We've heard that a million monkeys at a million keyboards could produce the complete works of Shakespeare – but could they get to the commerce section of your site?

OK, maybe primates aren't your target market, but before you launch your site you had better make sure that your users can use it. Usability doesn't mean that your best friend thinks it's "cool," your designer calls it "cutting-edge," or your VC sees it as "viable." What it means is that you get a group of users to hammer on it, and you watch them hammer. That way, you find out what paths they follow, which graphics they click, where they get lost, and, most importantly, when they lose interest.

Usability testing is crucial to your site's success and profitability. Office supplies giant Staples has seen online sales soar since calling in usability experts Human Factors International to help with the redesign of their site.

"Every time I went in to speak to them after the redesign, they were announcing record-breaking days for sales," says Ross Moser, a senior specialist at Human Factors.

As a result of usability testing, Staples added a shopping cart showing a perennial list of items purchased, the ability to create a shopping list of frequently bought items, and a quick way for shoppers to link to their favorite aisles. After the redesign, drop-off rates during registration fell by 25 percent, and on the "enter your ZIP code" page the drop-off rate fell from 10 to 3.5 percent. "Due to our interaction with Staples, usability testing has become integral to their design process," says Moser. "They understand the importance of knowing their user."

The first step

Those words, "integral to their design process," are the most important part of that statement – you must integrate usability testing before, during, and after you complete your site design.

"Usability goes more than screen-deep. It goes right down to the bone," says Jon Meads, president of Usability Architects, which specializes in Web site testing. "The whole company needs to be geared to the user-development process."

Adam Fischler, usability specialist at Agency.com, agrees: "Early usability testing will save a lot of money later. We try to make testing as fast and cheap as possible. We don't want to tie up development. To be effective, it has to be very easy to do. It has to fit into a schedule and budget."

Early testing lets you examine and use market research to determine the value of each feature and how well people recognize and value each one. "Every feature you add adds cost and increases the amount of testing you will have to do," says Meads.

"During the conceptual phase, there are lots of questions," Fischler concurs. "Do users value what you are trying to build? Is there any sort of knowledge that the user needs before using the site? At this point you don't need feedback on the interface. You need to gather more audience analysis."

Audition the Competition

You can get your best tips for what is right and what is wrong in your field by checking out the competition. "In the early conceptual stage, we might do a competitive evaluation of Web sites and ask testers to complete a task such as buying a computer on three different sites," says Agency.com's Fischler. "We'll look at what functionality was offered on each site. This helps us understand what makes one site more valuable than another."

"You can do different forms of tests from an early stage," says Catherine Gaddy at Human Factors, who holds a Ph.D. in applied experimental psychology and the CHFP (Certified Human Factors Professional) credential. "You can do simple tests even when you just have a list of features. You might have a list of 20 or 30 things. You can ask people to rank features or choose the top ten most important features."

Magnet Interactive has set up a usability style guide that dictates design standards such as measurements for designable space – taking into account different browser default settings and platforms – and page sizes. "It's critical to set up these guidelines," says Lewis Francis, VP of media technologies at Magnet Interactive Group. "A lot of studies show that a third of traffic disappears if users have to wait longer than 15 seconds for a page to load. So we set target sizes for basic page types like splash pages and home pages."

Drawing it out--or not

Paper prototyping is an early, rapid-testing technique that is done before any coding takes place. Simple screen drafts are created with a drawing program, and users are asked to walk through tasks using the paper screens.

"This is where you are trying to knock out early conceptual flaws," says Fischler. Moser agrees: "It's a great and inexpensive way to see how well your design matches up with the user's mental model. Graphics and a more complete look can influence the user. With a more completely designed interface, users are influenced by how you want things to work, instead of us finding out how it would work best for the user."

Magnet Interactive avoids the paper trail, however. "Some folks swear by paper testing. We have never done that," says Francis. "We create mock-ups of our sites. We probably spend more time and budget on creating mock-ups. But we find it more valuable watching people use the site. It's important to sit with the target audience, not with the VP of marketing or the head of the department who hired you to do the job."

