USEFul: A Framework to Mainstream Web Site Usability through Automated Evaluation

Alexiei Dingli & Justin Mifsud
International Journal of Human Computer Interaction (IJHCI), Volume (2) : Issue (1) : 2011 10
USEFul: A Framework to Mainstream Web Site Usability
Through Automated Evaluation
Alexiei Dingli alexiei.dingli@um.edu.mt
Intelligent Computer Systems
University of Malta
Msida, MSD 2080, Malta
Justin Mifsud justinmifsud@gmail.com
Computing & Information Systems
Goldsmiths University of London
London, SE14 6NW, England, UK
Abstract
A paradox has been observed whereby web site usability is proven to be an essential element in
a web site, yet at the same time there exist an abundance of web pages with poor usability. This
discrepancy is the result of limitations that are currently preventing web developers in the
commercial sector from producing usable web sites. In this paper we propose a framework whose
objective is to alleviate this problem by automating certain aspects of the usability evaluation
process. Mainstreaming comes as a result of automation, therefore enabling a non-expert in the
field of usability to conduct the evaluation. This results in reducing the costs associated with such
evaluation. Additionally, the framework allows the flexibility of adding, modifying or deleting
guidelines without altering the code that references them since the guidelines and the code are
two separate components. A comparison of the evaluation results carried out using the
framework against published evaluations of web sites carried out by web site usability
professionals reveals that the framework is able to automatically identify the majority of usability
violations. Due to the consistency with which it evaluates, it identified additional guideline-related
violations that were not identified by the human evaluators.
Keywords: Usability, Automated Usability Evaluation, Usability Guidelines, Usability Problems.
1. INTRODUCTION
The International Standards Organization's ISO9241 standard, defines usability as the
"effectiveness, efficiency and satisfaction with which specified users achieve specified goals in
particular environments" [1].
Whilst there exists a general agreement about the importance of web site usability, especially
within the technical communication professional and the academic communities [2, 3, 4], it is
given less priority in the commercial sector [3]. In fact, even in its early years, it was noted that
on average, the web sites on the World Wide Web were of a poor quality [5, 6, 7, 8].
Problems related to poor usability and accessibility in software and web sites, also prompted
some countries to also have their own guidelines and legislation for usability and accessibility of
web sites [9, 10]. Additionally, there is less accord about what constitutes usability [2] particularly
because some argue that usability is perceived in different ways by different users based on their
characteristics such as age, gender, education level, technology skills and culture [11, 12].
Thus, it can be observed that a paradox exists whereby web site usability is proven to be an
essential element in a web site, the absence of which confuses users and results in loss of
revenue [13, 14] and at the same time it is not commonly applied with the commercial sector. The
main question that this study aims to address is “How can web site usability be automated and as

a result mainstreamed?”- meaning how can an automated tool be developed that can make it
easier and more possible for more web designers and developers to produce usable web sites.
2. BACKGROUND
2.1 Current limitations that Prevent Web Site Usability from Going Mainstream
Usability Evaluation (UE) is the process of measuring usability and recognizing explicit usability
problems [15]. Its main goal is to identify the main issues in the user interface that may lead to
human error, terminate the user interaction with the system and cause user frustration [16].
Although there exist a number of widely accepted usability evaluation techniques such as
Heuristic Evaluation [17], Cognitive Walkthrough [18], Think aloud testing and Query techniques
[15, 19], the development of usable web sites is not common because of the following limitations:
• Usability evaluation requires the engagement of experts to conduct it, and there is a
shortage of such experts [20, 21, 13, 22].
• The process of conducting usability evaluation is expensive and some companies do not
have the finance to afford it [23, 24].
• Conducting usability evaluation and improvement of web sites is becoming increasingly
difficult because of the number of web sites being developed, their size and the regularity
at which they are updated [25].
• Time is an issue since the web site life cycle is fast due to market pressure and absence
of distribution barriers [5]. So as to meet such demanding deadlines, evaluation many be
overlooked, thus resulting in less usable web sites.
• Tobar et al. [24] state that all forms of quality measurement of a web site such as
usability evaluation can only be carried out up to a limited depth.
• Studies also show inconsistencies in the reported usability violations when the same
web sites were evaluated by different usability experts [26, 20, 27, 28].
2.2 How Automation Helps in Mainstreaming Web Site Usability
In this study, automation is being chosen as the primary method to mainstream web site usability.
This is because our proposed framework is based on research carried out by Beirekdar et al. [20],
Ivory and Hearst [29] and Brajnik [5] who identify automated usability evaluation as a viable
approach that can overcome the limitations of its manual counterpart. Since most of the
advantages of automating web site usability that they propose actually overcome the current
limitations, outlined in Section 2.1 of this document, then automation has been chosen as the
technique to mainstream web site usability. This is because they state that automated web site
evaluation:
• Reduces the costs of usability evaluation: Through automation, the evaluation can be
done more quickly and hence more cheaply.
• Reduces or eliminates the need for usability experts to carry out the usability
evaluation: The use of such a tool will be of assistance to designers and developers who
do not have such expert skills in web site usability.
• Overcomes inconsistency in the usability problems that are identified: By removing
the human element, automation removes the inconsistencies in the usability problems
that are detected as well as any misinterpretations and wrong application of usability
guidelines.
• Enables the prediction of the time and costs of errors across a whole design: Since
automated evaluation tools perform usability evaluation methodically, they are more
consistent and cover a wider area in their evaluation and thus, one can better predict the
time and cost required to repair usability errors that are identified

