UI Design Newsletter – June, 2001

It is the nature of most engineering activities that designers never have enough time to truly perfect their products. Engineers are continuously confronted with the real-world constraints of tight schedules and limited resources. Designing Web sites is no exception.

Most Web site designers would like to develop sites that could be easily used by anyone who attempted to access their site. This is hardly ever possible. In the end, designers must wisely use the limited resources available to them to design Web sites that will accommodate the most users with the least amount of design and testing effort. The following example uses only four variables (out of many) to show how designers can effectively limit their many options. The data used in the example were recently reported by www.thecounter.com/stats/.

In each case the goal was to include enough options to handle at least 90% of users. To accommodate the largest number of users with the least investment of time and other resources, restrict design decisions to the following:

1. Design only for the Windows operating systems from Microsoft (these account for 93% of all operating systems being used):
   • Win 98 (72%),
   • Win 95 (10%),
   • Win 2000 (6%), and
   • Win NT (5%).
2. Design only for Internet Explorer 5 (77%), Internet Explorer 4 (9%) and Netscape 4 (9%). These three browsers account for 95% of all browsers being used.
3. Design primarily for 800x600 pixel resolution (53%), but also accommodate pixel resolutions of:
   • 1024x768 (31%),
   • 640x480 (5%), and
   • 1280x1024 (3%).

   These resolutions are used on 92% of all computers.

4. Design primarily for a 16-bit color depth (55%), but also accommodate color depths of
   • 24-bit and 32-bit (38%), and
   • 8-bit (5%).

These color depths are used on 98% of all computers.

Even focusing only on these four variables, designers would need to conduct 108 separate tests to determine how the interfaces will perform with the different combinations. For example:

• Test 1 - Win 98 and Internet Explorer 5 and 800x600 resolution and 16-bit color
• Test 2 - Win 98 and Internet Explorer 4 and 800x600 resolution and 16-bit color
• Test 3 - Win 98 and Netscape 4 and 800x600 resolution and 16-bit color
• Test 4 - Win 95 and Internet Explorer 5 and 800x600 resolution and 16-bit color
• Test 5 - Win 95 and Internet Explorer 4 and 800x600 resolution and 16-bit color
• Test 6 - Win 95 and Netscape 4 and 800x600 resolution and 16-bit color etc.

In other words, unless there is a very good reason to do so, do not waste time designing for and conducting usability tests on:

• The Macintosh, WebTV, Linux, Unix operating systems, or any Windows operating system prior to Win 95,
• Any browser except recent versions of Internet Explorer and Netscape,
• Monitors that have resolutions less than 640x480 or greater than 1280x1024, and
• Monitors having less than 8-bit color.

As designers, we must clearly understand our intended user population, design specifically for them, and learn to systematically give up trying to accommodate all users. Our design time is finite and our resources are limited. We cannot design a usable website for everybody! Good designers make good decisions about what not to do.

Bob points out how many different technical variations there are in Web design. Indeed, it is impossible to test all their interactions. But most of these variations can be handled by understanding the range of technologies and checking the design in different resolutions and browsers. The tough news is that the USER community has far greater variation then the technical infrastructure. Users vary on functional requirements, taskflow, skills, knowledge, motivation, and environment. Now think of the interactions for that!!!

Luckily there are also a set of understandings and general principles that we can apply to handle the range of users. These let us accommodate a wide range of users based on human psychophysics. THEN, there is a methodology that custom fits the interface to the target user/taskflow/environment. Without these research-based principles and methodologies your design has a negligible chance of full success. Just think of the number of interactions!