What I Learned Redesigning My Logo

When I rebuilt this site in February, I gave it a logo the same afternoon. Two outlined triangles, tip to tip. The top one with a short stem under it was a Y. The top bar, the long diagonal and the bottom bar made a Z. Together the two triangles were an hourglass, because time is the thing I think about most, and I liked that a mark for my initials also said that. It took an hour or two, it went into the browser tab, and I did not think about it again for six months.

This month I started to dislike it. Any two triangles tip to tip are an hourglass. The Y and Z were there if you knew to look, but so is a Y in most hourglasses. It could have belonged to a timer app or a scheduling startup. It said nothing that was only mine.

Around the same time I had been noticing startup logos. There are so many companies now, and every one of them has a mark, and a surprising number of the marks are ugly. I could see that they were ugly. I could not say what was wrong with them, or what a good one would have done instead, and that bothered me more than the logos themselves. I build things for a living and I could not explain the difference between a good mark and a bad one.

So I decided to learn by doing one properly, from the beginning to the end, on the only brand I own. Mine.

It turned out to be far more complicated than I imagined. The redraw took about sixty versions over two days, for a shape that in the end is two strokes and a triangle. This post is about what those sixty versions made me think, because most of it I had not thought before.

Where the logo actually shows up

The first thing I did was ask a question I had skipped in February: why does one person need a logo at all? A company needs one because strangers have to recognise it on a shelf or an app store. Nobody needs to recognise me. My name is at the top of every page on this site, and a name already does the job of a mark.

Then I checked where the mark actually appears, and the answer surprised me. It does not appear on any page. The homepage does not show it, the posts do not, the footer does not. It shows up in exactly three places: the browser tab, the icon a phone puts on its home screen if someone saves the site there, and the top corner of the card that appears when someone shares a link to a post.

Look at those three places and they have something in common. A browser tab is sixteen pixels tall. With twenty tabs open, no name is readable at that size, but a shape can be. A home screen icon is a rounded square the size of a fingernail, and three characters of a name in it read as a smudge. On the share card, my name is already printed along the bottom as the signature, so a second copy in the corner would repeat it. Every place the logo lives is a place where the name is either unreadable or already there.

That gave me the honest answer. A person needs a logo for the places their name cannot go. Nowhere else.

There are more of those places than the three it lives in today, and the ones I would actually use make the same point harder. A red seal, the kind pressed onto a notebook page or a print, since my surname 朱 is the word for that red. The chest of a basketball jersey, where a team would put its crest. A sticker on a snowboard. The end of an email signature. None of these will take a name either. A seal works in negative space. Embroidery cannot hold a thin line. A sticker gets read from across a room, and an email icon is smaller than a favicon.

So the places told me the constraints before I had drawn anything. The mark has to survive at sixteen pixels, because that is the smallest place it lives. It cannot depend on its corners, because the home screen rounds them off and a round crop takes them away entirely. It has to work in one colour, on a dark card, on a light page, and in a single red on paper. None of these were my choices. They came with the places.

I knew all of this when I sat down to draw. I still started at the wrong end.

Trying to pack every meaning in

My idea of a logo, going in, was a suitcase. I had things I wanted to pack: the letters Y and Z, the hourglass, the number two because it was my jersey number and I have always felt it fit me, and something from my Chinese name. Design, as I imagined it, was finding a shape big enough to hold all of them. A mark that carried four meanings was doing four times the work of a mark that carried one, so more was better.

The first versions came straight out of that idea. The Y got its stem back. The Z was drawn heavier so it read before the other diagonal. Then a version where the Y and the Z shared a stroke, and one where the top-right corner of the Z curled round into the hook of a 2. That last one had everything I had asked for, and it looked like the monogram on a hotel towel.

Every element was present and correct, and together they were ugly, and I could not say why. I could tell in a second that a version was wrong. I could not produce a single sentence about what was wrong with it. Being able to tell and being able to make turned out to be two different abilities, and I had only the first one. Until that week I had assumed the first one was most of the job.

Good logos are two things at once

What got me out of it was looking at logos I actually admire, which are mostly basketball players’ logos. Kawhi Leonard’s is a hand, and the fingers are his initials and his number at the same time. Kyrie Irving’s is his initials folded into each other so that the K and the I share strokes. None of them packs things side by side. Each one is a single shape that happens to be two or three things, and you notice the second thing a moment after the first.

That is a different job from packing. Instead of a shape that holds four meanings, you want one shape that already is a second thing, so nothing extra has to be added to carry the second meaning. To find that, I put everything of mine on the table, not just the letters, and looked for two things that already had the same skeleton.

The match was in my given name. The character 益 starts, at the top, with two short strokes that open upward, the component 丷. The old form of the character below that is a bowl with water spilling over its rim. Two strokes opening upward are also the top half of an hourglass with its lid off. They are also the two arms of a Y. A solid triangle under them is the sand that has already fallen, the base a Y stands on, and the bowl that has filled up. One shape, three readings. The number two comes along without being drawn, in the two strokes above and the two materials, thin lines over solid mass.

Nothing in that shape is there in order to mean something. Each part is just what it is, and the meanings are ways of reading it.