• Increases the coverage of the usability aspects that are evaluated: Automation
overcomes commercial constraints such as those associated with time, cost and
resources, which typically limit the depth of evaluation
• Enables the evaluation between different potential designs: Commercial constraints
limit the evaluation against one design or a group of features. Automated evaluation
software provides designers with an environment where alternative designs can be
evaluated.
• Facilitates the evaluation in various stages of the design process: An interface can
be evaluated and any usability issues identified and resolved early, thus saving time and
costs that would be incurred should it be addressed at a later stage.
• Is of immediate value in the web design and development domain: Brinck and Hofer
[25] state that due to the large number of web designers and developers, a tool that
enables the evaluation of a web site is something that will appeal to this large community.
2.3 Attempts at automating web site usability evaluation
Chi et al [6] state that there are two types of tools that can perform automated usability
evaluation. These categories refer to tools that:
• Make use of conformance to standards
• Try to predict the usage of a web site
Through the research carried out for this study, four tools have been identified that perform the
usability of a web site, all of which fall in the second category. These are Cognitive Walkthrough
of the Web - CWW [30] Web Tango [31], WebCriteria Site Profile [5] and Bloodhound [6]. All three
solutions base their usability evaluation through usage prediction - something which various
researchers such as Groves [32], Winckler et al. [33] and Murray and Costanzo [34] argue
against since this method is based on prediction algorithms that can provide misleading data.
3. MAINSTREAMING WEB SITE USABILITY THROUGH USEFUL
3.1 The Components of the USEFul Framework
The framework that is being proposed in this study has been named USEFul (USability
Evaluation Framework). Unlike the previous attempts at automated usability evaluation, USEFul
falls in the first category proposed by Chi et al [6].
This is because it is based on research conducted by Jeffries et al. [21], Otaiza et al. [19] and
Tobar et al. [24], who identify Heuristic Evaluation, that is, the evaluation of the interface with
respect to a set of usability principles [35] as the usability evaluation technique that manages to
detect the majority of global usability issues, from all the usability evaluation techniques that they
evaluated. Similarly, studies by Comber [7], Ivory and Hearst [29] and Tobar et al. [24], show that
adherence to guidelines can effectively contribute towards making a web site more usable.
Thus, the USEFul framework will reference web site usability guidelines and use them to
automatically assess the usability of a web site that is being evaluated.

FIGURE 1: A visual representation of the USEFul framework (Source: Authors)
The different components of the USEFul framework as illustrated in Figure 1 are the:
• User - the person who is conducting the usability evaluation of a web site
• Web site - the means by which the user can interact with the USEFul framework. To
specify which web site needs to be evaluated, the user needs to key in the web site's
name, its tag line and URL.
• Web service - The web service communicates with the library by calling the execute
function from the library and passes it the parameters that it needs to evaluate the web
site. Once the evaluation is complete, the web service passes the results of the
evaluation back to the web site.
• Library - The library contains the program that carries out the evaluation. In order to
carry out the evaluation, it retrieves the data from the database.
• Database - The database is an SQL relational database that contains 4 tables
o Usability Category table - stores the usability categories available (Section 3.1)
o Implementation Level table - stores the implementation levels available (Section
3.2)
o Guidelines Definitions table - stores the guidelines that will be used in the
framework, expressed in natural language. This table also stores the Priority Rating
(Section 3.2) of each guideline and references the usability and implementation level
tables.
o Rule Type 1 table - In this table, the guidelines from the guidelines definitions table
that have green or amber implementation level (Section 3.2) are expressed in a form
that the library can interpret to carry out the evaluation. For each record, the fields
contain the HTML tag along with its additional data such as its attribute and size that
the library needs to search for so as to find the pattern that identifies a specific
guideline. The rule type 1 table also allows the comparison or searching of two
HTML tags or tags within tags. An important column in this table is the "ruleSuccess"
column as the fields in it store a value that the library interprets as the conditions
under which the guideline is considered to have been violated or not. This is
important since it distinguishes between guidelines that must be adhered to and thus
must be present in the web site and those that must never be found or can only be
found once as otherwise they would cause a usability violation.
At this point, it is worth mentioning that the current build of the USEFul framework
contains just 1 rule type (Rule type 1 table), that enables the library to identify the
guidelines that relate to HTML tags or CSS selectors. It is envisaged that future builds will

enable the addition of new rule types such as ones that enable the library to evaluate
usability guidelines related to images and other resources used by the web site. Also,
should new structuring tools such as a JavaScript parser be incorporated into the
framework, then these would require new rule types to be incorporated.
3.2 The Set of Guidelines That Will be Used by the USEFul Framework
Over the years, a number of usability guidelines have been published such as those by Smith and
Mosier [36], Norman [37], Nielsen [38], Comber [7], Sano [39], Borges et al. [27], Spool et al. [40],
Fleming [41], Rosenfeld and Morville [42], Shneiderman [43], Nielsen [44], Dix et. al. [15] and
Nielsen and Loranger [45].
However, the problem with Usability guidelines is that there is no set of guidelines that has been
established as a standard [15]. Thus, a set of 240 guidelines has been compiled for this study
from the results of usability studies carried out by researchers and experts in the fields of
cognitive psychology, technical communication, computer science, human factors and usability.
Since most of the proposed guidelines have been retrieved from the U.S Department of Health
and Human Services' (HHS) Research-Based Web Design & Usability Guidelines [46], the same
categorization has been used, that is, each guideline was placed in 1 of the 15 categories shown
in Table 1
TABLE 1: The number of guidelines used by the USEFul framework in each category (Source: Authors)
3.3 Assigning the Priority Rating and Level of Implementation to each Guideline
Each guideline used in the USEFul framework has been assigned an Implementation Level
which denotes the ability (or otherwise) to translate that guideline into a form which can then be
referenced by the program. This gives an indication as to what automation level each guideline
has. The parameters that will be used for this classification are as shown in Table 2:
Usability Category Number of Guidelines
Optimizing the user experience 29
Hardware and software 4
The homepage 12
Page layout 9
Navigation 27
Scrolling and paging 3
Headlines, titles and labels 18
Links 21
Text appearance 18
Lists 13
Screen based controls (widgets) 27
Graphics, images and multimedia 17
Writing web content 18
Content organization 8
Search 16
Total Guidelines 240

