Hierarchy: deciding what gets read first, second and third
Sort everything on the piece into three levels, then make the jump between levels visible at the reading distance using size, weight, position and space.
- Start here: the essentials
- What hierarchy is, and what it is for
- Three levels, and what goes in each
- Making the jump visible
- The four tools, in the order to reach for them
- The walk-back, which is the only real test
- Headings that say something, and the layer-cake scan
- When three levels will not hold it
- See it: a flyer, annotated
- Practice: sort, set, walk back
- Your piece
- Check your work
- Knowledge check and scenario
Start here Essentials
This is the decision that does the most and costs the least. It needs no new font, no new color and no new software.
- Put every piece of text into one of three levels. First read, second read, third read (three levels).
- Set everything in a level the same way as everything else in that level. Same size, same weight, same alignment (making the jump visible).
- Make the jump between levels obvious at arm's length. A level that is slightly larger than the one below reads as a mistake rather than a level (the four tools).
- Walk back to the distance you paced and look again. Not at the screen, at a printout, from the real distance (the walk-back).
Minimum time: about 30 minutes, plus one test print. Done when: standing at the reading distance, you can say what the piece is about before you can read anything else on it.
1. What hierarchy is, and what it is for Understand
Hierarchy is the order a reader takes things in. Not the order you wrote them, and not the order they sit on the page from top to bottom, but the order the eye actually goes.
A piece with no hierarchy is not ugly. It is slow. Every line is the same weight, so the reader has to read all of them to find the one they wanted, and at a bulletin board they are not going to. A piece with hierarchy answers the first question before the reader has decided to ask it.
The question a reader asks first is usually "is this for me". If the piece answers that at a glance, they will look for the second thing, which is usually "when" or "where". If it does not, they look at the next flyer.
So the three levels are not three sizes. They are three questions:
- What is this? Read from the paced distance, before the reader has committed.
- When and where, or whatever the second question is for your piece. Read once they have stopped.
- Everything they need once they have decided. Cost, contact, details, the small print.
Preference Three is a working number rather than a published one. It is small enough that the jumps between levels can be large, and large enough that most ordinary pieces fit. Some pieces have two. A sign read from a car usually has one.
The design curricula this manual compared converge on the same middle. The CalArts fundamentals program has a module called Composition and Visual Hierarchy, covering cropping, hierarchy, direction, scale, weight and texture; the LinkedIn Learning path gives layout and composition two courses. Where those differ from this chapter is that they teach it as a compositional craft with many outcomes, and this chapter teaches it as one decision with a test attached. Preference
2. Three levels, and what goes in each
Take the fact list from The words come first and put a 1, a 2 or a 3 beside every line. Decide by asking what the reader needs at that moment, not by how important the line feels to you.
| Piece | Level 1 (from the paced distance) | Level 2 (once they stop) | Level 3 (once they have decided) |
|---|---|---|---|
| Market flyer in a shop window | ARBOR STREET MARKET | Second Saturdays, 8 AM to 1 PM | Fourteen vendors, free, the lot address, the contact |
| Corner sign for a driveway sale | SALE and an arrow | (none) | (none) |
| Fence sign at the house | SALE | Saturday 8 to 2 | 1409 Vance |
| Retirement supper invitation | Teodoro Fitz | Saturday, March 14, 6:00 PM | Address, parking, cost, RSVP |
| Hallway sign to a room | OAK ROOM, with an arrow | Story Hour, Thursdays 10:30 | (none) |
| Slide in a council chamber | The one idea of the slide | The labels on the picture | (none; it goes in the handout) |
Three things are worth noticing in that table.
- The date is level two more often than people set it. It is rarely the first thing read, because few people scan a board looking for a date. But it is the second thing, and it is routinely left at the size of the contact details, which is level three.
- Some pieces have no level three, and trying to give them one is how a sign ends up unreadable.
- Level one is short. Two or three words, often. If your level one is a sentence, it is a level two wearing a hat.
A reader will over-read "make the headline big" into a headline that eats the page and a date left at 10 point. The levels are relative to each other, not absolute. The test is not whether level one is dramatic; it is whether level two still reads at the distance a reader will be standing when they look for it. A piece whose headline is enormous and whose date is invisible has a worse hierarchy than one with a modest headline and a readable date, because the reader gets stopped and then stranded.
3. Making the jump visible
Once things are sorted, two rules do most of the work.
Everything in a level is set the same way. Same size, same weight, same alignment, same color. If two lines are at the same level and look different, a reader reads that difference as meaning, and spends a moment working out what it means. That moment is the cost.
The jump between levels is large. A level that is a little bigger than the one below reads as an accident. The rule of thumb that works is that each jump should be obvious from across a room without reading the words, and the easiest way to test it is to blur your eyes or to look at a thumbnail.
Here is the same line at a set of sizes, so you can see what a jump looks like against what a nudge looks like. The captions state each size; the lines themselves are decoration.
