“Give Customers Short Paths to What They Want” by Moira Dorsey
(aka the “three clicks” (or the “well, duh”) paper)
Forrester found that on a site with a flat navigation structure, where important content is three or fewer clicks from the home page, users get where they want to be faster and with fewer errors. This has been a standard practice in web design since the last millennium, right? Obviously it's much easier to design this way if your site has a limited number of purposes (e.g., if all you do is sell widgets or allow users to post pictures and text). One important point: if your site is content-rich in a way that means that not all your content is surfaceable in two or three clicks, your search box needs to be obvious and your search engine good. It took me six clicks to get to the MCDM website from the UW homepage by selecting links, but only two when I searched on “MCDM.”
“Web Page Layout” by James Kalbach and Tim Bosenick
The Kalbach and Bosenick paper was interesting (and not just because it used RH nav for its own table of contents). At first I had some difficulty parsing the two hypotheses, but finally understood them thus:
LH navigation will be faster than RH overall, BUT
RH navigation will be initially slower, but will speed up as the tasks progress until there is essentially no speed difference between RH and LH navigation
From looking at their chart of results, it appears at first glance that these two somewhat contradictory hypotheses were borne out in their experiment, but Kalbach and Bosenick claim that there is in fact no difference between the two navigation schemes. The difference for Task 1 and Task 2 looks significant until you look at the standard deviations—and you have to give those more than a cursory look, because it turns out that nearly a third of the participants were unable to find an e-mail link (task 1) and a physical address (task 2) listed on the site within the three-minute limit. Users who were unable to complete a task in three minutes were not counted as data points; this cuts the number of points used for these calculations down sharply. Three minutes is roughly five sigma from the mean, and it seems bizarre for so many users to have ended up so far outside of that limit. It's impossible to know whether this stemmed from a design problem on the site or a problem with the way the study was conducted (I'm guessing site problem, though).
I find it hard to understand why they used such a small study population in the first place. When they remove the slow people, they're left with ~20 data points, which would make me nervous if it were my study. I'm not a statistician, have never played one on TV, but have taken a couple of statistics classes, which means I know just enough to make me want to know more about the sampling. The findings themselves seem reasonable, however.
The discussion section brings up a number of interesting points regarding how people scan a site. The Audi site is extremely spare, with a short right-hand menu, black-and-white graphics, and plenty of white space, which may well have contributed to the lack of difference in the outcomes. It would be very interesting to conduct a similar but more complex study with longer navigation menus. I think the main point here should really be that if you stick to good design principles throughout your site, it doesn't matter so much what your navigation scheme is as long as it is consistent and clean.
“Evolution of Web Site Design Patterns”
This is a very long, data-heavy paper, and difficult to treat critically in detail. Basically, the authors examined 1500 sites over the period 2000-2003 for 157 interface factors in order to determine whether website design has evolved towards better usability and accessibility. Examples of some of the these factors are number of links per page, average number of words included in a text link, font style and size, and number of HTML errors per page.
One problem the authors note is that much of the enormous amount of web design guidance available is confusing and contradictory. They grouped sites according to their rating by Webby Award judges in order to assess practices. Interestingly, after examining a number of highly-rated sites, the authors recommended that the Webby committee adopt a more rigorous evaluation procedure.
Overall, for the period 2000-2003, the authors found some improving trends, some disappointing trends, and a fairly shocking number of accessibility problems. One thing that's not made clear in the paper is whether the increase in HTML errors and the decrease in accessibility is due to the increase in the number of HTML-inexperienced web authors, using web interface tools that may not have been well designed.