TABLE 2: How the guidelines should be interpreted in terms of their Implementation Level (Source: Authors)
Since the resources to tackle usability violations are typically scarce, the evaluator carrying out
manual usability evaluation prioritizes them so that the violations that cause the highest problems
are addressed first [47, 45].
One of the most used prioritization techniques is the severity scale, whereby each guideline is
given a severity rating [48]. For this project a severity scale called Priority Rating (PR) is
proposed, whereby each guideline is assigned a PR from 1 to 5 where a guideline with PR 5 is
very important in terms of its contribution towards making a web site usable, whilst a guideline
with PR 1 provides minor contribution. This prioritizes the list of usability violations identified by
the program.
3.4 Incorporating the Guidelines into the SQL Database
Usability guidelines are occasionally abstract and difficult to interpret and apply [26, 20, 27, 28,
49]. This has been addressed in the USEFul framework through the use of the guidelines
definitions and rule type 1 tables (Section 3.1). The process through which a guideline is entered
into these 2 tables is illustrated below through the use of one of guideline#81 which is one of the
guidelines used in the USEFul framework:
3.4.1 Guideline Expressed in Natural language
Guideline#81: URLs should not be complex and should ideally be less than 50 characters. This is
beneficial for both usability and SEO [45]
3.4.2 Guideline as Entered in the Guidelines Definitions Table
When the guideline is entered in the guidelines definition table, its primary key is 81. The
guideline and its explanation have been split into the fields under the "Guideline" and "Reason"
columns respectively. The value in the field "ruleType" is 1 since this guideline needs to be
evaluated using the rule type 1 rule. The guideline is a green guideline, hence the value 1 under
the "ruleCat" column and its Priority Rating is 5, hence the reason why there is a "5" under the
Implementation
Level Category
Interpretation
Green • Guideline can be fully implemented in the database within the USEFul
framework.
• The framework is able to automatically determine whether this guideline applies
to the web site being evaluated.
• The results returned by the framework when referring to this guideline are
conclusive since these types of guidelines are typically measurable, with clearly
defined parameters.
Amber • Guideline is harder to fully implement in the USEFul framework.
• Certain patterns that automatically identify if this guideline may apply to the web
site being evaluated have been implemented in the database.
• This guideline can be converted into a "green" guideline by incorporating within
the USEFul framework additional Artificial Intelligence algorithms.
• The results outputted by the framework when referring to this guideline consist
of data that can assist the human evaluator in checking whether it applies to the
web site being evaluated
Red • This guideline is typically abstract and requires user intervention or very
advanced algorithms from the field of Artificial Intelligence or additional
technology to make it possible for it to be implemented in the framework.
• Through the use of advanced algorithms or technology, it can be converted into
"amber" or "green" guideline
• In its current build, the framework lists this guideline so that the human
evaluator can manually check if it applies to the web site being evaluated

ruleSeverity column. Since this guideline belongs to the "navigation" usability category, the value
"5" has been entered in the field under the ruleGroup column.
TABLE 3: How guideline#81 is represented in the guidelines definition table (Source: Authors)
3.4.1 Guideline as Entered in the Rules Type 1 Table
To check whether the guideline is found in the parsed HTML, the execute function needs to
search for the following pattern:
<a href="any text as long as it is less than 70 characters"> .. </a>
Therefore, the guideline is converted to a form that the execute function can understand and this
is stored as a record in the rule type 1 table (Section 3.1) as shown below:
TABLE 4: How guideline#81 is represented in the rules type 1 table (Source: Authors)
Thus, the value "81" is a foreign key that references the primary key "81" in the Guidelines
Definition table. The execute function will look for is the "a" tags which have an "href" attribute as
stated by the contents in the fields under the "tagA" and "attributeA" columns respectively. The
exact content in between the inverted commas of the "href" attribute is irrelevant, hence the
reason for the NULL value in the field under the "valueA" column.
However, for the guideline not to be violated, this content needs to be less than 70 characters
long, as stated by the contents in the fields under the "compareOperator and "sizeA" columns
respectively. Since the guideline is not dependent on any other HTML tags, the fields under the
four tag B columns are all set to NULL. If the guideline matching the pattern in this record is
found, then it is a good thing, hence the reason why the value under the "ruleSuccess" column is
True. Also, for the guideline to succeed, all the content of all "href" attributes within all the "a" tags
found must be less than 70 characters. This is set by the "0" value in the field under the
"mustSucceed" column.
4. HOW THE PROPOSED FRAMEWORK EVALUATES WEBSITE USABILITY
This section will describe the process that takes place from when the user accesses the USEFul
web site and keys in the parameters pertaining to the web site they would like to evaluate to when
the results of the usability evaluation are reported on the web site. This description will thus
discuss on a high level the interactions that take place between the various components within
the framework represented in Figure 1.
pk Guideline Reason ruleType ruleCat rule
Severity
rule
Group
81 URLs should
not be
complex
URLs should ideally be
less than 50 characters.
Such URLs are beneficial
for both usability and SEO
1 1 5 5
pk ruleFk tagA attributeA valueA sizeA tagB
12 81 a href NULL 70 NULL
attributeB valueB sizeB rule
Command
compare
Operator
rule
Success
must
Succeed
NULL NULL NULL NULL < True 0

4.1 Step 1 - The user passes the data to the web site
This step refers to when the user states what web site the USEFul framework will need to
evaluate. They do this by filling in the text fields pertaining to the web site's name, URL and
(optionally) the tag line. The website's GUI can be seen in the screenshot below (Figure 2)
FIGURE 2: Screenshot of the USEFul web site (Source: Authors)
4.2 Step 2 - The Web Site Passes the Data to the Web Service
When the user fills in the data in Step 1 and presses the "Evaluate Web Site" button, the web site
passes the data as parameters to the web service.
4.3 Step 3 - The Web Service Configures the Library
When the web service receives the parameters from the web site, it communicates with the
library and creates a new instance by setting the configuration values in the library according to
these parameters. It is important to note that in reality, the library is actually contained within the
web service. The only reason why the library was illustrated as a component outside the web
service in Figure 1 is to create a distinction between the two components for explanation
purposes. Therefore, the phrase "communicates" is being used to illustrate the flow of data
between the library and the web service.
Thus, the web service sets the path of the web site, the company name and the tag line. The web
service then uses the library's functionalities to load and parse the web site to create parsed
HTML and CSS documents. These parsed documents are stored in the web service in static
variables.
4.4 Step 4 - The Web Service Fetches the Guidelines
The web service communicates with the library which in turn communicates with the SQL
database to fetch the data stored in the guidelines and rule type tables. The returned data is
stored inside the web service in a data table as shown in Table 5 below:

