DESIGN INTELLIGENCEJuly 23, 2026·Mary · DEPIX Design Intelligence

Pave Where They Walk: The Design Lesson of Desire Paths

The most honest piece of design feedback in the world is a line of trampled grass. A desire path — the unofficial track worn where people actually walk, cutting the corner the paved route went around. Also called a cow path, a social trail, an elephant path, it appears wherever the designed route and the human one disagree, and it is drawn, one footstep at a time, by everyone who decided the official way was too long.

They are not random. A German researcher found that people form a desire path once the prescribed route is roughly 20 to 30% longer than the direct line — a remarkably consistent threshold for how much detour a human will tolerate before voting with their feet, in obedience to the plain principle of least effort.

Here is why a landscape architect should love the worn grass rather than resent it: a desire path is the purest user research that exists. The paved path is a hypothesis about where people want to go. The desire path is the data. It is revealed preference made physical — not what users say they will do in a survey, but what they actually do, recorded in the ground, continuously, for free, and with nobody performing for a researcher. No usability lab ever produced a cleaner signal than a muddy shortcut across a lawn.

Which is why the wisest response is the humblest one: pave the cow paths. The instructive practice — some campus planners genuinely wait to see where students wear the grass before laying the concrete — treats the original layout as a guess to be corrected by use, not a decree to be enforced. The opposite response is the one you see everywhere: a fence, a "keep off the grass" sign, a chain across the shortcut. That is a designer fighting the users to defend a wrong guess, and it never works. It only makes the wrongness official. A sketch of the idea is almost comic: the more determined the barrier, the more it advertises that the design lost the argument.

And the pattern is not limited to grass. The best digital features are paved cow paths too — emergent conventions the makers formalised only after users invented them: the hashtag was not designed and handed down — users invented it out of an older convention, wore the path, and the platform eventually paved it. Everywhere, people are drawing desire paths in the features they misuse, the workarounds they invent, the steps they quietly skip. The tracks are always there for a designer willing to read them.

The concept-phase problem is that you usually cannot wait for the grass to wear. You have to pour the path before anyone has walked it. So the discipline is to simulate the desire path in advance — to predict where people will genuinely want to go, not where your composition wishes they would, and to design for the real route rather than committing to a beautiful wrong one you will spend years defending with barriers. The most expensive paths are the ones set in concrete before anyone knew where the walking would happen.

There is an honest limit, though, and good designers hold it. Desire paths are not always right to pave. They optimise one walker's shortest route, not the collective good — the health of the lawn, the safety of a shortcut sprinting across a road, the accessibility of a step-free route that the worn track ignores, or the intended experience of a garden path deliberately routed the long way to reveal a view. Sometimes the friction was the point, chosen at the concept phase for a real reason. The skill is not blindly paving every trampled line. It is reading which desire path exposes a genuine design error and which one just shows a user optimising for something the design, rightly, refused to give them.

But the default should tilt toward listening. Every design is a bet about what people will do, and use is the only real test of the bet. Desire paths are simply the version of that test you can see from a plane. In a product, an interface, a service, the users are wearing them too — and the designers who win are the ones who read the tracks as data and pave them, instead of putting up a sign.

Sources: