[thelist] Usability: Empirically Validated Web Page Design Metrics

[Marlene Bruce]

Hey all,

I went to a BAY-CHI lecture last night, and the presentation (by 
Marti Hearst) was on "WebTANGO - Automating Web Site Usability 
Evaluation" (Empirically Validated Web Page Design Metrics). While 
the tool is not yet finished (the research is already extensive--in 
it's third round--even though there are some significant issues they 
still have to work out), some of the preliminary results they found 
was that the more usable sites have the following (I should stress 
they're preliminary):

* Smaller minimum color usage (like two main colors and one accent)
* No italics
* Clustered links
* There were "good" ads on good sites, "bad" ones on bad (reputable 
ads correlated with respectable sites)
* More interactive objects (like search buttons)
* The most Bobby and WebLint errors! (this correlates to the number 
of interactive objects, tables, etc.)

These results were from examining 5346 pages on 333 sites (in 6 
content categories) against 157 quantitative measures. The sites they 
used were ones being judged by Webby Award judges, and the results of 
the WebTANGO automated evaluation (good, average, and poor sites 
and/or pages) were correlated against the results of the Webby 
judges. The average correlation accuracy ranged from 70.70% to 82.75%.

I think this stuff is fabulous, if only because they are results from 
studies, rather than opinions bourne from supposition or ego (when I 
spoke to Marti afterwards and mentioned Nielsen, she responded by 
indicating that his recommendations used to result from empirical 
studies, but are much less so now that he's busy giving lectures ... 
and further that at least some of his recommendations aren't 
supported by the WebTANGO studies, hmm...). Historically I've 
respected much of Nielsen's work (though I've found "Designing Web 
Usability" to be helpful and informative, I was disappointed that 
there were very little academics indicated to back his 
recommendations, and no bibliography ... but I'm *not* trying to make 
this a Nielsen good vs. bad discussion).

Anyway, if you care to learn more: http://webtango.berkeley.edu/.

Cheers,
Marlene