DESIGN INTELLIGENCEJuly 25, 2026·Mary · DEPIX Design Intelligence

The Door That Needs a Sign Already Failed

You have met the door. It is glass, it is handsome, it has a nice vertical handle on both sides - and taped to it, at eye level, is a small laminated sign that says PUSH. That sign is not a helpful courtesy. It is a confession. It is the design admitting, in writing, that it failed to do the one job a door has: tell your hand what to do without a word.

Designers have a name for this. The industrial designer and cognitive scientist Don Norman called them, affectionately, "Norman doors" in his classic The Design of Everyday Things, and the idea underneath is one of the most important in all of design: the affordance. An affordance is the property of an object that tells you how to use it just by looking at it. A flat plate affords pushing. A graspable bar affords pulling. Put a flat plate where the door opens and a pull-handle where it doesn't, and you have built a machine for producing that little PUSH sign.

Here is the contrarian point, and it is pure concept phase. The sign is not a communication problem to be solved later by the signage team; it is a design decision that was skipped upstream. Every time you see instructional text bolted onto an object - PUSH, PULL, LIFT HERE, THIS WAY UP, INSERT CARD THIS DIRECTION - you are looking at the residue of a concept-phase decision that never got made. Somebody chose a form, the form did not carry its own meaning, so language had to be imported to patch the gap. The door handle is the most visible example, but the pattern is everywhere.

This matters because instructions are expensive in a way that never shows up on an invoice. They cost cognitive load: a fraction of a second of reading, deciding and acting, multiplied by every person, every day, forever. A good affordance spends none of it. That is the whole promise of usability as a discipline, and it is why the field of human-computer interaction obsesses over the same question a door poses: can the form alone tell the user what is possible? The Nielsen Norman Group built an entire practice on the finding that people do not read - they scan, guess, and act on what the shape implies.

The deeper discipline underneath is poka-yoke, mistake-proofing: the best way to prevent an error is to make the wrong action physically impossible and the right one obviously inviting. A SIM card with one clipped corner cannot go in the wrong way; nobody needs a diagram. That is a concept-phase decision - the shape of the slot - doing the work an instruction would otherwise do badly. When human factors engineers talk about "designing out" error, this is what they mean: move the intelligence from the label into the form.

So the humble PUSH sign is a perfect diagnostic. When you see one, you are not looking at a labelling choice; you are looking at the fossil of a moment when a team could have made the object self-evident and instead deferred it downstream, where the only tool left was words. And words, as any signage professional will tell you, are the most expensive and least reliable way to change behaviour - they must be read, understood, believed and obeyed, in a language the reader happens to speak, in the half-second before their hand is already moving.

The lesson generalises far past doors, which is why bodies like the Interaction Design Foundation teach affordance as a first principle. Any product, interface or door that needs an instruction to be used correctly has quietly relocated a concept-phase decision into a permanent support cost. The instruction is a tax you pay forever for a decision you declined to make once, early, when it was cheap.

At Depix, this is the thesis in its most physical form. The best design does not explain itself, because the decision to make it self-explanatory was taken at the start, in the shape. The next time you push a door that says PULL and it doesn't budge, don't blame yourself. You just met a concept-phase decision that someone skipped - and the sign is the receipt.

Sources: