In advertising, no phrase is more cliched than "new and improved." Simplicity and ease of use in an existing product are often no defense against incremental "improvements". In many cases, results are predictable: a perfectly reasonable interactive experience suffers some unfortunate change.

Usability testing for entirely new systems can in some ways be less problematic than for systems that have been upgraded. In the latter case, there's a strong temptation to take shortcuts. Designers may have preconceptions about the importance of what has and has not changed, and these expectations may lead to potential problems being ignored. One of my students gives an example:

One day my young cousin asked if she could play with my new handheld video game. A few moments later she came back to me and said that it didn't work. When she gave it back to me, I saw that she'd put in the game cartridge backward. With my older game system, this would have been impossible because of a little sliding mechanism that makes sure the cartridge goes in the right direction, but my new game doesn't have the same design. It took me several minutes to get the backward cartridge out.

It is possible that the designers believed that users would be experienced enough with previous products to use the new product correctly, or that visual guides instead of physical constraints would be enough for correct usage, or that this problem only arises in some unlikely circumstances. For individual users who run into difficulties, of course, this is little consolation.

Usability evaluation of upgrades to a system can identify obvious problems. Sometimes, however, the problems that arise can be subtle:

In the store where I work, we sell a combination night light/room deodorizer, which plugs into an electrical socket and releases a scent. It has a little flower on top. The problem with our product is that it covers up both sockets in a standard wall outlet. If you plug it into the top socket, the light bulb covers the bottom one, and if you plug it into the bottom socket, the flower part covers the top one. People are going to have to unplug the night light every time they need to plug something else in, unless they have a lot of outlets in the room.

In this case, usability problems with the improved night light might not appear if it is evaluated in isolation (if an evaluation is even considered for such a simple device) but rather only in the larger context of how the light is integrated into a real environment. The functionality added to the light is offset by the loss of existing functionality that is not, strictly speaking, part of the light itself. Varying context can makes potential problems harder to identify.

Other problems can be seen when new functionality is given to users who don't understand it or don't need it:

There's a new feature that Volkwagen has put in their Jettas. If you turn your key all the way to the right when unlocking the door, all four of the windows roll down at the same time. For the life of me I do not understand why a car needs to have all the windows roll down when you want to get in. When I first got the car two years ago I was constantly frustrated, but eventually I learned the art of unlocking my car. When my boyfriend or one of my family members wants to borrow my car, I always have to remind them that my car has this crazy feature.

It might seem obvious to an automobile designer that in a hot climate, opening all the windows easiy and quicky would be a desirable function. For one user, however, it's not even conceivable how this could be useful.

Yet another case shows how a lack of distinguishing features between an old product and its replacement can lead to problems:

When I bought a new car last year, I put a "North Carolina State University" sticker in the rear window. The new sticker looked just like the one in my old car, with red letters on a clear plastic background. I put it on and smoothed out the air bubbles. Unfortunately, it was not exactly like the older stickers. After a few months the background plastic started to peel off, and I realized you're supposed to press the sticker on and then peel it off immediately, leaving only the letters. By that time the background had become brittle and came off in pieces, along with some of the lettering. I had to scrape it all off with a razor blade and start over.

When I describe these examples to the students in my human-computer interaction class, a common initial response is to wonder how users can fall into such obvious errors. As the students gain more familiarity with the concepts, it is easier for them to recognize categories of such problems in changing an existing system and to apply standard lessons for avoiding them.

• Preserve the usability merits of the existing system, if used standalone or combination with other systems.
• Accommodate usage patterns for the existing system, so that previously learned behavior does not lead to errors.
• Make it easy to learn new functions that are added to the existing system; ensure that new, easily learned or applied functions are actually desirable.

The first and probably most important lesson to learn in improving a system, though, might be cast as a designer's version of the Hippocratic Oath: First, do no harm to usability.