Book Review, by Bill Skeet
Web Site Usability: A designer's guide
------------
Jared M. Spool, Principal Investigator,
Tara Scanlon,
Will Schroeder,
Carolyn Snyder,
Terri DeAngelo
User Interface Engineering, 1997
$39.95
------------
I'm summarizing this report because its findings were frequently discussed at the recent KR design meeting. I'm beginning to think this may be the most important book about web design I've ever read. Most surprising are the results of the study that contradict conventional wisdom about design.
The book is available online: http://world.std.com/~uieweb/bookform.htm or call (800) 588-9855.
Scope of the study:
Nine sites were including Edmund's, HP, WebSaver, Travelocity, Inc., C-Net,
Fidelity, Disney. Basically, they watched a whole bunch of people doing informational
searchs--trying to look up specific facts to compare or make judgements. Accomplishment
of thses tasks was the basis for determining "usability." The researchers also
asked users to rate the sites.
------------
The book sites five major implications of the study.
1) Graphic design neither helps nor hurts: Some sites were professionally designed, and did well, others were also graphically intense and scored at the lower end... they could not identify a significant correlation between graphic design elements and users' success. Edmund's, the top-scoring site is mostly text.
2) Text links are vital; Many users used text links first, ignoring nearby graphics. The predictiveness of the link is important; the better users could predict where a link would lead, the more successfull they were in finding info. The text link is the way users prefer to navigate sites.
3) Navigation and content are inseparable; the "shell strategy" of building a navigational structure and hierarchy and then plugging content into it often leads to generic links and static pages. The sites that were most successful were those where content and navigation were inextricably linked. The problem with shells is that their generic links make it harder for users to predict what they will find. This implication makes development of large web sites significantly more difficult, because it suggests that the home page and high-level links may need to change more frequently
4) Information retrieval is different than surfing; When users surf, they are browsing and click on the most "interesting" looking things; this study focused on one specific activity: information retrieval within a large site. When looking for information, users are more focused. They tend to click on the link most likely to yield the information they're hunting for. Advertisements and especially animations were seen as visual "noise" and irritated some users so much they covered them up. Implies that sites aimed at information retrieval need to be designed differently from sites aimed at surfing--conflicting goals.
5) Web sites aren't like software. Usually, when comparitive testing software, users are asked which product they like most after they've worked with them all. Almost without fail, users choose the same product that they were most successful with-- user preference is a good proxy for measuring usability. Not true for web sites, however. While some people "liked best" the site they were most successful with, others did not. These users like a site because of it's content, rather than the site's ability to help them find information. They'd say things like "I liked Disney, it seemed more interesting," even if they had gotten completely lost and failed to complete any of the tasks.
------------
Chapt 2: Navigation
Users don't form mental models of sites: found that the structure of
the site made little difference whether users would be successful.
Frames: didn't see as many problems as expected. Scrolling can cause probs.
Sites with navigation buttons or links at the top and bottom of pages did slightly better than sites with navigation buttons down the side of the page.
Site maps: The users who used Fidelity's site map were twice as successful at finding answers on that site as the users who did not.
About one-third of users always tried to answer questions by going to the site's search facility.
Many peole used the browser Back button to go back one or two pages. If something was further back, most users would try to use the page navigation.
------------
Chapt 3: Links
The success of a link depends on 1) how well the user can predict where the
link will lead, 2) how well the user can differentiate one link from other,
nearby links. Also, link layout and where links lead can affect usability.
Descriptiveness aids prediction: Links that explicitly describe the content of the page they lead to (e.g., "How to Read the Pricing and Rating Listings"). Generic links caused problems on Disney and C|Net site.
Image links: A weak positive correlation exists between image links (such as graphics that look like buttons) and user success. However, because they don't change color after the user clicks them, there is no visual cue that they have already been traversed. Users did use visual cues to identify image links by moving the mouse across the page and watching for the cursor to change to the pointing finger.
Embedded links: putting text around links doesn't seem to work well. A strong negative correlation between embedded links and user success in finding info.
Wrapped links: anything that causes a link to wrap across multiple lines can damage it's effectiveness. Users didn't always know a wrapped link was one link, not two.
