Content uploaded by Javier Andres Bargas-Avila
Author content
All content in this area was uploaded by Javier Andres Bargas-Avila on Oct 28, 2015
Content may be subject to copyright.
Online Form Validation: Don’t Show Errors Right Away.
Javier Bargas-Avila, Glenn Oberholzer
STIMMT AG, Webernstrasse 5, 8610 Uster, Switzerland
javier.bargas@stimmt.ch, glenn.oberholzer@stimmt.ch
Abstract: Online form validation can be performed in several ways. This reports discusses an empirical study
with 77 participants, which has found evidence that the best way of validating online forms is by reloading the
form with erroneous fields highlighted. ISO recommended immediate error feedback came in last in this study.
Additionally, a modal theory for form completion derived from the results postulates that users are either in
Completion or Revision Mode when filling out online forms. These modes affect the users’ way of interaction
with the system.
Keywords: Online Forms, HTML, JavaScript, Validation Mechanisms, Interaction Design, Interaction
Processes
1 Introduction
The use of the World Wide Web as a sales channel
places a lot of emphasis on interaction. HTML as
the base technology for websites offers only limited
interaction mechanisms. New technologies like
Macromedia Flash™ or Java™ promise to make up
for the shortcomings of regular HTML.
Nevertheless, HTML is still the most frequent way
to implement interaction – mostly due to its
standardization and widespread acceptance. The
exchange of data between users and servers happens
almost exclusively through forms in applications
like registration processes, shopping carts and guest
books.
Every human data entry is prone to errors –
validation of data is often a must to ensure data
quality and consistency. Several methods are used to
achieve this. To the authors’ knowledge, no research
has been conducted discovering the most effective and
most user friendly form validation method for HTML
forms. The study presented in this paper explores
this matter.
2 Theory
2.1 Form Validation
Due to the passive nature of HTML, several
helper technologies are used to validate forms on the
Net (i.e. Goodman 2001). Validation happens either
server-sided or client-sided (Wilde, 1999). If
validation is performed on the server, the technology
employed is irrelevant for the client. Client side
validation is performed mainly using JavaScript. The
validation mechanisms used in the study are
explained in the following sections.
Figure 1: Server-Side Validation – simultaneous
display of error messages (SV).
2.2 Server-Side Validation – Simult-
aneous Display of Error Messages
Validation is performed on the server and a page is
returned to the client, which highlights the fields
with missing or wrong entries (Figure 1). From an
HCI point of view, this method represents an elegant
solution; erroneous fields are explicitly highlighted
therefore reducing the cognitive load of the user.
Errors and the correct syntax do not have to be
Human-Computer Interaction -- INTERACT'03
M. Rauterberg et al. (Eds.)
Published by IOS Press, (c) IFIP, 2003, pp. 848-851
memorized and reconstructed. Also the locus of
control is directed to the critical parts of the form.
Figure 2: Client-Side Validation using JavaScript
Simultaneous Display of Error Messages (JAIO).
2.3 Client-Side Validation – Simult-
aneous Display of Error Messages.
Validation is performed on the client. The user is
presented with one pop-up-window displaying all
field names with wrong or missing entries at once.
Ideally, the correct syntax is repeated (Figure 2). This
puts high cognitive load on the user (memorization).
The problems with recall effects are well documented
.
2.4 Client-Side Validation using
JavaScript – Sequential Display of
Error Messages.
Validation takes place on the client. The user is
presented only with one incorrect field feedback at a
time. Even though the cognitive load is low due to
the sequential nature of the validation, the work
process is interrupted and attention is drawn from the
task at hand. Additional clicks and confirmations
have to be performed.
2.5 Client-Side Validation using
JavaScript’s OnBlur Event
Again, validation takes place on the client. The User
is presented with an error message immediately after
leaving a field she has filled out incorrectly. The
error message appears in a Pop-Up window. Users
notice the mistakes immediately after they happen.
This enables an instant correction and prevents a
tedious revision process later. However, this form of
validation is not well known on the World Wide
Web. The International Standard Organization
implicitly recommends this way of form validation
(ISO, 1996). The 9241 standard demands immediate
correction of erroneous entries.
3 Experimental Hypotheses
3.1 Precise Formulation of the Problem
The goal of the study is to compare the efficiency of
commonly used methods on the WWW to find the
most efficient way to implement form validation.
