Usability Guidelines for Web Sites

Research Methods and Usability
Guidelines for Ecommerce Web
Sites
Mary Czerwinski
Microsoft Research
Note: Many of these slides came from a Keynote address by Kirsten Risden at
Web99
Key ecommerce ingredients

Most sources agreed that
navigation
 clear content organization
 product information availability
 trust


Extras:
personalized services and/or
 providing updated content or activities.

Building trust






Building Trust: What It Takes
A clearly stated return policy
A clearly stated security/encryption policy and
seals of approval from authoritative
organizations
The ability to back out of a transaction
Fast and easy navigation
Source: eCommerce Trust Study, Cheskin
Research and Studio Archetype/Sapient,
January 1999
From browsers to buyers….

Seth Gordon recommends:





Make Buyers Feel Comfortable: good content and
a personality
Get Users to the Goods: make the products
easy to find and educate along the way
Make It Easy to Complete a Purchase: no
roadblocks in your checkout tunnel
Build Buyer Loyalty: deliver on your promise and
give reasons to return
From Builder.com Design section, January
25, 2000.
Serco ecommerce guidelines




Make it easy for users to enter the store
Use a meaningful store layout and
product categories
Ensure descriptive terms or pictures
are used
Allow users to find and use search
facilities
Serco ecommerce guidelines





Provide meaningful and relevant search
results
Make it clear whether items are available in
the on-line store
Allow users to see what's in their shopping
basket
Provide sufficient product information and
explain technical terms
Clearly flag the financial security features of
the site
How do we get there?
Information Structure



Site organization should reflect user’s
conceptual categorization of content
Can be very different from designer’s
categorization of content
Derived empirically

Card Sort


Need < ~200 distinct information types
Find Task (a.k.a. reverse card sort)

If > ~200 distinct information types
Card Sort Study
Cluster Analysis
Find Study (a.k.a. Reverse Cart Sort)




If more than 200 items, make initial
design prototype….(s)
Record user paths through browse
hierarchy of ecommerce site
Map out “confusability”
Determine where problems are and why
Find Study
us
in
es
sFi
na
H
ea
nc
lth
e
C
om
-F
itn
pu
es
te
E
s
nt
rs
er
-In
ta
te
in
rn
m
et
e
S
nt
po
r ts -Me
di
-R
a
ec
re
Tr
at
R
io
a
ef
ve
n
er
l-L
en
ce eisu
-E
re
du
c
H
o m atio
n
eN
Fa
ew
m
sP
ily
In
eo
f
o
pl
rm
eat
C
om
io
In
n
m
te
un
re
st
iti
ses
Li
fe
st
yl
es
B
% redundancy
“Reckless redundancy”
reduces learnability
100
90
80
70
60
50
Redundancy
40
30
20
10
0
Category
Information structure
take-aways

For ease in finding information, ecommerce
site organization must


fit user’s conceptual categorization
be easy to learn
Presentation

Support and guide “scan and focus”
visual processing through design



Layout and graphical design
Labels
Text
Layout and Graphical Design







Use mostly text and avoid pictures
Provide landmarks
Use simple configurations
Use “blocking” appropriately
Don’t break configurations up
Make high contrast choices
Avoid animations
Use Text and Avoid Pictures
At least . . .keep navigation
elements away from graphics
Use clean, simple
configurations
Use “blocking” appropriately
Use High Contrast Colors
Click Here
Avoid animations….
Labels




Use distinctive, differentiating terms
Concrete terms are more “attractive” than
abstract terms
Put supporting detail in tool tips or “look
ahead” text
Asking users to generate labels is highly
instructive
Use tool tips or “look ahead”
text
User generated labels



For groupings in card sort
For content
For subordinate categories
Delivery

Performance



First page download should be < 10
seconds on 28.8.
Secondary pages can be somewhat slower
to download.
Resolution


60% of Media Metrix sample using 6X4
Number is changing slowly.
Key Take-aways




Match user’s conception of how
ecommerce domain is organized.
Remember that you are supporting a scan
and focus approach to visual information
processing.
Words can be worth thousands of pictures.
There is a high bar with little room to error
when it comes to performance.