for a social infographic because it matches how the feed already crops images.
The trap is filling the extra height with a second argument. A vertical plate with four separate claims has no primary element, and the eye has nowhere to land first. One claim, one chart, one takeaway.
Carousels are a sequence, not a poster cut up
Slicing one tall infographic into five slides produces five fragments that each need the previous slide to make sense. Instead, treat each slide as a complete idea with a reason to swipe:
- Slide one makes the claim. It is also the slide most likely to be seen alone, so it has to stand up without the rest.
- Slides two to four earn it. One idea each, in a consistent frame, so the reader knows where to look on every swipe.
- The last slide offers the action. A save, a link, a question.
Keep the same type scale and margins across the sequence. The carousel’s cohesion is what signals “this is one document”, and that is the signal that gets a save.
One-pagers are for saving
A one-pager is read deliberately, often after someone bookmarks it. That buys density: a full dataset, footnotes, sources. It is the one format where dense-modules and comparison-matrix belong.
Two things still apply. Put the thing worth saving at the top, because the screenshot people take is the first screen. And leave a quiet band at the bottom for sources — the evidence is what makes a dense page trustworthy rather than decorative.
Loops earn the third second
A three-second loop showing a line climbing, a before dissolving into an after, or a map filling in does one job a still cannot: it shows direction. Keep the motion to a single change. Two competing animations read as noise.
Loops are the weakest format for text. Anything you would label on a still should be short enough to read in one beat, or it belongs in the caption.
The thumbnail test
Before publishing, shrink the image to about 120 pixels wide and look at it. You should see one shape and one direction. If you see an even grey field of little tiles, the plate is flat and no amount of detail will rescue it in a feed.
This is the same test we apply inside the Skill: one entry point, a proportional ladder, one accent, and an exclusion zone around the ornament. It is easier to fail it in a square than in a poster, which is why the square is the format most worth designing last.
A quick chooser
Ask what you want the reader to do, then pick the row:
- Remember one number — square, one element, no second claim.
- Follow a process — carousel, one step per slide, consistent frame.
- Save something for later — one-pager, dense, sources at the bottom.
- Feel a trend — loop, one motion, no reading.
- Compare two things — vertical binary comparison, equal halves, difference highlighted once.
Everything else is a compromise between two of those, and the compromise is what you will pay for in the crop.
How this post was made
Written by Alex Zhang, founder of NeoSpark, from the layout catalogue and composition rules in the infographic Skill; the format sizes are the platforms’ published dimensions. The density and attention estimates are our working judgment, stated as such rather than as data. Related: what an AI infographic generator actually decides and the poster version of these rules. More pieces like this sit in the Formats archive.
