InformationWeek.com

Welcome Guest. | Log In| Register | Membership Benefits

April 2, 2001

RADICAL SIMPLICITY
Remember The User's Point Of View

continued...page 2 of 2

Illustration by Nicholas Wilton

ONLINE EXCLUSIVE CONTENT:

Simple Twist Of Fate
Simplify That E-Mail

RADICAL SIMPLICITY SPECIAL:

Radical Simplicity Home Page
InformationWeek Research: Overcoming IT Complexity
InformationWeek Events: The Spring 2001 InformationWeek Conference

When Carly Fiorina became HP's CEO in September 1999, HP worked to reflect Fiorina's "Invent" ad campaign on the site by making its features more innovative. Last year, Baldwin's team built in smarter routing of tech-support E-mail from customers, making sure requests are sent to the appropriate product specialists. On the horizon are efforts to better localize HP.com for foreign markets and to make the site accessible from information appliances as well as PCs. "I'm not going to claim we've achieved the goals of an inventive site yet," Baldwin says, "but we're headed down that path."

Site redesigns and user testing can be expensive. The Nielsen-Norman group, which claims former Sun Microsystems engineer and usability sage Jakob Nielsen as its other namesake, charges as much as $20,000 a day for its consulting services. Ed Chi, a research scientist at Xerox PARC who studies how computer users find information, says usability testing involving dozens of users generally costs companies $35,000 to $60,000--sometimes much more.

According to Norman, gaining expertise in usability techniques can pay off later. Norman and Spool advocate building prototypes--special prerelease "versions" of software designed to gather information about how the applications are used--as early and often as possible, and incorporating findings into the design process. They also advocate paper prototypes, the approach that worked for Metavante.

But that's not how commercial software engineers usually design products. The standard process involves drawing up a specification that details the functions a program should include and what the commands ought to be, then writing the program to the specs. Usability testing starts about three-quarters of the way through the design process, when the software has taken form. By then, it's often too late to incorporate the results into the current release.

While Norman says his approach saves companies time and money later, some contend that the constant tweaking costs more up front. "Rapid prototyping is definitely a great ideal, and you can probably get more usable products that way," says Jack Breese, an assistant director at Microsoft Research. This isn't always possible, Breese adds, because resources aren't limitless--even at a company Microsoft's size. Still, things are changing. Sullivan says that while feedback from field testing often hasn't made it into products until later releases, "we don't think that's true anymore." An example: About 20% of Windows usability engineers are working on a future version of Windows XP while the current version is still in beta testing.

Photo by David Turek One business that has followed the Nielsen-Norman approach with good results is Google Inc., the privately held search-engine company that has won numerous awards for its accurate results, minimalist interface, and lightning-fast performance. "There's a direct correlation between a good user interface, good results, and people coming back [to the site]," Google director of technology Craig Silverstein says.

That's important for a company that makes its money selling ads on its site (no banners, just fast-loading text) and licensing its search technology to sites such as Cisco Systems, Netscape, and Yahoo. Nielsen consults for Google and sits on its technical advisory council--and his influence shows. Google's interface is so clean that its fast-loading home page takes up just 4 Kbytes of space.

The key to usable Web software, in Google's view, lies in thinking about the front and back ends of the application in concert. Once a month, Google crawls the Web and downloads more than 1 billion pages to its 8,000 Linux-on-Intel systems, which are spread among five U.S. data centers. When a user sends in a query, Google's algorithm refines results that match the query with a proprietary PageRank score, which measures how many other Web sites link to a pertinent page and gives more weight to links on sites that are considered "important" in their own right.

Linux is another key to Google's speed: The company's IT team optimizes the open-source-code operating system to read data from disk quickly. Google has a Linux project under way that aims to read data from disk more directly.

Google also spends lots of time tuning its HTML. "We regard tables as purely evil," says Marissa Mayer, a Google software engineer, referring to HTML-coded lists of information that have a common look on Web pages displayed on various platforms, but which take precious seconds to load. "We've pretty much given up controlling what things are going to look like on every computer." Every page that's written in Google's IT shop is run on a variety of operating systems, browsers, and monitors. Silverstein's personal copy of Netscape Navigator version 1 sometimes gets roped into testing. He keeps it around because it's fast.

George Moromisato Photo by Mark Ostow Building user-pleasing software is tougher when customers aren't familiar with an application. It's a conundrum that Groove Networks Inc. faces as it prepares to launch version 1 of its product, a peer-to-peer communication and file-sharing app designed by Lotus Notes creator Ray Ozzie. Groove's product aims to make it easier for remote workers to collaborate on documents, trade comments in a shared area, and even draw on a virtual whiteboard, but it's unfamiliar to most people. The way Groove handles data is fundamentally different from the way client-server apps deal with it. The data is stored locally, traded peer-to-peer when people are online, and synchronized when they log on after an offline session.

Chief product designer George Moromisato says Groove's biggest challenge has been to introduce new concepts to users while grounding them in things they already know. "It's very easy [for people] to screw up a process they don't understand," he says. Groove "glommed on to" familiar concepts such as chat rooms for its shared-space interface, and tried to expose users to as few configuration options as possible to reduce complexity, he says.

"People are pretty smart," adds Moromisato, who has also done design work on Notes at Lotus. "We just have to give them clues here and there." That's advice IT shops can take to heart.