Chapter 5: Evaluating Web Sites
User TestingWhile the evaluation process just described is useful to uncover many types of site problems, it is important not to limit evaluations just to inspection. Developers may focus on certain things and completely miss problems commonly encountered by users. Further, this form of evaluation does not adequately reflect how users actually use a site.
Looking at log files can provide valuable insight into how a site is used. Log files will show who is looking at a site (by IP address or domain name, mostly), what pages users commonly look at, when they look at these pages, the paths users take through a site, the links followed to get to a site, and even what kind of browsers are being used. The log file really does show if content is popular and may provide a great deal of information related to site usability. For example, a tremendous number of users leaving the site from a certain page may indicate a problem. Log files can be used to verify assumptions or even show places to look for problems.
While log files provide a great deal of useful information, they really say very little about a user's feelings about a site. An invaluable way to evaluate a site is to watch how users actually use a site and try to solicit feedback from them. Conducting a user test can be difficult. Be careful to focus more on what users do and not on what the say. Users typically don't want to look stupid and will often indicate that they understand something when they don't.
Rule: Pay attention more to what users do than to what they say.The best way to deal with this problem is not to let users know that they are taking a test; you might even try to casually watch them without their knowledge. If you ask users to take a usability test, you may find that they pay more attention or try harder to figure things out than they might usually. The assumption almost seems to be that test administrators will be pleased at how proficient they are. At the opposite end of spectrum, on occasion testers will purposefully look for errors. In either case, it should be evident that testing conditions may not always be the same as user conditions.
A very important aspect of testing is making sure not to get too involved. For example, if you ask users to evaluate a site, don't guide them through it. If you co-pilot the users' browsing sessions, they will uncover only what you want them to and maybe not use the site as they might normally. If you talk too much, showing off the features of the site, you may not give users a chance to say what they think. User testing can be very difficult for site designers who want to put their work in the best light possible, and they may be very unwilling to listen to user criticisms.
Suggestion: Consider having a person not involved in the site design process conduct a user test.You can certainly be very scientific about user testing: using two-way mirrors, recording mouse travel and keystrokes, and even monitoring pauses or mistakes made by the user during a typical task. Some might go so far as to watch facial expressions or even monitor the blood pressure of the test subject. However, the end result is often really the most important aspect of the test. Remember that, in the final analysis, probably the only real important things to users are whether they were successful in their mission and enjoyed the visit. This does not mean that the study of usability lacks reasonably measurable characteristics; it just suggests that, as imperfect creatures, humans may not always act logically and may even quickly forget the difficulty of performing a task if there is a wonderful reward at the end. Readers interested in understanding more about user testing and usability, particularly the theory and practice of conducting usability tests, should visit http://www.useit.com and http://www.usableweb.com.
Overview | Chapters | Examples | Resources | Buy the Book!