Touchscreens are terrible and you shouldn't put them in your products

Display screens these days are great. Cheap, full color displays running at hundreds of pixels per inch are everywhere. The laptop I’m typing this on is three thousand pixels across. My phone runs at close to 500 ppi: one has to work very hard to see individual pixels without a magnifying lens. Text is smooth and clear, pictures bright and fully focused, information density can be (and sometimes is) nice and high, and life is good.

Many screens have a layer that responds to your fingers, allowing one to interact with the screen by looking at the images on the screen and touching or swiping the screen itself. A touchscreen!

Done right, the effect feels like magic. It’s as though rubbing your finger on your eyeglasses was enough to manipulate objects in the real world. These touchscreens are the main way one uses a phone, for example; billions of devices with this amazing interaction mode. I’ve owned phones like this for about fifteen years and they’re pretty nice.

Done wrong the effect feels like looking at the world through dirty, smudged glasses after having had a minor stroke. Stuff looks weird and responds unpredictably. There’s no mechanical sympathy to be had.

These things are built such that both the onscreen graphics and the responses to touch input events are done in software. (For sufficiently cheap computers, people like to say “firmware” instead of “software”.) So far so good.

But people who build hardware like to freeze the hardware parts of a manufacturing design as early as possible because it reduces uncertainty in the manufacturing process. Software to operate the touch interface, by contrast, can usually be tweaked right up until (and often after) the device ships. For something like a phone or a tablet, where the touchscreen and the software are pretty much the whole product, this works great because everyone is incentivized to make things work smoothly: otherwise the product is terrible and no one will buy it.

But for everything where the touchscreen display is just another part of the device, this leads to trouble. Suddenly, people who need to build a big thing that’s incidentally supposed to have knobs and buttons on it start thinking, “Hey, a touchscreen and tiny underspecced computer are cheaper and more flexible than a bunch of physical knobs, and we can commit our hardware way earlier and ship sooner,” and everyone goes ahead and removes all the buttons and knobs and replaces them with the screen.

This causes problems for users.

• Touchscreens require visual attention. You have to look at them and read them in order to interact, which means that they’re only good for situations where you’re looking anyway: phones, tablets, etc.

  • This is the one that’s baffling: It’s impossible to use a touchscreen in a car except by taking your eyes off the road! The very same NHTSA that requires us to have a backup camera and airbags and tiny windows and every car built like a bunker has somehow decided that it’s okay to put driver-facing systems in cars that cannot be operated except by failing to concentrate on driving.

  • Let’s review: To reduce manufacturing costs, they give us automotive controls that can only be operated by taking your attention off the road.

• Things move around on the screen. The things you touch are in a different place every time because too many buttons are being replaced, so the programmers are asked to cram lots of “buttons” into the software, and they inevitably move around on the screen because they don’t all fit at once.

• Touchscreens don’t work when they’re damp, so if you have a touchscreen fitness tracker, plan on not being able to use it if it rains, or if it’s humid, or if you sweat when you exercise, or if you sweat in a hot car. True story: my former employer made otherwise excellent fitness trackers that were advertised as waterproof and had swim tracking functionality—but that had a touchscreen instead of buttons, and therefore didn’t work when you got them wet. Bug closed, they’d say: functions as designed.

• The screens lag, breaking the illusion of “just pushing buttons”. This could happen because of underspeccing the computer that drives the screen, or because of latency in the screen driving hardware, or a bunch of other reasons. So instead of pushing buttons, users are left to “push the touchscreen to try to get it to register the fact that I pushed it.” Sometimes the software doesn’t register a swipe as a touch, so you have to touch super carefully. It’s much, much worse than a button.

  • Fun anecdote: in my Subaru, touching the screen results in up to four seconds of delay before the push is registered. Turning on Bluetooth takes tens of seconds once the “connect phone” button has registered a press. It’s not generally possible to know when a press has been registered except by waiting tens of seconds to see whether music starts playing. Then you get to guess, up to five times a minute if you’re lucky, whether the problem was with Bluetooth or with the screen or with the phone or with something else. Instead of just showing up and playing music and starting to drive.

  • I have an external monitor that I plug into my laptop, and it has an obvious half-second of extra lag compared with the basically instant laptop screen. The external monitor is useless for video calls, games, animations, or anything that requires accurate depiction of motion. I have no idea why it lags this way.

  • It’s worth acknowledging that CRT screens were terrible in every way except for their supremely low latency. Coupled with fast hardware like a Nintendo Wii controller, all one had to do was to will the controls to do something and it would already be happening over on the TV. With laggy modern displays, fast twitch arcade games like Wii Sports basketball and baseball used to be awesome, and now just don’t work. But nobody mourns the loss of a CRT’s x-ray radiation, power consumption, and bulk.

• Occasionally a team of heroic firmware developers can make things work pretty well anyway (c.f. my former employer!), but most consumer goods makers cheap out on programmers and end up with buggy, shitty software. So the buttons jump around, or freeze, or just stop registering touch events.

• When the screen inevitably fails, there’s nothing to fix. Knobs can be replaced or soldered in, but replacing a car’s touchscreen and the computer behind it usually means spending thousands of dollars for a by then several generations old touchscreen and computer. Even if you pay extra to Apple to replace an iPhone screen the phone is never quite right again.

  • Picking on Subaru again: The touchscreen in my wife’s Subaru got hot in the sun once and now doesn’t work. (Touch events fire randomly, or not at all. It beeps unbidden, sometimes a lot.) Replacing it would cost four thousand dollars, over ten percent of the purchase price of an entire pretty nice brand new car.

But in very rare circumstances, “touch” is great. The clickwheel iPod was awesome, and Apple’s trackpads are the best input devices I’ve ever used. They’ve even used haptic pseudoclicks for years, invisible and therefore perfect. (How other manufacturers can’t figure out how to make trackpads as good is kind of a mystery; Apple’s been superior for as long as I can remember.)

Most mobile phones, at least as shipped, have incredible touchscreen interfaces including sometimes vibrating just enough to let you know you hit a “button”. The swiping behaviors are nicely dialed in, and pretty much always do what you want. When done correctly you don’t even have to think about it: the swipe works without you even noticing. But always when you’re supposed to be looking directly at the screen anyway. And only when a real button or knob wouldn’t have worked to begin with.

Even the slightest imperfection in the user experience—perceiving lag, questioning whether it will work, or needing to look at the road—make a whole product feel cheap, amateurish, and broken.

Real physical buttons and switches and knobs for me, please. And keys.