TABLE 5: The Structure of the Data Table stored in the Web Service (Source: Authors)
The components of the data table shown in Table 5 are the following:
• Guideline definition: This contains a copy of the data found in the guidelines definitions
table (Section 3.1)
• Rule type properties: This field contains a copy of the rules type 1 table (Section 3.1)
• Results fields: The results fields are additional fields initially set as empty when the data
table is created in the web service. These fields will eventually contain the results that the
web service will communicate back to the web site after the evaluation is completed.
These are:
o Tags: will contain the number of times the HTML tag or CSS selector found in the
field under the rule type properties column in the data row is found.
o Success: The number of tags or selectors found whose properties match the
properties of the guideline being referenced
o Fail: The number of tags or selectors whose attributes match the properties of
the guideline being referenced but their value or size properties do not match
o Null: The number of tags or selectors whose attributes, sizes or properties do not
match with the property of the guideline
o Success%: This value is the result of the equation Success/(Tags-Null) x 100
o Passed: This field will eventually contain a True/False value that will indicate
whether the guideline has been violated or not
4.5 Step 5 - The Web Service Uses the Library to Evaluate the Web Site
At the end of Step 4, the web service contains a copy of the parsed HTML and CSS documents
which are stored in static variables. It also has a data table as shown in Table 5.
The web service then takes the first data row and calls the execute function from the library. The
execute function takes 1 data row (1 row of the data table as indicated in Table 5) as a
parameter. When the execute function in the library receives the data row from the web service, it
sees what rule it has to work on in order to evaluate whether the usability guideline defined in that
data row is being violated.
We will use guideline#81 as an example to illustrate the steps the library performs to assess
whether the web site being evaluated adheres to a guideline or if it violates it. It is being assumed
that guideline#81 has been entered in both the guidelines definition table and the rules table as
indicated in Tables 3 and 4 respectively.
When the web service passes calls the execute function with the data row pertaining to
guideline#81, each time the execute function is run, and it finds an "a" tag it follows the logic tree
shown in Figure 3 below. As it can be seen in Figure 3, the execute function can take different
paths, depending on what the fields contain and the type of pattern matching that it needs to
perform.
Guideline Definition Rule Type Properties Results Fields
1 data row

FIGURE 3: The Logic Tree for Rule Type 1 (Source: Authors)
At the end of this process, the execute function returns the evaluation result which can have 1 of
3 values: True, False or Null. In the case of guideline#81, these are interpreted as follows:
• True: The "a" tag has an "href" attribute whose content is less than 70 characters
• False: The "a" tag has an "href" attribute whose content is 70 characters or more
• Null: The "a" tag does not have an "href" attribute
Suppose that the execute function finds 3 instances of the "a" tag in the parsed HTML document
and these are as follows:
• True: 2
• False: 1
• Null: 0
For each evaluation result, the execute rule compares it with the value of the field under the
"ruleSuccess" column in Table 4. During the same process, it uses the result of the comparison to
increment the counters of the values that will be written in the results fields of the data table in the
web service. The method of comparison is modeled on the XNOR truth table and can be seen in
Table 6 below:
Get Tag
Execute Function
Get Attribute
Get Value
Get Size
M
M
M
C
Match Attributes
Match Tags
Match Values
Match Sizes
M
C
Check With Numeric
Check With Text
C
Process performed by the execute function
Connector - The logical flow can proceed
from any connecting point
Logical flow followed by the execute function
Legend

TABLE 6: How guideline#81 is represented in the guidelines definition table (Source: Authors)
Thus, assuming that in the case of the guideline#81 example, the sequence in which the
evaluation results are issued by the execute function are 2 True, 1 False and 1 null, then the
execute rule would make the following comparisons:
TABLE 7: The comparisons made by the execute function for the guideline#81example (Source: Authors)
In this way, the values for the results fields of the data row for guideline#81 in the web service
(Table 5) would be as shown in Table 8 below:
TABLE 8: The values in fields of the data row for guideline#81 (Source: Authors)
As can be seen in the table above, at this stage, it has not been determined whether the guideline
has passed or not. In fact, to set this value, the execute function looks up the value of the field
under "mustSucceed" in Table 4 to see under which conditions it can be stated that guideline#81
has not been violated.
Since the value in this case is "0", the execute rule interprets it that for the guideline not to be
violated, this guideline must not fail, meaning that the fail counter in Table 8 should be "0". Since
this is not the case, the execute rule sets the value of the "Passed" field in Table 8 to "FALSE".
This effectively means that the guideline has been violated.
This value is then stored in the passed field of the data row for guideline#81 in the web service.
4.6 Step 6 - The Web Service Sends the Data Table to the Web Ste
Once the data table is complete, the web service sends it to the web site which creates 3 data
views, one for each implementation level. It also sorts the violations in each implementation in
descending order of priority rating and displays them as shown in Figure 4.
Input A (Evaluation Result) Input B (ruleSuccess) Output (A XNOR B)
False: Guideline not found False: Bad guideline True: Increment success counter
False: Guideline not found True: Good guideline False: Increment fail counter
True: Guideline found False: Bas guideline False: Increment fail counter
True: Guideline found True: Good guideline True: Increment success counter
Null: Guideline not applicable True: Good guideline Null: Increment null counter
Null: Guideline not applicable False: Bad guideline Null: Increment null counter
Input A (Evaluation Result) Input B (ruleSuccess) Output (A XNOR B)
True True True: Increment success counter
True True True: Increment success counter
False True False: Increment fail counter
Results Field Value
Tags 3
Success 2
Fail 1
Null 0
Success% 66.7
Passed NULL

FIGURE 4: Screenshot showing evaluation results carried out by the USEFul framework (Source: Authors)
5. EXPERIMENTS AND RESULTS
Since web site usability professionals are scarce (a current limitation mentioned in Section 2.1), it
was decided that the effectiveness of the USEFul framework will be assessed by comparing the
results of the evaluations carried out on web sites against published evaluations of the same web
sites carried out by web site usability professionals.