- Arbor Street Market40 px
- Arbor Street Market28 px
- Arbor Street Market20 px
- Arbor Street Market17 px
- Arbor Street Market14 px
The step from 40 to 28 reads as two levels. The step from 20 to 17 reads as the same level, set slightly wrong. Test it These rows show the sizes relative to each other and they scale with the text size you have set for this page, so what they demonstrate is the proportion, not any printed size.
A workable starting point is that each level is roughly half again to twice the size of the one below. That is arithmetic, not a published figure, and the real check is the walk-back.
4. The four tools, in the order to reach for them
Four things make a level look different from another level. Reach for them in this order, because the early ones are free and the later ones cost you something.
- Size. The strongest and the cheapest. It costs space, which is the only thing it costs.
- Space. Nearly as strong, and often invisible to people who have not noticed it. A line with a lot of space around it reads as more important than a line jammed between two others at the same size. Space and alignment is the chapter.
- Weight. Bold, or a heavier weight of the same family. Cheap, and it goes stale fast: a piece where half the lines are bold has no bold. Butterick's rule is to use bold or italic as little as possible, and not together. Preference That is one practitioner's recommendation, with no study behind it, and it is a reasonable default.
- Position. The top of a piece and the left edge of a block are read first in English, so putting something there does some of the work. Position is a weaker signal than the first three and a cheaper one than color.
What is not on that list, and what people reach for first:
- A second typeface. It makes two lines look different without making one look more important, and it gives you another thing to keep consistent. One family with two or three weights produces a clearer hierarchy with less work. Type you can read goes through what the sources say and do not say about this.
- Color. Color can carry emphasis, but it cannot carry it alone: WCAG's use-of-color criterion asks that information not be conveyed by color alone, and a reader printing your flyer in black and white loses it entirely. Contrast and color has the measurable part.
- Boxes and rules. A box says "these things belong together", which is a job that grouping and space already do. A page full of boxes has told you nothing about order. Space and alignment asks you to take them out and see what falls over.
- All capitals. Capitals look emphatic and read slower, which is a poor trade on anything longer than a line or two. The federal plain language guidance says all caps is harder to read and is treated as shouting; Butterick says caps are fine for less than one line, and asks for extra letterspacing whenever they are used. Those two agree, from different directions. Convention
5. The walk-back, which is the only real test
Print it. Take it to the place. Stand where a reader stands. Look.
Everything before this is a prediction. Three things go wrong between the screen and the wall, and none of them is visible at a desk:
- The screen lied about size. A document at 70 percent zoom on a large monitor is nothing like an 8.5 by 11 inch sheet six feet away.
- The competition was not in the mock-up. Forty other flyers, a window reflection, a poster behind yours.
- The light was wrong. Your screen is backlit and the wall is not.
Test it The question to ask yourself, out loud, standing there: can I say what this is before I can read anything else on it? If you have to look for the answer, level one is not working, and the fix is size or space rather than a new idea.
Then take one step closer, to the distance somebody stands when they have stopped, and ask the second question: when and where? If level two has to be hunted for, it is set too close to level three.
Write what you could and could not read in the Distance test log on Before it goes out. The log matters more than it looks: the first column is the finding and the last is the fix, and keeping them apart stops you from changing things you have not tested.
Headings that say something, and the layer-cake scan Go deeper
The Nielsen Norman Group's 2019 eye-tracking article names several patterns people use on text they have not committed to reading, and ranks them, in its own words, worst to best for effectiveness: F-pattern, spotted, layer-cake, commitment. The layer-cake pattern, where a reader scans the headings and drops into the one that matters, it calls by far the most effective way to scan short of reading nearly every word.
What produces a layer-cake scan is headings that say something. Its 2017 article on the F-shaped pattern is explicit that the F-shape is what happens when there are no strong cues: when text has little or no formatting, when the reader is being efficient, and when they are not committed enough to read every word. It calls headings, bullets, bold text and good formatting the antidotes.
So the practical claim this manual draws is narrow and worth stating precisely. Test it A heading that names its content gives a skimming reader somewhere to land. A heading that says "General" or "Information" does not, which is also what the federal guidance on headings says when it calls topic headings sometimes so vague that they are not helpful.
Two cautions. The heatmaps in those articles come from small groups, and the articles describe them as illustrations rather than statistical tests. And the 2019 article says patterns are highly contingent on task, assumptions, layout and content type. Neither article is about flyers, and neither measures whether a better scan produces an action.
- Text Scanning Patterns: Eyetracking Evidence: the 2019 article that names and ranks the patterns.
- Headings: the federal guidance on question, statement and topic headings.
When three levels will not hold it Go deeper
Sometimes the sort produces four or five levels, and the usual cause is one of three things.
The piece is doing two jobs. A flyer that both invites guests and recruits volunteers has two level ones. The honest fixes are to split it into two pieces, or to pick the real action and demote the other to a line at the bottom. What the piece has to do is where that decision belongs.
A level three list is long. Fourteen vendors, eight sponsors, a twelve-line program. Those are not levels; they are a block, and the block is one item in level three with internal structure of its own. Set the block small, consistent and grouped, and let it be a texture rather than a set of lines competing for attention.
