A button should make its purpose clear

Recognising a control still leaves us to understand what will happen. A comparison of three ways to save puts wording, clues and outcomes to the test.

Graphic and interaction design
Two paper paths lead from a blue key: one to a tray of sheets, the other to a sheet leaving a frame.
A recognisable control can lead to different outcomes. AI-generated illustration: a metaphor, not a reproduction of an interface.

‘Continue’ is a convenient word for whoever designs a page. It takes up little space and promises something next without committing to what that might be. It can be less convenient for the person about to press it: sometimes it opens a summary; sometimes it sends a request. The gesture is identical; the consequences can be very different.

The button itself, meanwhile, may be impeccable. It has the right colour, a crisp border, a brief animation when we touch it. Everything suggests that we can act there. We still have to work out what we are setting in motion. It is a small distance, easily lost while we concentrate on how the page looks.

Before the gesture

Recognising a control and anticipating its effect are separate steps. A raised rectangle may suggest a button without telling us whether we are saving a draft or making it public. Its wording should help bridge that distance, together with the information around it.

A distinction made by Don Norman is useful here. An affordance concerns the actions possible in the relationship between an agent and an environment; the clues that allow us to recognise them are another matter. Drawing a shadow beneath a shape on screen does not, by itself, make it operable. It may suggest that it is worth trying. In Affordances and Design, perception and conventions of use are kept distinct from the material possibility of a gesture.

The term signifier, used in Signifiers, not affordances, directs attention to those perceptible clues. A border, a word or a position beside a field helps us understand how to proceed. For the designer, the question becomes more precise: what am I leading someone to believe is possible here?

That promise can remain too vague even when the control is perfectly recognisable. Knowing where to press still leaves open the question of what will happen.

What ‘Save’ leaves unsaid

For this note, I have prepared three small controls that add a text to a temporary collection. The action is the same, but its presentation changes: a bookmark symbol, the word ‘Save’, a label naming the destination. The three rows can be tried separately; each works with its own copy of the same note.

The same note, three controls

Each row has an independent collection. Saved items exist only on this page and are cleared when it reloads. The wait of roughly half a second is simulated.

The symbol alone

Waiting for the gesture.

One word

Waiting for the gesture.

Destination and outcome

Collection empty.

The variants change several elements at once, allowing us to examine design choices. They do not measure readers’ understanding.

The bookmark may evoke a familiar gesture for someone who has encountered it elsewhere. In a space where every row repeats the same function, a small icon may be enough. On its own, however, its meaning has to be reconstructed: will it keep the text here, download it or add it to the browser’s bookmarks? Context can answer; when it is absent, the symbol has to carry a request we may have assigned to it too hastily.

‘Save’ offers a verb. ‘Save to collection’ also narrows down where to look for the result. That is a useful difference in this example, without becoming an invitation to lengthen every label. A toolbar used every day and a page opened for the first time can afford different amounts of explanation.

The variants change several elements at once. Looking at them and pressing them helps us discuss those choices; it does not demonstrate that the third makes every person understand the control more quickly. That would require a comparison with real readers. Here we can establish something more limited: what information each version makes available.

Afterwards, silence

The third control continues the conversation after the click. During the simulated wait, it shows that saving is in progress; it then confirms that the note is in the demonstration collection. The next action lets us remove it. The result remains readable, and ‘saved’ has a precise scope: reloading this page empties the collection.

The other two respond with ‘Done’. The gesture has received a response, but that response says little about what now exists and where to find it. Even a successful animation can stop at this level: it acknowledges the touch, then hands back the uncertainty we wanted to resolve.

Feedback must explain the outcome it can actually guarantee. If an application has merely received a request, ‘Request received’ and ‘Operation completed’ are not interchangeable. The difference concerns the state of the work before it concerns the tone in which we describe it.

This also reveals the limit of design focused on the moment of clicking. That moment follows an expectation and leaves something behind. When the result immediately disappears, or can no longer be distinguished from what was there before, the person using the page has to reconstruct whether the gesture had any effect.

Two trays side by side: a blue sheet hangs above the first; in the second, the sheet rests inside.
Starting an action and confirming its outcome require different information. AI-generated illustration; it does not document a real system.

A control for different bodies

The clues change with the way we reach a button. Moving a pointer over it can reveal a border or some text; when touching a screen, that preparation may be absent. Information needed to decide should also be available without waiting for that movement.

With a keyboard, the problem takes another form: we need to recognise which control will receive the action. The focus outline, marking the element we are reaching, gives the next gesture a position. Removing it to tidy up the design risks removing precisely that clue. The W3C guidance on visible focus concerns this ability to find our way.

In the demonstration, the three controls can be reached with Tab and activated with Enter or Space, following the expected behaviour for buttons. The icon also has an accessible name that a screen reader can announce. That name serves a different mode of use from looking at the symbol; writing it into the code does not automatically resolve ambiguity for someone who sees only the icon.

The space around a control also belongs to its use. A target that is too small or too close to another asks for a precision that not every hand can offer in the same way. The W3C guidance on target size and spacing helps us examine this relationship. A design has to be reachable as well as recognisable.

Making a choice easy

Making a control clear, though, leaves another question open: which action are we making easy? A page can give ‘Accept all’ great prominence and leave refusal in a barely visible line. The first button is perfectly understandable. It is the relationship between the alternatives that becomes opaque.

The same care used to guide can therefore tilt a choice. Colour, spacing and wording carry weight even when each element works on its own. Counting how many people pressed the most prominent control is not enough to establish how free they were to choose another route.

I return, then, to ‘Continue’. If the next step sends a request, ‘Send request’ sets out a consequence we can pause over. That pause may be necessary. A button that makes its purpose clear must also leave time to decide not to press it.

Two trays side by side: a blue sheet hangs above the first; in the second, the sheet rests inside.
Starting an action and confirming its outcome require different information. AI-generated illustration; it does not document a real system.