Link destinations: Observations imply that users may have a mental model that links will take them to another page within the same site. Within-page links were slightly positive, but sometimes caused confusion. Links to other sites often left users puzzled by disappearance of the link structure and appearance of the initial site and inability to get back to the home page.
Will experience help? Over time, as users become familiar with the content, will they learn to make accurate predictions based on terse, lower-content links? Since the study only tested first-time use of the sites rather than repeated use, don't know how true this may be. However, they saw little evidence of learning during the tests-- once users got stuck on something, it kept giving them trouble.
------------
Chapt 4: Searching
Users were often confused about what parts of a site the search engine would
examine (scope of search). They tended to assume that a search would cover the
entire site. If there were obviously several different ways to search, users
didn't know which one to try first.
Search results didn't like seemingly unorganized search results. Search results that did not return enough information made information retrieval difficult. Users didn't like it when a search engine returned redundant information.
------------
Chapt 6: Readability and Page Layout
Readability was measured by indexes that average number of words, sentences,
and words with more than three syllables. Theorized that the easier a page is
to read, the more usable it would be; however, found exactly the opposite.
The less readable a site was, the more users were successful with the site. The less readable a site was the more users found the site authoritative, clear, complete, satisfying and useful.
Researchers note that most people didn't read all of the information on a page. They scanned the text as they searched for an answer. Only when they found something that looked plausible did they actually read the text. Page layouts that made skimming easy fared better than those that didn't.
Another theory: hard-to-read sites have fewer conjunctions and standard grammatical structures. Makes sense that the extra words we normally use to complete a sentence may get in the way of the meaningful words in the text. Removing them, and thus making the text harder to read by traditional measures, may increase scanability.
White-space: Surprisingly, the more white-space on a site, the less successful users were at finding information. The more white-space, the lower users rated the site in terms of finding things easily, ease of reading, ease of searching, overall appearance, ease of use and productivity.
Theory why sites with only a little white space do better than sites with a lot: When people are skimming/hunting for information, they benefit by covering a lot of ground quickly. White space spreads out the info and slows the hunt.
Also, looking at text on a web page is different from looking at text in a newspaper, for example. In a newspaper, the reader can see the whole page at once. This is not true for web sites, which are viewed through the window of a browser. The reader is forced to construct the page in his or her head.
Scrolling and "the Fold": Never saw any user frustration with scrolling. When counting "first clicks"--the first place people clicked when they came to a new site -- clicks were just as likely to be above the fold as they were to be below it.
Rules: some users had trouble with horizontal rules that fell at the "fold". Repeatedly, users did not scroll below these lines, even though they did scroll down long pages.
------------
Chapt 7: Graphic design on the web
Two graphic issues: 1) in the sites studied, download time was not an issue,
2) users could not concentrate when there was animation on the screen.
Quality of graphic design seems to have little impact on users' ability to find and process info. Measured more than a dozen factors (number of graphics, size of graphics, colors of links, background color, presence of advertising, number of graphic links, etc) and found that sites with any of these graphic attributes did as well or as poorly as sites without them.
No correlation, positive or negative, between graphic design elements and the users' success at finding info.
Only one graphic attribute strongly tied to user success is browser-default link color. Use of default colors (blue, unvisited/ purple, visited) is helpful so users don't have to relearn every time they go to new site.
Download time: When users thought an image would have interesting content, they would wait for it to load; not patient for decorative stuff. Users skimmed text and alternate text as pages loaded. If they found an interesting link, they jumped before the graphics had a chance to load. This caused problems for some users when the images provided valuable content and they didn't wait.
Movement in ads: None of the sites tested actually used animation to support the content--the animation was gratuitous. Users find this animation uniformly irritating. Animation makes it considerably harder for users to read or skim. Advertisements generally contained the most movement, but users masked out the ads when searching for content.
------------
Chapt 8: User preference
Asking users if they like a site is not a good indication of whether they can
successfully use it. One of the least liked sites was one that users were fairly
successful with. When asked why, users uniformly said they liked sites the best
that had content that was interesting and relevant to them. Sites users liked
were often different from the ones they could successfully use.
Results are disturbing because they imply that designing a site that users like and designing a site that they can use may be conflicting goals. On the other hand, these results may mean simply that good content is so important to users that other factors are secondary. In any case, we cannot depend on user satisfaction to indicate a usable web site.
------------
BUY THE BOOK.
Bill