Four validation mechanisms are tested:
1) Server Side Validation (SV)
2) JavaScript Validation with simultaneous display
of errors (JAIO: JavaScript All in One)
3) JavaScript Validation with sequential display of
errors (JOAT: JavaScript One At a Time)
4) JavaScript Validation using the OnBlur Event
(JOBE: JavaScript with OnBlur Event).
When looking at efficiency, the following
questions are investigated: Which method leads to the
least number of mistakes? Which method allows the
most time efficient completion of forms?
3.2 Hypotheses
Two hypotheses are derived from the above stated
questions (Table 1).
Hypothesis 1
Validation methods SV and JOBE lead
to significantly less consecutive
errors than JAIO and JOAT.
Hypothesis 2
Validation methods SV and JOBE lead
to significantly shorter completion
times than JAIO and JOAT.
Table 1: Hypotheses
Reasoning Hypothesis 1
SV presents the user with a comprehensive
feedback on the errors made after the user has
completed the form. The entire form is displayed
again and the erroneous field are visually highlighted
and commented. Users can correct all mistakes
without further clicks.
JOBE on the other hand provides instant feedback
as long as the question to be revised is still in the
locus of control. When leaving the field, the system
reports errors in a pop-up window. This way, the
error can be corrected immediately and the process
can continue.
Due to the comprehensive and always visible
nature of feedback with SV and immediate feedback
with JOBE, the authors assume that these methods
will provoke fewer errors than JAIO and JOAT.
Reasoning Hypothesis 2
The reasoning for this hypothesis is derived from
Hypothesis 1: A greater number of Consecutive
Error leads to additional cognitive processing and
interaction processes (clicks and entries). This should
result in longer completion times for JAIO and
JOAT compared to SV and JOBE.
4 Methodology
4.1 Experimental Design
An online questionnaire is used for the study.
Officially, it is conducted as an „Online-Community
Study“. Only the last part of the questionnaire
actually concerns itself with validation methods. The
experiment consists of three parts: Community
Questionnaire (37 questions) and Demographic
Questionnaire as well as a Subjective Evaluation
Questionnaire (11 questions for subjective rating of
the form). The Demographic Questionnaire includes
15 straightforward questions (i.e. name, address etc.)
with five built in obstacles (see Table 2).
77 subjects are recruited for the experiment. They
are randomly placed in one of four groups, each of
which is presented with only one validation method:
1) Server Side Validation (SV): n=18
2) JavaScript All-In-One (JAIO): n=17
3) JavaScript One-At-A-Time (JOAT): n=19
4) JavaScript OnBlur-Event (JOBE): n=22
4.2 Obstacles
Key to the study are the obstacles built into the
demographic questionnaire:
Field
Obstacle
Name
Had to be entered in capital letters:
[FIRST NAME NAME].
Email
Had to include an «@»: [name@place.org].
ZIP/City
Had to be separated by an «/»: [9999/City].
Country
Had to be the two digit country code in
capital letters: [CH].
Date of
Birth
Day-Month-Year separated by «/» Day and
Month two digits, Year four digits:
[14/02/1977].
Favorite
Website
Had to start with «http://», had to include
the domain name and had to be completed
with an «/» at the end of the URL:
[http://www.place.org/] or
[http://place.org/].
Table 2: Obstacles in the demographic
questionnaire.
Of course, these validation criteria are not likely
to exist on a real web site – hopefully. The
unrealistic setting is taken into account, because the
goal of the study is to research classic error
correction and not the validation of good form
design.
4.3 Experimental Procedure
Testing is carried out in a usability lab during a
two week period in September 2002. The subjects
are first introduced to the setting and given
instructions. The error rate is recorded electronically
as are all relevant time measures. All demographic
data is extracted directly out of the completed forms.
5 Results
A variance analysis of the demographic factors shows
no significant differences between the experimental
groups’ gender distribution (F=8.93, p>0.45), age
distribution (F=.971, p>0.41), computer knowledge
(F=.352, p>0.78). Effects due to demographic
differences can therefore be excluded.
A significant evaluation of the error rate can only
be performed when categorizing the errors into two
subgroups as follows:
First Time Errors: The term «First Time Error»
refers to an error made the first time a field is filled
out by the participant. They occur due to the
obstacles built into the form. The participant has no
chance to prevent them. This rate should not differ
among the four experimental groups.
Consecutive Errors: The term «Consecutive
Error» refers to an error made after the form has been
validated once. This means that a user has been
presented with one of the feedback messages and still
enters the data incorrectly. The authors assume that
this rate differs significantly between the groups.
