Bad design examples: 8 everyday things people get stuck on

Everyone has pulled a door that needed a push. Most of us feel a bit silly and blame ourselves. We shouldn't. In almost every case, the design set us up to fail.

I make a short video about one of these moments for our Head scratchers series. Here are eight bad design examples from the series so far, the design principle each one breaks, and the fix. Each one also has a twin on websites and in AI assistants, which is where I spend most of my working day.

1. The door that says pull when it means push

A door has a handle, and a handle says pull. So you pull, and it doesn't move, because this door needs a push.

People call this a Norman door, after the designer and researcher Don Norman. Doors are one of the first examples in his book The Design of Everyday Things,¹ and a 2016 Vox video with Norman made the term widely known.² The problem is a mixed message: the shape of the handle tells you one thing and the door does another. Norman's word for the cue that tells you what to do is a signifier, and here the signifier lies.

The fix: a flat plate on the push side. There's nothing to grab, so you push. If a door needs a sign saying PUSH, the design has already failed.

Online: text that looks like a button but isn't, and buttons that look like plain labels. If people hover around wondering whether something can be clicked, it's a Norman door.

2. The supermarket sign that doesn't name what you came for

You need baking paper. You look up at the aisle signs and they list categories, not the thing in your head. So you walk up and down the aisles, hunting.

The signs are organised the way the shop thinks, not the way shoppers think. One of Jakob Nielsen's ten usability heuristics covers exactly this: a design "should speak the users' language".³

The fix: put the answer where people look. Name each zone overhead in shoppers' own words, and mark each section at the shelf.

Online: navigation built around the company's org chart or jargon. Menus called "Solutions" and "Resources" are aisle signs that don't name the baking paper.

3. Four burners, four knobs, and the wrong one lights

Four burners sit in a square. Four knobs sit in a row. You turn a knob and the wrong burner starts heating.

This is Norman's favourite example of poor mapping: controls that don't line up with the things they control.¹ Researchers have been testing better layouts since 1959, and I went through that research in a longer article on stove knobs.

The fix: lay the knobs out like the burners, so the front-left knob works the front-left burner. No little diagram needed.

Online: filters a long way from the results they change, and settings pages that list options in a different order from the screen they affect.

4. The lift buttons you have to decode

The doors are closing, someone is running for the lift, and the two buttons are arrows pointing in and out. Which one holds the doors?

A 2003 study of lift door signs found that the words OPEN and CLOSE were the easiest to recognise, and the arrow signs did poorly.⁴ Symbols save space, but only if everyone reads them the same way, in a hurry.

The fix (our suggestion): write "Open" on the button, make it the bigger one, and put it first, where hands go.

Online: icon-only buttons. A row of small symbols in a toolbar or app asks people to decode each one. A word beside the icon costs a little space and saves a lot of guessing.

5. The shower mixer that hides how it works

Hotel shower mixers often have hot and cold marks, but they wear off. So you guess, and you get a blast of cold water.

Norman points out that even a single-lever mixer leaves you four things to work out: which way changes the temperature, which way is hotter, which way changes the flow, and which way gives more.⁵ A worn label can't answer any of them.

The fix: remove the question. A progressive mixer only turns one way, from off to cold and then warmer. One example is the Elementi Tita by Robertson Bathware, which still works if its labels wear off.

Online: toggle switches where you can't tell whether a setting is on or off, and sliders with no labels at either end. If people have to try a control to find out what it does, they'll sometimes get the cold water.

6. The crossing button that never answers

You press the button at the crossing. No light, no beep, nothing. So you press it again, and again.

That's missing feedback. Nielsen's first heuristic is that a design should keep people "informed about what is going on, through appropriate feedback".³ Feedback only helps if the press really counts, though. In 2004 the New York Times reported that more than 2,500 of New York City's 3,250 walk buttons did nothing at all.⁶

The fix: a light and a beep on every press, so you know it heard you, then wait for the green man. The Aldridge iTouch by Traffic Technologies does this: its sensor light turns green when it picks up your hand.

