Command Links

Jakob Nielsen's Alertbox for May 14 is now online at:
  > http://www.useit.com/alertbox/command-links.html
		
Summary:
Application commands can be presented as buttons or as links,
which offer more room for explanation.
For primary commands, however, buttons are still best.
		
----------------------------------

Usability Week 2007 conference

  > San Francisco, June 18-23

In-depth training:
  > 3-day Intensive Camp: Usability in Practice
  > 3-day Immersion: Tog on Interaction Design

10 specialized full-day tutorials:
  > Fundamental Guidelines for Web Usability
  > Application Usability (2 days)
  > Intranet Usability (2 days)
  > B2B Sites
  > Websites That Sell
  > Presenting Company Information on Corporate Websites
  > E-Mail Newsletter Usability
  > Writing for the Web

  > http://www.nngroup.com/events

DEADLINE to save 10% with early registration: this Friday, May 18

----------------------------------

BBC reports from our London conference last week

"Web 2.0 distracts good design":
   > http://news.bbc.co.uk/1/hi/technology/6653119.stm

----------------------------------

DESIGN OF WEB SATISFACTION QUESTIONNAIRES

I caution against relying too much on survey responses for your design
decisions. What people say and what they do are often different. Still,
there's value in asking users questions to complement true observational
data. Particularly for assessing the users' level of satisfaction.

Two academics (Paul Van Schaik, University of Teesside, and Jonathan Ling,
Keele University) have researched the mechanics of presenting survey
questions on websites. Their findings:

(1) When people are asked how much they agree with a statement, the
following two choices work equally well for presenting the response
options:
   (a) linear slider: a visual analogue scale
   (b) list of numbers (usually 1-7): a Likert scale
This is good, because it's much easier to deal with numbers. Also, 82% of
the test participants preferred the Likert response format. Thus, use a
list of numbers.

(2) It is better to present one question at a time across multiple screens
than to show the entire survey on a single page. With only one question
visible, users can focus on that without distractions.

In this case, I DISAGREE with the study findings. The study used a
questionnaire with 23 questions, which is much too many for commercial
projects, where it's essential to MAXIMIZE the response rate by MINIMIZING
the burden you impose on customers. Yes, with a huge number of questions,
showing them all at once can be overwhelming. But the answer is to reduce
the number of questions. (Then run another survey another day, if
necessary – after all, any decent website has plenty of traffic, so there
are always more users available.)

63% of users preferred having all the questions on a single screen, as
opposed to having to page through the questions. Since practical Web
surveys rely on *voluntary* compliance, it's important to make them as
pleasant as possible and to follow users' preferences. (This is in
contrast to the research study, where participants were *required* to
complete the survey.)

(3) It's better to show people their 1-7 rating choices as a list of radio
buttons than as a drop-down menu. This, of course, confirms the usability
finding we have had for 7 years that it's usually best to avoid
drop-downs:
   > http://www.useit.com/alertbox/20001112.html

In this study, it took 16% more time for users to use a drop-down than to
use radio buttons.
 
Full scientific paper (paid access required):
   > http://doi.acm.org/10.1145/1229855.1229859

The biggest usability problem in questionnaires on real websites is that
they are inevitably TOO LONG. This again leads to low response rates,
which invalidates the entire exercise. See:
   > http://www.useit.com/alertbox/20040202.html

Survey bloat is a much bigger problem than whether the responses are shown
as radio buttons or drop-downs. But you might as well get the details
right, of course.

Also remember to use the correct control when deciding between radio
buttons and checkboxes (needed for some forms of questions):
   > http://www.useit.com/alertbox/20040927.html

To conclude: use Likert scales with radio buttons. (And I would add: keep
surveys short and fit them on a single, non-scrolling screen.)

----------------------------------

STUDENT VOLUNTEERS NEEDED

We need students to help staff our conference in San Francisco in June
(see above). You get to attend the world's most elite usability event for
free in return for pretty light duties.

To apply, send a short resume plus the name and email address of your main
professor to [email protected]