The most reliable source for this evaluation was identified to be the book "Homepage Usability -
50 Websites Deconstructed" by Nielsen and Tahir [50]. The reasoning behind this chosen method
of experimentation is based on the following points:
Dr. Jakob Nielsen is considered to be a web usability guru [51, 52] and has been hailed
as "one of the world's foremost experts in web usability" [53].
The book itself illustrates in a very clear manner the usability violations that have been
identified by Nielsen and has received numerous positive reviews [54, 55].
Nielsen evaluates the web sites featured in this book by referencing web site usability
guidelines. This usability evaluation technique is the same technique used by the USEFul
framework. This eliminates any possibilities that any difference in the list of identified
violations is as a result of different techniques being employed.
The set of guidelines used by Nielsen for this evaluation is a subset of the HHS
Research-Based Web Design & Usability Guidelines [46]. In fact, this book is listed as
one of the cited sources. As stated in Section 3.2, the majority of the guidelines
implemented in the USEFul framework are from the HHS guidelines.
Although the HHS Research-Based Web Design & Usability Guidelines have been
retrieved in February 2011 and thus it can be assumed that they are still relevant today,
any flaws in these guidelines does not affect the performance of the USEFul framework
since the guidelines are not hard coded into the library itself. Moreover, using the same
set of guidelines as those used by the human evaluator for these tests eliminates the
possibility that any discrepancies in the results were due to different sets of guidelines
being used.
On inspection of the results reported by evaluation carried by Nielsen and Tahir it was noticed
that they also mention some positive usability characteristics. The guidelines that have been
observed which have led to these positive traits have also been incorporated in this experiment.
Since the USEFul framework only reports usability violations, the absence of these guidelines in
the list of detected violations was thus interpreted as a positive result. In this regard, since what
Nielsen and Tahir reported were both positive as well as negative comments, the term "usability
aspects" will be used instead of usability violations so as to avoid the negative connotation
associated with the word "violation".
Due to their expertise in web site usability, Nielsen and Tahir also list a number of positive as well
as negative usability aspects which are specific to the web site being evaluated and their linkage
to any of the guidelines could not be established. These site-specific aspects were incorporated
in this study with the red guidelines since they could not be automatically evaluated by the
USEFul framework in its present form.
Ten web sites from the book were selected on the basis that their evaluation contained less site-
specific recommendations and more green and amber guideline related usability aspects. This
means that from the authors' evaluation, it was easier to identify which guidelines from the HHS
Research-Based Web Design & Usability Guidelines were being violated. Special attention was
taken to select web sites that violated different guidelines so as to increase the set of guidelines
that will be incorporated into the database for evaluation.
So as to ensure that the versions and contents of the web sites evaluated are identical to the
ones evaluated by Nielsen and Tahir, the Internet Archive's Way Back Machine [56] was used.
Using this tool, the exact web sites were loaded by utilizing the dates present on the screenshots
in the book. Based on these criteria, the 10 web sites chosen for this experiment are those found
in Table 9.

.
TABLE 9: The web sites that were used for this experiment (Source: Authors)
So as to be able to identify the usability violations identified by Nielsen and Tahir in these web
sites 62 guidelines (36 Green, 27 Amber and 0 Red / Site-Specific guidelines) from the set of
guidelines mentioned in Section 3.2 were used for this experiment. The results of the evaluations
can be seen side by side in Table 10:
.
TABLE 10: Usability evaluations carried out by Nielsen & Tahir against USEFul (Source: Authors)
As can be seen in the figures presented in the column Results 1 of Table 10, the USEFul
framework was able to correctly identify the guideline-related usability aspects, 95.86% of the
time when compared to Nielsen and Tahir's manual evaluation. When the code was inspected to
identify why there was a 4.14% discrepancy it was found that the main reason why the framework
Web Site Tag Line URL
About The Human Internet http://replay.waybackmachine.org/200106110
62521/http://www.about.com/
Accenture Now It Gets Interesting http://replay.waybackmachine.org/200107111
42024/http://www.accenture.com/
Asia Cuisine Asia's Leading Food and
Beverage Portal
http://replay.waybackmachine.org/200107030
30929/http://www.asiacuisine.com.sg/
Barnes & Noble (No Tag Line) http://replay.waybackmachine.org/200102031
220/http://bn.com/
BBC Online Welcome to the UK's Favourite
Website
73705/http://www.bbc.co.uk/
Boeing Forever New Frontiers http://replay.waybackmachine.org/200105221
94801/http://www.boeing.com/
DIRECTV America's Leader in Digital
Home Entertainment
15718/http://www.directv.com/
FedEx (No Tag Line) http://replay.waybackmachine.org/200105250
31739/http://www.fedex.com/us/
Red Herring The Business of Innovation http://replay.waybackmachine.org/200105152
22459/http://redherring.com/
The Art Institute of
Chicago
(No Tag Line) http://web.archive.org/web/20010630180812/
www.artic.edu/aic/index.html
Web Site Usability aspects identified by
Nielsen and Tahir
Nielsen and Tahir’s usability aspects
identified by USEFul
Green Amber Red Total Results 1:
As a percentage of
implementable
usability guidelines
(Green & Amber)
Results 2:
As a percentage of
total usability
guidelines (Green,
Amber, Red & Site
Specific)
About 7 7 11 25 100.00% 56.00%
Accenture 6 8 13 27 85.71% 44.44%
Asia Cuisine 7 3 11 21 100.00% 47.62%
Barnes & Noble 11 4 13 28 93.33% 50.00%
BBC Online 17 6 15 38 100.00% 60.53%
Boeing 9 3 12 24 100.00% 50.00%
DirectTV 14 1 12 27 86.67% 48.15%
FedEx 10 8 11 29 100.00% 62.07%
Red Herring 9 3 17 29 91.67% 37.93%
The Art Institute of
Chicago
7 5 10 22 100.00% 54.55%
Total 97 48 125 270 Average: 95.86% Average: 51.48%

failed was due to bad coding in the web sites being tested. In fact, the primary cause was the use
of images to represent text instead of using actual text. So as to minimize the impact of this
limitation, most of the guidelines' interpretation in the table also referenced the alt attribute of
images. When the alt attribute was not present, then the USEFul framework was not able to
parse the text represented by those images since it does not currently have Optical Character
Recognition (OCR) facilities. Additionally, this discrepancy can also be attributed to the lack of
proper usage of certain HTML tags such as the use of the <p> tag instead of the <h1>..<h6> tags
for headings.
Since only 53.71% of the usability aspects were directly related to green and amber guidelines, it
can be seen that overall, the number of violations reported by the USEFul framework on average
relates to just 51.48% of the total usability aspects identified by Nielsen and Tahir. This finding
shows why various researchers [6, 57, 29, 58] suggest that any tool that automatically evaluates
a web site cannot replace a human being. In this case, because of their experience and expertise
in web usability, Nielsen and Tahir were able to identify almost as many usability guidelines that
have been classified as red guidelines or site-specific recommendations as those that were
classified as green or amber. In this regard, it is clear that the USEFul framework cannot
implemented without the inclusion of a human evaluator.
An interesting observation in these experiments is that since the USEFul framework checks for
the presence of each guideline in the database, it performs a consistent evaluation and thus it
was able to identify more usability violations in each of the tested web sites. At this point it is
important to note that the term being used is usability violations since the USEFul program can
only report usability violations. The HTML and CSS code was then inspected manually so as to
confirm that these additional usability violations were correct.
When the additional usability violations identified by the USEFul framework are compared to the
total aspects identified by Nielsen and Tahir (manually), it can be noted that the framework was
able to identify on average 128.15% usability violations (Table 11: column Results 3) i.e. 28.15%
more violations than what Nielsen and Tahir actually identified. The total number of additional
violations identified are shown in Table 10.
TABLE 11: Additional usability violations identified by the USEFul framework (Source: Authors)
This means that despite the fact that only 53.71% of the usability attributes identified by Nielsen
and Tahir can be converted into green and amber guidelines, the USEFul framework was still
able to detect 28.15% more usability violations using this limited set of 62 guidelines.
Web site Additional usability violations
detected by USEFul
RESULTS 3:
Total usability aspects detected
by USEFul as a percentage of
total usability aspects identified
by Nielsen & Tahir
Green Amber Total
About 11 9 20 136.00%
Accenture 13 9 22 125.93%
Asia Cuisine 16 15 31 195.24%
Barnes & Noble 13 12 25 139.29%
BBC Online 7 10 17 105.26%
Boeing 9 8 17 120.83%
DirectTV 9 7 16 107.41%
FedEx 9 9 18 124.14%
Red Herring 14 12 26 127.59%
The Art Institute of Chicago 6 9 15 122.73%
Total 107 100 207 Average: 128.15%

Another interesting find in these experiments suggests that through the USEFul framework, a
usability expert is still likely to identify more usability violations. Table 12, which summarizes the
results of all the tests carried out with the 10 web sites, illustrates this point.
TABLE 12: Summary of the results from the experiments
From the results shown in Table 11, Nielsen and Tahir were able to comment on an average of
27 usability aspects per web site (270 ÷ 10). whilst USEFul was able to identify 34.6 usability
aspects (mainly violations) per web site (346 ÷ 10) which translates to 28.15% more usability
violations being detected.
If a usability expert were to make use of the USEFul framework, then they would be able to detect
the green, amber, red and site-specific usability aspects whilst the framework would still report
the additional usability violations. This would mean that using the USEFul framework to evaluate
the 10 web sites above, the expert would have commented on a total of 477 usability aspects
(145 + 125 + 207), that is, an average of 47.7 usability aspects per web site (477 ÷ 10). This can
also be interpreted as an increase of 76.67% in the number of usability aspects that the expert
evaluator can make per web site.
6. LIMITATIONS
As recommended by various researchers [6, 57, 29, 58], the purpose of any tool such as the one
being proposed is to provide assistance to human evaluators. Ivory and Chevalier [57] advise that
such tools should always be used with caution and one should never completely rely on their
results alone. This is because with current technology, it is difficult to develop a tool that can
behave like a human and exhibit human attributes such as common sense [16, 58]. This is
partially addressed in the USEFul framework through the assignment of the Implementation Level
to denote the possible level of automation for each guideline. Still, it was observed that the
USEFul framework was not able to handle Nielsen's site-specific recommendations. Such
recommendations can be made by a human evaluator through the application of logic, experience
and techniques such as grouping guidelines.
Another limitation is the difficulty encountered with incorporating certain guidelines into the rule
types tables in the framework's database, particularly because of their abstract nature. At
present, the only way to incorporate such guidelines into the USEFul framework is to introduce
certain assumptions as recommended by Dix et al. [15] and Vanderdonckt and Beirekdar [59].
Additionally, the proposed guidelines are aimed at evaluating the usability of web sites that have
business goals, meaning that these web sites serve to promote and/or sell products and/or
services either online via the web site itself or through offline channels [45, 46]. Thus, web sites
that do not fall in this category require a different set of guidelines. The framework facilitates this
Item # Usability aspects detected by
Nielsen & Tahir USEFul
1 Number of web sites evaluated 10 10
2 Green and Amber guideline-related
usability aspects detected
145 139
3 Red and Site-Specific usability
aspects detected
125 0
4 Additional usability violations
detected
0 207
Total 270 346

process since it reduces it to incorporating the new guidelines into the database tables without
modifying the code.
7. CONCLUSION
From the results obtained in this evaluation, it can be concluded that the USEFul framework is
very effective at identifying usability aspects that violate usability guidelines. However, bad coding
practices can adversely affect the results obtained. The USEFul framework references each
guideline that has been implemented in its SQL database to see if the web site has violated it. In
the experiments that have been carried out, this factor has enabled it to identify more violations
than the usability experts. However, the framework cannot detect what have been classified as
red or site-specific violations because it lacks the logic, experience and expertise that an expert in
web usability has. In this regard it has been concluded that the framework cannot replace a
human evaluator but should be used to assist an evaluator. In fact, the results indicate that using
the framework, a usability expert is likely to be able to detect more usability violations
Additional research needs to be carried out to make the framework more flexible so as to be able
to implement more abstract guidelines that are currently being classified as having a red
implementation level. Currently, the framework can parse HTML and inline and internal CSS
code. Therefore, further enhancements that are planned include the ability to parse external CSS
stylesheets as well as Javascript parsing since these can considerably affect the way the user
sees the web site when rendered through a web browser. So as to overcome the problems with
analyzing the content of images in web sites, image processing and Optical Character
Recognition algorithms can help in addressing this issue. By implementing these enhancements
as well as other algorithms that may be deemed as beneficial, the framework can truly contribute
towards mainstreaming web site usability.
8. REFERENCES
1. International Organisation for Standardisation (ISO) DIS 9242-11, (1998) Available at:
http://www.iso.org
2. P. Paolini. “Hypermedia, the web and usability issues” In IEEE International Conference on
Multimedia Computing and Systems. pp. 9-11 IEEE, 1999
3. M. Swaak, M. De Jong, P. De Vries. “Effects of Information usefulness, visual attractiveness,
and usability on web visitors’ trust and behavioural intentions.” In IEEE International
Professional Communication Conference. P. 1-5. Waikiki, HI, USA, 2009
4. P. Vora. “Designing for the web: a survey”. Interactions. pp. 13-30 (1998)
5. G. Brajnik. “Automatic web usability evaluation: what needs to be done?”. In Proceedings of
the 6th
Conference on Human Factors and the Web, Austin Texas, USA, 2000
6. EH. Chi, A. Rosien, G. Supattanasiri, A. Williams, C. Royer, C. Chow, et al. “The
Bloodhound project: automating discovery of web usability issues using the InfoScent™
simulator” In Proceedings of the ACM CHI 03 Conf.. pp. 505-512. Ft.Lauderdale, FL, USA,
2003
7. T. Comber. “Building usable web pages: an HCI perspective”. In Proceedings of the
Australian Conference on the Web AusWeb'95. pp. 119-124. Ballina, Australia, 1995
8. J. Nielsen. “User interface directions for the web” Communications of the ACM. 1999, pp.65-

72
9. Cabinet Office. (2007) Available at http://archive.cabinetoffice.gov.uk/e-
government/resources/quality-framework.asp [Accessed 9 October 2010]
10. European Agency for Safety and Health at Work in European Union Council Directives.
(1990) Available at: http://eur-
lex.europa.eu/LexUriServ/LexUriServ.do?uri=CONSLEG:1990L0270:20070627:EN:HTML
[Accessed 9 October 2010]
11. SA. Becker, FE. Mottay. “A global perspective on website usability” IEEE Software, Jan/Feb,
18(1): pp.54-61, 2001
12. P. Fraternali, M. Tisi. “Identifying cultural markers for web application design targeted to a
multi-cultural audience” In the 8th International Conference on Web Engineering ICWE'08.
pp. 231-239. Yorktown, NJ, USA, 2008
13. F. Montero, P. Gonzáles, M. Lozano, J. Vanderdonckt. “Quality models for automated
evaluation of web sites usability and accessibility”. In International COST294 Workshop on
User Interface Quality Model. Rome, Italy, 2005
14. R. Ruiz-Rodríguez. “An auxiliary tool for usability and design guidelines validation of web
sites” In Proceedings of the 15th International Conference on Computing - CIC'06. pp.304-
308. Mexico City, Mexico, 2006
15 A.Dix, J. Finlay, GD. Abowd, R. Beale. “Human Computer Interaction” 3rd
Ed. Pearson
Education Ltd. Essex, (2004)
16. K. Norman, E. Panizzi, “Levels of automation and user paricipation in usability testing”
Interacting with Computers, p. 246-264, 2006
17. J. Nielsen, R. Molich. “Heuristic evaluation of user interfaces” In Proceedings of the SIGCHI
conference on Human Factors in Computing Systems: Empowering People. pp. 249-256.
Seattle, Washington, USA, 1990
18. P. Polson, C. Lewis, J. Rieman, C. Wharton. “Cognitive walkthroughs: a method for theory-
based evaluation of user interfaces” International Journal of Man-Machine Studies, pp. 741-
773, 1992
19. R. Otaiza, C. Rusu, S. Roncagliolo. “Evaluating the usability of transactional web sites” In
Third International Conference on Advances in Computer-Human Interactions. pp. 32-37.
Saint Maarten, Netherlands, Antilles, 2010
20. A. Beirekdar, J. Vanderdonckt, M. Noirhomme-Fraiture. “KWARESMI - knowledge-based
web automated evaluation tool with reconfigurable guidelines optimization” In Proceedings
of the 9
th
International Workshop on Design, Specification and Verification of Interactive
Systems DSV-IS. pp. 362-376. 2002