Online: a Pay or Submit button that gives no sign it's working. People click again, and that's how orders and forms get sent twice. Show that the click registered straight away, even if the work takes longer.

7. "Unexpected item in the bagging area"

You put your handbag down at the self-checkout, and the screen tells you there's an unexpected item in the bagging area. It doesn't say which item, or what to do about it.

That's a dead-end error: it stops you and gives you no way forward. In a 2014 UK survey of 1,017 adults, 83% of self-checkout users said they found this message annoying.⁷ Nielsen's ninth heuristic says an error message should "precisely indicate the problem, and constructively suggest a solution".³

The fix: say what's wrong and what to do next, ideally with a picture. "Something extra is on the bag shelf. Lift it off to carry on."

Online: "Something went wrong. Please try again." It's the bagging area of the internet. So is a form that turns red without saying which field is wrong, or why.

8. The chatbot that won't let you talk to a person

You ask a customer service chatbot for a person. It replies "Sorry, I didn't get that", again and again.

The US Consumer Financial Protection Bureau has a name for this: a doom loop, where customers get repetitive, unhelpful answers with no way through to a human.⁸ This one doesn't need an online twin. It is one, and it's the example I work on most.

The fix: the bot should own the failure and offer a real person as soon as it starts to loop. In a Nielsen Norman Group study of chatbots, people generally saw that kind of escape hatch favourably.⁹ Our own rule of thumb is to pass the conversation along too, so nobody has to explain their problem twice.

What these bad designs have in common

Look at the eight together and a pattern shows up. Each design makes people do work that the designer should have done: guessing which way a door opens, translating a sign, matching knobs to burners, decoding arrows, testing a mixer, pressing a button again, working out an error, or escaping a chat.

Then there are the signs. The PUSH sticker on the door, the diagram beside each knob, and the hot and cold marks on the shower are all patches over a design that doesn't explain itself. Norman's point, all through his book, is that people tend to blame themselves when the fault lies with the design.¹ When you see a sign taped to something, you're looking at a design problem someone noticed but didn't fix.

How to spot bad design in your own product

You don't need a lab to find your own head scratchers. Start with these questions:

1. Where have you added signs? Tooltips, help text and "Note:" lines explaining a control often mark a design that needs fixing.
2. Does every action get an answer straight away? A click, a tap or a message should show it registered, even if the result takes longer.
3. Do your labels use your customers' words? Check your navigation and buttons against the words people type into search or say to support.
4. Does every error say what went wrong and what to do next? If not, rewrite it before you touch anything else.
5. Is there always a way out? Back a step, to a person, or to start again. No doom loops.
6. Have you watched real customers use it? Ask a few people to do one task while you watch, and keep quiet. The places they hesitate are your Norman doors.

That last one is the heart of usability testing, and it's the quickest way I know to find out where people get stuck.

Footnotes

1. Don Norman, The Design of Everyday Things (1988; revised and expanded edition 2013). The book covers doors, signifiers and the mapping of stove-top controls. ↩ ↩² ↩³

2. Vox, It's not you. Bad doors are everywhere (February 2016), with Don Norman and Roman Mars of 99% Invisible, as shared on Norman's own site. ↩

3. Jakob Nielsen, 10 Usability Heuristics for User Interface Design, Nielsen Norman Group (1994, updated since). ↩ ↩² ↩³

4. Yeoun, M. and Harada, A. (2003). An Evaluation of Recognizability for Elevator's Open/Close Signs. Bulletin of the Japanese Society for the Science of Design, 50(2). ↩

5. Don Norman, The Design of Everyday Things, revised edition (2013), chapter 4, "The Faucet: A Case History of Design". ↩

6. Michael Luo, "For Exercise in New York Futility, Push Button", The New York Times, 27 February 2004. ↩

7. Viewsbank survey, reported in Nearly one in two UK shoppers need help with self-service checkout, Retail Customer Experience (May 2014). ↩

8. Consumer Financial Protection Bureau, Chatbots in Consumer Finance (June 2023). ↩

9. Raluca Budiu, The User Experience of Chatbots, Nielsen Norman Group (2018). ↩