DESIGN INTELLIGENCEJuly 21, 2026·Mary · DEPIX Design Intelligence

The Shopping Cart Was Perfect. Nobody Wanted It.

In 1937 Sylvan Goldman, who owned the Humpty Dumpty grocery chain in Oklahoma City, solved a problem every retailer had and none had named: shoppers stopped buying the moment their handbasket got too heavy to carry. His fix was almost comically simple — he put a basket on the seat of a folding chair and wheels on its legs — and debuted it on 4 June 1937. On paper it was a perfect object: it let one person carry far more, painlessly, straight to the register. It should have sold itself.

It didn't. For nearly two years, almost nobody would use it — and the reason had nothing to do with the cart.

The resistance was social, and it was specific. Men refused: pushing a cart felt like an admission they weren't strong enough to carry their own groceries. Women rejected it too — it looked and felt like pushing a baby pram, exactly the chore many came to the store to escape. Older shoppers thought it made them look infirm. Goldman had designed a brilliant answer to a physical problem and walked straight into a psychological one: the object asked people to perform a new behaviour in public, and that behaviour carried a meaning they did not want attached to them.

Here is the part worth studying. Goldman did not respond by redesigning the cart. He barely touched it. He redesigned the adoption. He hired models — attractive men and women of different ages — to push carts around his stores, chatting and shopping and looking entirely at ease, and stationed a greeter at the door to hand one to each customer and explain it. In other words, he manufactured social proof: he made the new behaviour look normal, even desirable, by showing people like them already doing it. Within months the resistance collapsed. Within a decade, supermarkets were being designed around the cart's existence — wider aisles, bigger stores, bulk displays — and the cart had quietly reshaped the whole architecture of retail and made Goldman a multimillionaire.

The design-intelligence lesson is the one every team learns late: the object is only half the design. The other half is the behaviour it requires, and that half is just as designable — and just as easy to ignore. Goldman's cart did not fail on its engineering; it failed on its adoption, and it succeeded only when he treated adoption as a design problem with a real solution rather than something that would take care of itself if the product were merely good enough. That is the whole content of diffusion-of-innovations research: a new thing spreads not on how good it is but on how easily people can try it, watch others try it, and adopt it without social cost. That is a concept-phase question, not a launch-day one.

Most product failures are not bad objects. They are good objects that asked for a behaviour nobody wanted to perform, from designers who assumed a strong-enough product would drag the behaviour along behind it. The self-service supermarket itself had hit the same wall a generation earlier — Piggly Wiggly had to teach a public accustomed to being served by a clerk that they should now walk the aisles and choose for themselves. Every genuinely new product carries a hidden second brief: not "what is the thing," but "what does the thing ask people to do, and will they do it?"

At Depix this is the uncomfortable reframing of what a concept actually is. It is tempting to treat the concept phase as purely about the artefact — the form, the function, the feel. But the concept phase is also where you commit, implicitly, to a behaviour: a new way of holding, wearing, sitting, buying or moving. If that behaviour is unnatural, embarrassing or effortful, no amount of craft downstream will rescue it, because the customer never gets far enough to notice the craft. Goldman's real invention wasn't the folding basket carrier; it was the realisation, decades ahead of the textbooks, that you have to design the reason people will let themselves be seen using it. The cart was easy. The permission was the hard part — and it was a design problem all along.

Sources: