Signifier

The most common sentence in a design review is "I can't tell whether this is clickable." For a long time that problem got reported as a missing affordance — affordance being the property that lets you guess, from how a thing looks, what you can do with it. The trouble is that one phrase covers two completely different situations. In the first, the thing genuinely can't be clicked. In the second, it can, and the screen just never says so. The first needs a feature built; the second needs a mark added. Calling both by the same name means the conversation starts in the wrong place.

The sheet of paper reading "push" taped to a door shows the difference. That door could always be pushed; the property was there from the start and simply wasn't reaching people. The paper doesn't create a capability, it announces one that already exists — and a signifier is that piece of paper.

On a screen the distinction matters more than it does on the door. A door has a body — a handle, a flat plate — whose shape leaks information about what to do. A screen has color and shape on a sheet of glass and nothing else. Marking an element clickable in code and having a person notice it are separate events, so every cue has to be placed on purpose.

Cues divide into three kinds by how they attach. The first two are visible before anyone acts; the third only appears once an action is already underway.

1. Cues carried by appearance — borders, fills, shadows, underlines. Underlining a link is the oldest example of this, and giving a button a filled background so it lifts off the surface is the same move. These arrive during a glance, which makes them the strongest cue to lead with.
2. Cues carried by words and symbols — the "push" sheet of the group. A verb inside a button, a downward chevron meaning there's more to expand, the six-dot grip that says a row can be dragged. The harder an action is to express through shape, the more the work falls to this kind.
3. Cues that only exist during the interaction — the pointer turning into a hand, a hover state where the color shifts under the cursor, a drop target lighting up once a row starts moving. These reach only people who already touched the element, so they can't announce a capability — they confirm someone is on the right track.

Three rules decide how much of each to use, and where.

Same kind of thing, same cue. If blue text meant a link on one screen, blue text has to mean a link on the next one. When the meaning shifts between screens people click things one at a time to find out, and after the fourth or fifth check the product feels untrustworthy.

Don't advertise a capability that isn't there. When an illustration or a statistics card carries a button's border and shadow, people click it, and when nothing happens they don't conclude they made a mistake — they conclude the product is broken.

Don't lean on a single cue. A button marked only by color is an unmarked button for someone who has trouble distinguishing colors, and a cue that lives only in a hover state disappears on a touch screen. Layer color, border and wording so that when one channel fails the others still carry the message.

Four situations bring the word into the room.

The first is a design review. You put one screen up and ask whether a newcomer could separate what's clickable from what's only there to read. At Pagedeck, a gray "Share" label in the top-right corner was skipped by three of eight test participants. Noa changed no wording at all — a border and a fill turned it into something button-shaped, and on the next round all eight found it within five seconds.

The second is maintaining a design system, where reusable rules and components are held to one standard, and cues get baked into the components. Decide once which border and fill a button carries, and what gets stripped away in a disabled state, and no screen has to make the call again.

The third is introducing an interaction that leaves no trace. Nothing on a screen states that rows can be reordered by dragging, and the same is true of long-press and swipe. Without a deliberate cue the capability exists and nobody finds it, which is why teams add a grip mark or a button that does the same job.

The fourth is an accessibility pass. Shadows and borders don't reach someone using a screen reader; the fact that an element is a button has to be written in the code before it can be spoken aloud. The visible cue and the cue recorded in markup are two separate pieces of work.

Against feedback, the split is chronological. Feedback tells you what happened in response to an action, so it arrives after a hand moves. A signifier sits there before. Lose the first and people never start; lose the second and, unsure whether anything registered, they press the same button twice.

"The feature is there, it just has no signifier. Nobody thinks to click here."

Nothing needs building — something needs announcing.

"That card looks like a button but doesn't do anything. Let's take the cue off."

It's advertising a capability that doesn't exist, so tone the appearance down.

"If that cue only lives in the hover state, it doesn't exist on a phone."

The cue vanishes wherever there's no mouse.

Pagedeck starts design reviews by flipping the screen to grayscale once. If what's clickable is still distinguishable with color gone, no cue is resting on color alone. Last time the status marks in the document list collapsed into the same gray, so Noa added a small shape in front of the label.

The first mistake is assuming more cues are better. A missing cue caused the problem, so adding more looks like it should keep helping. But once every element has a border and a shadow, which ones actually matter disappears again. Five emphasized buttons on one screen means deciding which to press first, and that load isn't much lighter than having no cues at all. Keep the strongest cue for one element and step the rest down a level.

The second is treating stripped-down screens as the more refined choice. A layout with the borders and shadows removed genuinely looks tidier in a mockup. The catch is that everyone reviewing the mockup already knows where to click. To a first-time visitor the same screen is a panel of text. When you do strip things back, pair it with progressive disclosure — pushing rarely used things one step out of the way — and leave the cues on the actions people reach for daily.