Discount Usability for the Web

by Jakob Nielsen (1997)

The introduction of the spreadsheet turned millions of people into programmers without the benefit of a computer science degree. Because of the resulting lack of knowledge about even the simplest debugging techniques, spreadsheet formulae and macros are riddled with bugs and million-dollar business decisions are sometimes based on calculation errors. It has been estimated that at least 40 percent of spreadsheets have bugs.

The introduction of the Web is causing a similar phenomenon in user interface design. My current estimate is that there will be about 10 billion Web pages on the Internet by the Year 2001. Intranets and extranets will probably hold at least 10 times that many pages. We already have two million pages on SunWeb (the intranet at Sun Microsystems).

Each Web page is a user interface design problem equivalent to that of a dialogue box: you must design a task flow that brings the most important items to users' attention and design alternative options for them to click on -- all the while keeping the meaning of these options clear for novice users. Considering that the world will design more than a 100 billion of these dialog-box equivalents in the next three or four years, extremely simple and inexpensive usability methods are crucial if we are to avoid a usability meltdown on the Web.

Amateur Designers

Most Web pages are designed by amateurs. Even at companies that believe in usable Web design, only the top few pages ever receive the attention of user interface professionals or the benefit of traditional user testing in the laboratory. The vast majority of pages are designed by marketing staff or others with little experience in interaction design or usability methods. Clearly, it is not an option to require that every Web page be designed by professionals or go through full-fledged traditional usability studies. Doing so would be equivalent to demanding that only people who can recite the collected works of Dijkstra in their sleep be allowed to build a spreadsheet. In the real world, such demands go nowhere: people want to develop their own spreadsheet analyses now, not wait months for the MIS pros to get around to helping them. Similarly, people want to build their own Web pages in their own departments without having to clear the results with the "design police" at corporate headquarters.

Inadequate use of usability engineering methods in software development projects have been estimated to cost the US economy about $30 billion per year in lost productivity (see Tom Landauer's excellent book The Trouble with Computers). By my estimates, bad intranet Web design will cost $50-100 billion per year in lost employee productivity in 2001 ($50B is the conservative estimate; $100B is the median estimate; you don't want to hear the worst-case estimate!). Bad design on the open Internet will cost a few billion more, though much of this loss may not show up in gross national products, since it will happen during users' time away from the office.

A usability loss of $100 billion may sound like a lot, but considering that in 2001 there will probably be about 200 million people designing intranet pages, each designer's work will contribute only $500 of that usability loss -- not nearly enough to justify the costs of hiring professional designers or paying for advanced usability work. Discount usability engineering is our only hope. We must evangelize methods simple enough that departments can do their own usability work, fast enough that people will take the time, and cheap enough that it's still worth doing. The methods that can accomplish this are simplified user testing with one or two users per design and heuristic evaluation.

Teach Usability in Schools

User testing and heuristic evaluation should be taught as part of the standard elementary school curriculum The proper role of Internet technology in schools is not the completely naive ideal that politicians have proposed: kids sending e-mail to the world's best scientists asking for help with their homework. No one would stay a leading expert long if they spent their time answering a flood of messages: personally, I can't even keep up with unsolicited mail from PhD-level students. Much more productive is to have kids use the Internet to build their own hypertext information spaces as part of their course work. It is much more inspiring to write for others than for the teacher's red pencil. They should usability-test their pages with one or two other students and conduct critical usability inspections of different designs for each problem.

The only way we can hope to teach usability engineering from the third grade up is to teach discount usability methods. Advanced and sophisticated methods that require, say, an understanding of statistics are clearly impractical.

Some critics will no doubt say that it is unacceptable to teach kids less than perfect methodology or to test interface designs with only two users or to do any of the other things I recommend here. I agree that these methods are imperfect. But the only realistic alternative is to do nothing. Given the amount of usability work we'll need in the coming years, it is quite simply not possible to do it all with deluxe methodology.

Just Do It

The true choice is not between discount and deluxe usability engineering. If that were the choice, I would agree that the deluxe approach would bring better results. The true choice, however, is between doing something and doing nothing. Perfection is not an option. My choice is to do something!