21. R. Jeffries, J. Miller, C. Wharton, KM. Uyeda. “User interface evaluation in the real world: a
comparison of four techniques” In Proceedings of the ACM Computer Human Interaction
CHI'91 Conference. pp.119-124. New Orleans, LA, USA, 1991
22. J. Vanderdonckt, A. Beirekdar, M. Noirhomme-Fraiture. “Automated evaluation of web
usability and accessibility by guideline review”. Lecture Notes in Computer Science, pp. 17-
30, 2004
23. A. Beirekdar, J. Vanderdonckt, M. Noirhomme-Fraiture. “A framework and a language for
usability automatic evaluation of web sites by static analysis of HTML source code” In 4th
International Conference on Computer-Aided Design of User Interfaces. pp. 337-349. 2002
24. LM. Tobar, PM. Latorre Andrés, E. Lafuente Lapena. “WebA: a tool for the assistance in
design and evaluation of websites”. Journal of Universal Computer Science, pp. 1496-1512,
2008
25. T. Brinck, E. Hofer. “Automatically evaluating the usability of web sites” In Conference on
Human Factors in Computing Systems CHI'02. pp. 906-907. Minneapolis, Minnesota, USA,
2002
26. A. Beirekdar, M. Keita, M. Noirhomme, F. Randolet, J. Vanderdonckt, C. Mariage. “Flexible
reporting for automated usability and accessibility evaluation of web sites” INTERACT, pp.
281-294, 2005
27. JA. Borges, I. Morales, NJ. Rodríguez. “Guidelines for designing usable world wide web
pages” In Proceedings of the Conference on Human Factors in Computing Systems:
common ground. pp. 277-278. Vancouver, British Columbia, Canada, 1996
28. M. Burmester, J. Machate. “Creative design of interactive products and use of usability
guidelines - a contradiction?” In J. Jacko, C. Stephanidis, D. Harris. “Human-computer
interaction: theory and practice", Lawrence Erlbaum Associates Inc., pp. 43-46 (2003)
29. MY. Ivory, MA. Hearst. “The state of the art in automating usability evaluation of user
interface” ACM Computing Surveys (CSUR), pp. 470-516, 2001
30. MH. Blackmon, PG. Polson, M. Kitajima, C. Lewis. “Cognitive walkthrough for the web” In
Proceedings of the SIGCHI Conference on Human factors in Computing Systems: Changing
our World, Changing Ourselves. pp. 463-470. Minneapolis, Minnesota, USA, 2002
31. MY. Ivory. “Web TANGO: towards automated comparison of information-centric web site
designs” In Proceedings of the ACM CHI 00 Conference on Human Factors in Computing
Systems, Student Posters. pp. 329-330. 2000
32. K. Groves. “The limitations of server log files for usability analysis” (2007) Available at:
http://www.boxesandarrows.com/view/the-limitations-of [Accessed 22 January 2011]
33. M. Winckler, C. Freitas, J. Lima. “Usability remote evaluation for the WWW” In Proceedings
of the CHI'00 Extended Abstracts on Human Factors in Computing Systems. pp. 131-132

34. G. Murray, T. Costanzo. “Usability and the web: an overview” Information Technology
Services, #61 National Library of Canada, Network Notes (ISSN 1201-4338)
35. J. Nielsen. “Enhancing the explanatory power of usability heuristics” In Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems: celebrating
interdependence. pp. 152-158. Boston, Massachusetts, USA, 1994
36. SL. Smith, JN. Mosier. “Guidelines for designing user interface software” Mitre Corporation,
Report No.: MTR-9240 (1986)
37. DA. Norman. “The design of everyday things”, Doubleday (a division of Bantam Doubleday
Dell Publishing). (1988)
38. J. Nielsen. “The usability engineering life cycle” Computer (IEEE): 12-22, 1992
39. D. Sano. “Designing large-scale web sites: a visual design methodology” Wiley Computer
Publishing, John Wiley & Sons Inc. (1996)
40. JM. Spool, T. Scanlon, C. Snyder, T. DeAngelo. “Web site usability: a designer's guide”,
Morgan Kaufmann Publishers Inc. (1998)
41. J. Fleming. “Web navigation: designing the user experience”, O'Reilly & Associates, (1998)
42. L. Rosenfeld, P. Morville. “Information architecture for the world wide web”, O'Reilly &
Associates, (1998)
43. B. Shneiderman. “Designing the user interface: strategies for effective human-computer
interaction”, 3
rd
Ed. Addison-Wesley, (1998)
44. J. Nielsen. “Designing web usability: the practice of simplicity”, New Riders Publishing,
(1999)
45. J. Nielsen, H. Loranger. “Prioritizing web usability”, New Riders Press, (2006)
46. U.S. Department of Health and Human Sciences in Research-Based Web Design &
Usability Guidelines (2006) Available at:
http://www.usability.gov/guidelines/guidelines_book.pdf [Accessed 31 October 2010]
47. M. Hertzum. “Problem prioritization in usability evaluation: from severity assessments toward
impact on design” International Journal Human Computer Interaction, 21(2): pp. 125-146,
2006
48. G. Sim, JC. Read. “The damage index: an aggregation tool for usability problem
prioritisation” In Proceedings of HCI 2010. Dundee, Scotland, 2010
49. J. Ratner, EM. Grosse, C. Forsythe. “Characterization and assessment of HTML style
guides” In Proceedings of the Conference on Human Factors in Computing Systems:
common ground. pp. 115-116. Vancouver, British Columbia, Canada, 1996

50. J. Nielsen, M. Tahir. “Homepage Usability - 50 Websites Deconstructed”, New Riders,
(2002)
51. M. Richtel. (1998) from The New York Times: Available at:
http://www.nytimes.com/library/tech/98/07/cyber/articles/13usability.html [Accessed 27
January 2011]
52. P. Marks. (2001) from CNN: Available at:
http://edition.cnn.com/2001/WORLD/asiapcf/east/02/23/web.usability/index.html [Accessed
on 27 January 2011]
53. J. Hamilton. (2000) from Businessweek: Available at:
http://www.businessweek.com/2000/00_47/b3708076.htm [Accessed on 27 January 2011]
54. M. Miller. (2002) from PC Mag: Available at
http://www.pcmag.com/article2/0,2817,1165443,00.asp [Accessed on 27 January 2011]
55. L. Thomason. (2002) from NetMechanic: Available at
http://www.netmechanic.com/news/vol5/review_no21.htm [Accessed on 27 January 2011]
56. Internet Archive. Available at http://www.archive.org/ [Accessed on 10 January 2011]
57. MY. Ivory, A. Chevalier. “A study of automated web site evaluation tools” Technical Report
(Report No.: UW-CSE-02-10-01), 2002
58. T. Tiedtke, C. Märtin, N. Gerth. “AWUSA - a tool for automated website usability analysis” In
Pre-Proceedings of the 9th International Workshop on Design, Specification, and Verification
of Interactive System DSV-IS'2002. pp. 251-266. 2002
59. J. Vanderdonckt, A. Beirekdar. “Automated web evaluation by guideline review” Journal of
Web Engineering, pp. 102-117, 2005

USEFul: A Framework to Mainstream Web Site Usability through Automated Evaluation

More Related Content

Viewers also liked (20)

Similar to USEFul: A Framework to Mainstream Web Site Usability through Automated Evaluation (20)

More from Waqas Tariq (9)

Recently uploaded (20)

USEFul: A Framework to Mainstream Web Site Usability through Automated Evaluation