A variance analysis of the result confirms that
the First Time Error rate is not significantly different
between the groups (F=2.274, p>0.87).
The key question of the study is whether
Consecutive Errors depend on the validation method
used (Hypothesis 1). Table 3 states the results.
Method
M
SD
SV
2
1.88
JOAT
1.58
1.84
JAIO
6.12
3.12
JOBE
5.73
2.64
Table 3: Consecutive Error Rate
A variance analysis shows that there are
significant differences in the Consecutive Error rate.
(F=18.416, p<0.00). Post-Hoc analysis shows that
both SV and JOAT differ significantly from JAIO
and JOBE (p<0.00). Neither between SV and JOAT
(p>0.59), nor between JAIO and JOBE (p>0.62)
exists a significant difference.
These surprising results only partially support
Hypothesis 1. As expected, server side validation
(SV) leads to only few Consecutive Errors and
presenting all errors at once in a JavaScript Pop-up
window (JAIO) leads to many Consecutive Errors.
However, only presenting one error at a time
when submitting the form led to an even lower error
rate than SV (difference however is not significant).
This is even more surprising as we assumed that
clicking away the pop-up windows will interfere
with the revision process. This is not the case.
Most astonishing are the results for the
validation method using JavaScript’s OnBlur Event
(JOBE). Despite the fact that immediate error
correction was possible, test participants made a
great number of Consecutive Errors. A closer
inspection of the taped sessions reveals that many
participants were clicking away the error message but
did not correct the fields. They continued to fill out
the form and were presented with additional error
messages when trying to submit the form.
Participants were irritated by the fact of being
presented with an error message before even
submitting the form for validation. It appears that
users were not in «Revision Mode» and therefore
ignored error messages when completing a form.
Hypothesis 1 can only partially be confirmed.
Validation SV and JOAT lead to significantly fewer
Consecutive Errors than JAIO and JOBE.
Another important factor when choosing a
validation method is the time needed to complete a
form (Hypothesis 2). A faster completion time is
better economically, because it takes users less time
and the probability to successfully terminate the
interaction process is higher.
Method
M
SD
SV
119.29
49.85
JOAT
142.71
104.04
JAIO
140.87
47.78
JOBE
134.02
72.25
Table 4: Time needed to send off form (in seconds).
Due to the high variance in time, no significant
difference can be observed (F=.533, p>0.59). Also
the single comparison between SV and JOAT (the
methods that led to fewer mistakes) leads to no
significant difference (F=.951, p>0.34). Mean
differences can therefore only be regarded as
tendencies. There seems to be a trend towards SV as
the fastest mechanism.
Hypothesis 2 is rejected for statistical reasons.
6 Discussion
Choosing the right validation method can have a
severe impact on the success or failure of a data entry
process. The results confirm the very common
approach of server sided validation by highlighting
all erroneous fields to be user friendly.
6.1 Theory of Mental Modes
The poor results for the OnBlur-Event validation
could hint to a modal model for the mental states of
users filling out online forms.
At first, users are in «Completion Mode». Their
goal is to submit the requested information to the
system. The main focus is on filling in all the fields,
knowing that there might be mistakes. As soon as
users have finished completion, their mental mode
switches. With the appearance of the first error
message, users enter the «Revision Mode». Now,
error messages are acknowledged, interpreted and
necessary actions are taken to fix errors.
This modal theory is also supported by behavior
in the real world. The natural process of filling out
forms is also different from revision by third parties.
Habituation might play an important role in this
behavior. The advantage of end validation lays in the
fact that the user is not constantly interrupted. To
prove this theory, further studies are needed.
Also the surprising result that JavaScript
Validation with sequential error correction is almost
as good as server-side validation can be explained by
the modal theory. After completing the form, the
attention is explicitly directed to one specific field.
This way, a user is guided through the revision
process and hence produces fewer mistakes.
7 Further Work
Further studies are currently conducted to verify the
modal theory for form validation. Also, the
subjective satisfaction with these validation
mechanisms must be researched. While subjective
data was gathered in this experiment, the results have
not shown clear indication towards one or the other
method. It would also be interesting to research even
different mechanisms .
References
Goodman, D. (2001). JavaScript Bible (Gold edition
ed.). New York: Hungry Minds.
ISO-9241. (1996-2002). Ergonomic requirements for
office work with visual display terminals (Parts 1-
17), on www.iso.org
Wilde, E. (1999). Wilde’s WWW technical foundations
of the World Wide Web. Berlin [etc.]: Springer.