Hocus Focus Groups

Task-based evaluations offer the most concrete pointers for rethinking your design. "Look for performance rather than preference when testing," says Gaddy.

"There is value to surveys and focus groups, but the danger [with focus groups] comes when you start interpreting information," says Fischler. "For example, in a focus group, you hold up an interface. One guy says he's not sure if he likes it. People are influenced by this comment. Behind the scenes, everyone runs around changing things. I trust behavior more. It's a stronger indicator of whether an interface is effective or not."

Good usability testing doesn't involve herding hundreds of users to check out your site. "Research shows that with four to six users you can determine 80 percent of any problems," says Fischler. "Pick the fewest number of people possible – the really big flaws, the truck-sized problems, will be obvious. You'll see those in the first four tests." The more diverse the audience you are trying to reach, the more testers you will need. Fischler suggests segmenting your audience into markets and selecting four to six users from each group.

The best test subjects

Should your interface design team be involved in the testing process, or should you hire neutral testers? Each route has value.

"A lot of people doing user interface work are going out and doing the testing themselves," says Meads. "Possibly they are biased, but also they are good." Fischler also shows very early site designs to a designer to ask for his or her expert opinion. "The feedback is valuable, even though you are not asking the end user," he says. But Fischler emphasizes the importance of testing on real users. "Test on people that are not designers or programmers. Test on the actual people you are targeting with your site. Programmers have a very good understanding of code, but then no one else can figure out how to do things," he says.

Test facilitators also have different approaches. "You need to be a cheerleader," says Gaddy. "If the [test users] get bogged down, you move them on to something else. Someone looking at a site in an office won't be as persistent as someone being paid for testing. If you are testing a program like MS Word, you leave [the testers] alone to figure it out. For Web sites, you need to be there with them to make sure they stay focused and stay on the site. It's easier to go off on tangents."

Fischler believes in the laissez-faire testing technique. "We try not to get in the way of the test," he says. "We don't prompt them – we cut them loose and watch them. What you think are moving elements, they might not. We have seen [people running the test] reach past the testers and show them how to get to the next step."

Eagle-eye observations

Observing the testers is a critical part of the testing process. "By observing, you can tell immediately when something is right," says Gaddy. "In a test when you've got the site right, people smile, they start reading, they start doing stuff. But when they look like those hula dolls you see at the backs of cars, you know you have to go back to the drawing board." Keen observation can spot problems that the testers don't mention. "On one site the font was really small," says Gaddy. "Nobody actually mentioned it, but all six people leaned forward and squinted."

Human Factors uses a test lab with a one-way mirror, so their clients also can watch testers. "This is very useful for them. They are often sobered by what they see," says Gaddy. "It's simple: either [the testers] can do things, or they can't."

Don't think you can sit back and relax once your site is launched. "You should also be measuring once you are up and running," Meads says. "Very few Web sites are static. Things are constantly changing – there are new products and new prices. You need to do market research and testing from the other end. You want to find problems before the end user does."

Trouble Areas

These trouble areas surface constantly for testers:

Feedback: Let your users know what has happened and what is going to happen. On an e-commerce site, users need to know whether they've made a purchase or not.

Privacy: Tell users how you will use their data and who is collecting it. People will bow out if your privacy policy is unclear.

Test smart: Don't run meaningless tests because they're in the book. Choose the right tests for your site at the right stage in development.

Think end-user: Most people design for their bosses, their corporations, or themselves. Put real people at the heart of the design process.

Graphics: Don't overuse fonts and colors. Making users wait for graphics fails to consider their requirements.

Plan to test: Think about testing early on throughout the site's life cycle.

Design a style guide: Create a usability guide early on and stick to it.

Vary connections: A lot of people design using the company's LAN. You need to do actual dial-up testing, across different ISPs and modem speeds.

Know your audience: Corporate workers, kids, young women – make sure you're designing for the intended target.

No mystery icons: If your site is light on text, make sure your target audience understands the graphic representations.