Widget

A widget is a guest. It is a view of one app, drawn on a surface that app does not own: a home screen, a lock screen, a desktop, a notification shade. Everything odd about designing one follows from that. The size is not yours to choose, because the host offers a small set of fixed slots and the reader picks which one to spend; the refresh schedule is not yours either, because the system decides how often it is worth waking you; and the whole thing has to make sense to somebody who did not navigate to it and is not going to scroll it. The implicit promise is that it saves a trip into the app, so a widget that only says “open me” has failed at the one job it had.

That is what separates it from the two components it is most often confused with. A card is a grouping inside your own product, where you control the width, the neighbours and the context the reader arrived with. A dashboard grid is a field of tiles inside a product someone is already looking at, laid out by importance, and a reader who wants more can click through the hierarchy it belongs to. A widget has none of that around it. On a watch the same idea shrinks again and is called a complication, which is a useful word to know because it names the extreme case: the amount of information that fits when the surface is a wristwatch dial.

The craft is choosing what falls away as the slot gets smaller. A widget should recompose rather than shrink, so a size class is a different edit of the same content, not the large one scaled down or truncated into a corner. Start from the smallest and ask what single fact earns the space, usually one number or one stat and the label that makes it mean something; then let each larger size buy a specific addition (the times as well as the titles, tomorrow as well as today) rather than more of the same rows. Content that is genuinely absent needs its own quiet empty state too, since a blank tile on a home screen reads as a bug rather than as a free afternoon.

Interaction is deliberately thin. A tap is an opening move, not a control: it should land the reader on the specific screen the tile was showing rather than dumping them at the app’s front door, and anything more elaborate than a tap or a couple of tap targets belongs in the app. Text has to survive the reader’s own type size, and the tile has to be legible over a wallpaper nobody on the team chose, which is why platform widgets lean on solid or heavily blurred backgrounds instead of transparency. Apple’s Human Interface Guidelines covers the size families and the timeline model in detail, and is the right place to check before assuming a widget can behave like a small window.