experimentsOctober 14, 20262 min read

Building a playground for design experiments

A small place to explore motion, test an interaction, and keep what you learn.

Press Play. Both dots leave at the same moment and arrive at the same moment, and one of them still feels faster.

Linear0, 0, 1, 1
Ease out0.22, 1, 0.36, 1
Same distance, same timeBoth dots leave together and arrive together. Only the curve differs.linear vs cubic-bezier(0.22, 1, 0.36, 1) · 150, 300 or 600ms

Same distance, same time, different trip

The top dot moves at a constant speed. The bottom one uses cubic-bezier(0.22, 1, 0.36, 1), which covers most of the distance early and spends the rest of the time settling in. An easing function maps elapsed time to progress, and MDN's page on cubic Bézier easing explains the four numbers. The little curve beside each track is that mapping, drawn from the same control points the dot uses.

.dot {
  transition: transform 300ms cubic-bezier(0.22, 1, 0.36, 1);
}

Switch to 150ms and the difference almost disappears, because both dots read as instant. At 600ms the linear dot starts to look dragged along on a string. I'd rather feel that in five seconds here than argue about it in a design review.

None of these numbers is a rule. The distance, what the motion is for, and how often someone repeats it all go into a duration. A menu you open forty times a day should move faster than a page transition you see once.

Keep the version that's losing

When a second version starts to look good, the first one tends to get overwritten. A playground makes keeping both cheap, so you compare them side by side with the same content instead of against your memory of yesterday.

There's research on comparing alternatives, though nobody in it was tuning easing curves. In Dow and colleagues' 2010 study, 33 non-professional designers made web ads. With time and feedback held equal, the group that worked on batches of alternatives in parallel produced more varied designs, and their final ads did better on click-through and expert ratings than the group that kept iterating on one.

Carrying that over to motion is my own extrapolation. It's still enough for me to keep the alternatives on screen while a decision is open.

Somewhere for the questions to live

Each piece in the Lab has a short note and the motion values it actually uses. The text toolbar asks how a selection should answer each change. The magnetic select asks how picking one thing might push its neighbours around. They can stay small.

The real test comes when a timing goes where it will be used. Two dots on a track can compare curves. They can't tell you whether 300ms belongs in a checkout, a music player, or a settings page you open once a year.