A small theory of taste

Here's a thing I've noticed: when people say an interface has taste, they usually can't say why. They'll point at the spacing, or the typography, or the way a hover state lands. But none of those is the answer. Those are consequences of the answer.

The thing they're noticing, I think, is fidelity to a small set of decisions.

The pattern

Take any interface you'd call tasteful. Read it carefully — not the parts, the posture. You'll find a handful of decisions that seem to govern everything else:

One typeface, or one pairing.
One accent color.
A small set of corner radii.
A consistent way of treating empty space.
A point of view about motion: yes, no, or sometimes.

That's it. Five or six things, repeated faithfully. The taste is in the repetition, not the individual choices.

What this implies

If taste is fidelity, then the way to develop taste is not to make better choices — it's to make fewer choices, and then refuse to break them.

This is much harder than it sounds. The temptation to add one more font, one more accent, one more radius is constant. Every component feels like it could be a little nicer with a tweak. And each tweak, on its own, is fine. The damage is cumulative.

A test you can run

Pick any screen of your interface. Count:

How many type sizes are in use?
How many colors?
How many corner radii?
How many shadow depths?

If any number is above five, you have a taste leak. Above ten, you have a taste flood. The cure is not to design more — it's to delete.

Why this is engineering

You might think this is a design post. It isn't, quite. The reason taste is hard to maintain is the same reason consistency is hard to maintain in software: every individual change is locally rational, and the global property only erodes.

What you need is the same thing in both cases: a system that makes the right choice the easy one. Tokens instead of values. Variants instead of one-off styles. A small palette enforced by the toolchain, not by attention.

When the system is right, taste becomes the cheap path. And then the question "does this have taste?" stops being interesting. The interesting question becomes "what does this make easy?"

That's the one worth asking.