Subtraction, starting with the Z

Once the shape had its own logic, the Z became a problem, and I spent several rounds trying to get it back in. A bar across the top of the two strokes turns them into a drawn triangle, and with the solid triangle below, the diagonal and the base make a Z. It worked as a letter and it closed the open top, which was the point of the form. So I tried half a bar. A bar with a gap in the middle so the two strokes still showed. A bar lower down, at the waist, which is where 益 itself has one, and which turned the mark into a yen sign. Sliding the top and bottom halves apart so a Z appeared in the outline. Each of these was a real Z. Each one made the mark a little worse than the version without it, and after five or six tries the pattern was too clear to argue with.

I had thought of taking things out as a matter of nerve, of being willing to lose a meaning. It is closer to a matter of listening. The shape had already decided what it was, and every stroke I added to carry a meaning it had not asked for showed up as exactly that, a stroke that was only there to mean something. A good mark has none of those.

There was also a better reason to let the Z go, which I only worked out afterwards. The Z is my surname. The players whose logos I had been studying build theirs from surname initials, and in the West that makes sense, because a surname is the half of a name that tells people apart. Given names repeat constantly, surnames run into the hundreds of thousands, and a formal room addresses people by their last name. Chinese works the other way round. A hundred surnames cover most of the country, 朱 is one of the most common of them, and a room where several people answer to it is normal. The given name is where a Chinese name carries its information. The letter Z is weaker still, because in pinyin it is also Zhang and Zhou and Zhao and Zheng, one letter shared by a couple of hundred million people.

So the part of my name that is actually mine is the given name, and the given name is the part that had fitted into the shape. The mark had kept the piece of the name that was me and dropped the piece that was everybody. I had spent five rounds trying to put back the wrong letter.

Rules I only found at sixteen pixels

With the direction settled, the remaining rounds were about details, and this is where I found out how little I knew. Every rule I ran into announced itself by being broken, and only at one particular size or in one particular place.

The first real candidate had a hollow triangle on top of the solid one, an outline over a mass. At full size it was the best thing on the sheet. At sixteen pixels the hole in the outline closed up, because a gap two units wide has no pixels left to be a gap with, and the mark turned back into two solid triangles, which is roughly where I had started in February. The next version, two curved strokes meeting at the waist, looked like the open form I wanted at drawing size and read as a plain X in the corner of the share card, where the eye takes in the shape at once and never gets to the triangle. Strokes of even width and strokes that thicken at the top and narrow toward the waist look nearly the same side by side, and on a dark card one looks drawn by a hand and the other looks stamped by a machine. Straight strokes read as the chevron from any icon library, and it took a slight outward bow, the way a glass bulb swells, to make them mine. Once the strokes bowed outward, the sides of the triangle had to bow inward, the way a pile of sand settles, because a version where both curved the same way read as one material instead of two. The point where the two strokes meet wanted to be rounded, or the mark stopped looking like something a brush could have made. And a base slightly narrower than the strokes above it read as top-heavy even after I measured the ink and found it split almost exactly in half.

I had no idea any of these were rules. Each one is obvious once you say it, and I would not have said a single one in advance. Someone who does this for a living carries all of them and a few hundred more, and applies them before the first stroke instead of after the sixtieth version, which is why their first version tends to look like my sixtieth.

Designers know why, I only feel it

It would be easy to file all of this under taste, and taste has a reputation for being something you either have or do not. What I saw over those two days pointed the other way. My reactions were subjective in the sense that I could not justify them, and every one of them had a reason anyway. Each time I said a version was wrong, the reason turned out to be something that could be written down, like the hole closing at sixteen pixels or the bar sealing the top. I had the reaction and someone else could have supplied the sentence. A designer has both.

The last rounds made this concrete. Once the shape was settled I put every dimension on one grid unit, so that the arm widths, the neck, the corner radii and the two chambers were all multiples of the same number. When I drew the construction diagram by hand and then measured the actual path, they disagreed: the neck was wider than the drawing claimed and the tip sat lower. So now one script writes both the mark and its diagram from the same numbers, and they cannot drift apart. The rule I kept from that week is that the structure follows the ruler and the posture follows the eye. When I could not tell two versions apart at real sizes, the one with the explainable numbers won. They may feel no more strongly about a shape than I do, but when they feel it they know why, and knowing why lets them fix it in one move instead of forty.

Which brings me back to the ugly startup logos. I still cannot fix them, but I think I can now say what is usually wrong. Most of them are suitcases. Someone had a list of things the company stood for and found a shape that held all of them, and every item on the list is visible as a stroke that is only there to mean something. The good ones are one shape that happens to be two things, drawn by someone who knew the sixteen-pixel rule before they started.

So the respect I came out with is specific. It is respect for the size of the rulebook, and for how completely a finished mark swallows it. Mine looks like nothing much now: two strokes and a triangle. Sixty rules, followed, look like that.

It reads as an hourglass mid-run, as a Y, as the top of my given name. It does not read as a Z, and I have stopped wanting it to. The meaning I had to leave out was the one that was never mine to begin with.