Why good interfaces never need explaining

A kettle handle asks to be held. A door plate asks to be pushed. We read these invitations without deciding to, and act on them before a single word of instruction is offered. The psychologist James Gibson called them affordances: the properties of a thing that tell you what can be done with it. Don Norman brought the idea into design, and it has quietly governed good work ever since.

Screens inherit the same logic. A button that looks raised asks to be pressed. A field with a cursor asks to be typed into. A card with a soft shadow asks to be tapped. When the appearance of a thing matches its function, people move through a product with a confidence they never think to notice. The interface feels obvious, which is the highest compliment most users will ever pay it.

The trouble begins when the signals lie. A line of plain text that happens to be a link. A coloured box that looks like a button but does nothing. A shadow used to decorate something that cannot be touched. Each false clue asks the person to stop, guess, and test the guess. They usually blame themselves when it goes wrong, which is the quiet cruelty of a bad interface. It makes ordinary people feel slow.

The pattern is easy to see on other people's websites and strangely hard to see on your own. A restaurant sets its phone number in the same quiet grey as everything else, and diners on their phones never realise it can be tapped. A joiner's site carries the words request a free quote in a handsome banner that reads as decoration, so almost nobody presses it. A booking page styles the confirm button and the cancel link identically, and a small, steady number of people press the wrong one. None of these failures arrive as complaints. They arrive as absences, the calls and enquiries that were never made, which no analytics report will ever label as a design fault.

None of this is solved by instruction. A tooltip, a guided tour, a paragraph of help text: these are apologies for a design that did not do its job. The moment you find yourself writing an explanation for how a control works, the honest response is to change the control, not to describe it more clearly. Good design removes the need for the manual rather than writing a better one.

Part of the answer is borrowed rather than built. Jakob Nielsen observed years ago that people spend most of their time on websites other than yours, which means they arrive carrying expectations formed everywhere else. Underlined text is a link. The basket lives in the top right. A magnifying glass means search. These conventions are affordances the whole industry has paid to establish, and any product can use them for nothing. That is quietly good news for a small business working from a template, because a decent template is mostly a bundle of conventions. The risk arrives with customisation, when a brand refresh strips the underline from links or swaps a familiar icon for something more distinctive, and trades decades of shared learning for a moment of style.

In practice the discipline is unglamorous. Make interactive things look interactive, and make static things look static. Use one consistent treatment for the same kind of action, so that learning a pattern once means knowing it everywhere. Give controls real edges and honest states, so that pressable, pressed, and unavailable each look like what they are. Resist the urge to decorate something that cannot be used, however handsome the result.

An interface that has to be explained has already failed the only test that matters.

It is fair to admit that fashion has not always sided with this argument, and that the argument is not absolute. Flat design spent a decade stripping shadows, borders and depth from interfaces in the name of cleanliness, and usability research duly recorded people hesitating over screens where nothing declared itself pressable. Yet some interactions we now consider obvious began with no affordance at all. Pulling a list downwards to refresh it signalled nothing until enough products taught it, and now it feels like a law of nature. The honest position is that a product has a small budget for teaching, enough for perhaps one new behaviour, and it should spend that budget deliberately while keeping everything around it strictly conventional.

There is a cost to all this that deserves naming, because it explains why the discipline is so often abandoned. An interface built from honest, conventional signals can look ordinary in a portfolio and unremarkable in a boardroom. The urge to be distinctive somewhere leads teams to be distinctive in the controls, which is precisely the wrong place. Character belongs in the voice, the photography, the pacing and the tone, the parts of a product that carry feeling without carrying function. A button is infrastructure. Making infrastructure surprising is a form of vanity, and the user pays for it in hesitation, a few hundred milliseconds at a time, on every visit, forever.

This matters more, not less, as interfaces begin to generate themselves. Products increasingly assemble a screen in the moment, in response to a request, and software that answers in plain language is becoming ordinary. When there is no fixed layout to learn, the affordance moves from the shape of a button to a person's sense of what the system can be asked to do. The new clue is not a shadow. It is whether the product makes its own capabilities legible, so that someone knows what is possible without having to try everything to find out.

That shift raises the stakes rather than removing them. A conversational interface that hides its limits behind a friendly prompt is the modern version of the box that looks like a button and is not. People will ask for what they cannot have, be told no, and quietly lose faith. The task is the same as it has always been. Show what can be done, in a form the person already understands, before they have to ask.

The reason these faults persist is not carelessness but proximity. Nobody can experience their own product for the first time, and after a few weeks of building or running one, every label seems clear and every button seems obvious, because you already know what everything does. The cheapest corrective in design is to sit beside one person who has never seen the thing, give them a real task, and say nothing. The moments they hesitate, the places their cursor drifts and doubles back, the instants they read a screen twice: each is a signal lying somewhere, and every one of them is invisible from inside.

When someone calls a product intuitive, they are not describing magic. They are describing a designer who took the trouble to match every signal to its function, and then removed everything that got in the way of it. The interfaces that feel intuitive a decade from now will rest on the same modest habit. Say what a thing does by the way it looks and behaves, and you will rarely need to say anything at all.

None of this shows up in a portfolio, which is part of why it is so often left undone. Getting every signal to agree with its function is patient, unglamorous work, invisible when it succeeds and expensive when it fails. It is the kind of care you notice only in its absence, and the kind worth insisting on from anyone you ask to build the things your customers will actually touch. A product that never needs explaining did not happen by accident. Someone decided it was worth the trouble.