Something is actually a level two in disguise. A deadline, a price, a "free", a "no registration needed". Those stop people or let them through, and they often belong beside the date rather than in the small print.
Hierarchy from one typeface and two weights. If you want to prove the point to yourself, build the whole piece in one family, with regular and bold, and get all three levels from size, space and weight alone. It is usually enough, and the piece that comes out is easier to keep consistent when somebody asks you to add a line a week later.
Hierarchy on a sign is a special case, because there is usually one level. If the sorting produces three levels on a sign that is read from a car, the sign has too much on it rather than too little hierarchy. Signs has the sizing, and the rule there is that anything you add comes out of the size of the thing that matters.
6. See it: a flyer, annotated See it
Dolores Reyes sorts the Arbor Street Market flyer. Sixty copies at 8.5 by 11 inches, in shop windows, read from the sidewalk at about six feet, which she paced.
Known. Second Saturday of the month, 8:00 AM to 1:00 PM, free to attend, fourteen vendors, Arbor Street municipal lot. She works in Canva on the free plan.
Assumed. That the shop windows are clean and the flyer is at eye height. Dolores's assumption; a flyer taped low on a door is read at a different angle and usually further away. She can settle it by looking at where the last flyer went.
Open. Whether the vendor list belongs on the flyer or on a sheet at the gate. Settled by: the market committee, who know whether people choose by vendor.
Her first layout had everything at two sizes: a large title and everything else at 11 point. Here is the sort, with the levels marked:
Two changes did almost all of it. The hours moved from level three to level two and got a size to match, and the vendor list became one line instead of fourteen, which bought the space for the hours. Nothing was added, no new font appeared, and the color did not change.
Then the walk-back. Dolores printed one, taped it inside a shop window and stood on the sidewalk. From six feet she could read the title and, on the second look, the hours. She could not read "Arbor Street municipal lot", which was set at level two but in a lighter weight than the hours, so it read as level three. The fix was to set it at the same weight as the line above it, which is the first rule in making the jump visible and the one that is easiest to break by accident.
7. Practice: sort, set, walk back Practice
Minimum: the sort into three levels and one walk-back with a printout. That produces the result Start here asks for. Time: about 45 minutes, plus the walk. Materials: your fact list, a printer, tape, and the place the piece will live.
- Write 1, 2 or 3 beside every line of your fact list. Do it on paper, fast, and do not negotiate with yourself.
- Count the level ones. If there is more than one, decide which is the real one; the other is a level two.
- Set each level the same throughout. Pick a size for each and apply it. Do not fine-tune individual lines; that is how a level loses its consistency.
- Look at it as a thumbnail. Shrink the view until you cannot read the words. The levels should still be visible as blocks of different weight. If they are not, the jumps are too small.
- Print one and walk back. Stand at the paced distance and ask the two questions out loud.
- Write what you could and could not read in the Distance test log on Before it goes out, then change one thing and walk back again.
The squint test, and why the thumbnail is better. Squinting blurs the whole page including the contrast, so a low-contrast piece passes a squint test by disappearing evenly. A thumbnail keeps the contrast and loses the detail, which is the comparison you actually want.
The stranger glance. Show the printed piece to somebody for two seconds, take it away, and ask what it was. Two seconds is roughly what a board gets. Whatever they say is your level one, whether you meant it to be or not.
Build it twice. Make a version with four levels and one with three, print both, and walk back. The three-level version usually wins at a distance, and seeing the other one lose is more convincing than being told.
8. Your piece Your piece
9. Check your work Check your work
Criteria
- At the paced distance, can you say what the piece is before you can read anything else on it?
- Is every line in a level set the same as every other line in that level?
- Are the jumps between levels visible in a thumbnail, with the words unreadable?
- Is the date where a reader will look for it, at a size they can read there?
- Did you get the hierarchy from size, space, weight and position rather than from a second typeface and a second color?
Worked example
Against the criteria, Dolores's flyer passes the first after the two changes, and failed the second before them: the location line was level two in size and level three in weight. The thumbnail test showed three blocks, which is what it should show. The date, in her case a recurring day rather than a date, is level two and reads at six feet on the second look, which is what level two is for. And nothing was added: one typeface, one accent, the hierarchy from size and weight.
Common mistakes
- A level that is slightly bigger than the one below. It reads as an error rather than a level. Make the jump large or merge the two levels.
- Two level ones. Usually a piece doing two jobs, and the fix is in the brief rather than in the layout.
- The date in level three. It is the commonest single failure on an everyday flyer, and it happens because the date feels like a detail while you are typing it.
- Emphasis spread over half the lines. Bold on everything is bold on nothing.
- Adding a typeface to make two lines look different. It makes them look different without making one look more important.
- Testing at a desk. The desk is where the hierarchy looks fine.
What is still unresolved
This chapter tells you the levels and the relationships, not the sizes in points or inches. The number for a line read from a distance comes from Letter height from distance; the ranges for body text come from Type you can read. It also does not settle the spacing between the levels, which is what makes the groups read as groups, and that is Space and alignment. And nothing here measures whether a clearer flyer brings more people. No source this manual read measures that, and this manual does not claim it.