ArticlePDF Available

Toward an Understanding of Real Estate Homebuyer Internet Search Behavior: An Application of Ocular Tracking Technology

Authors:

Abstract

We track and record five measures of eye movements of current homebuyers who are in the process of searching for homes on the Internet. Total dwell time (how long a person looks at the photo), fixation duration (how long a person spends at each focal point), and saccade amplitude (the average distance between focal points) are all found to significantly explain a buyer’s overall opinion of the home and its value. A secondary finding is that the sections of the Webpage that are viewed first are the photo of the home, the quantitative description section, distantly followed by the real estate agent remarks section. Finally, charm pricing, the marketing technique where agents list properties at slightly less than round numbers, works in opposition to its intended effect. Given our result that homebuyers dwell significantly longer on the first home they see, and since charm pricing typically causes a property to appear towards the end of a search when sorted by price from low to high, we question the wisdom of using a charm pricing strategy.
JRER
Vol. 34
No. 2–2012
Toward an Understanding of Real Estate
Homebuyer Internet Search Behavior:
An Application of Ocular Tracking
Technology
Authors Michael J. Seiler, Poornima Madhavan, and
Molly Liechty
Abstract This paper examines the eye movements of homebuyers
searching for homes for sale on the Internet. Total dwell time
(how long a person looks at the photo), fixation duration (time
spent at each focal point), and saccade amplitude (average
distance between focal points) significantly explain someone’s
opinion of the home and its value. The sections that are viewed
first are the photo of the home, the quantitative description
section, distantly followed by the real estate agent remarks
section. Finally, charm pricing, the marketing technique where
agents list properties at slightly less than round numbers, works
in opposition to its intended effect. Given that homebuyers dwell
significantly longer on the first home they view, and since charm
pricing typically causes a property to appear towards the end of
a search when sorted by price from low to high, we question the
wisdom of using a charm pricing marketing strategy.
While it is widely recognized that homebuyers rely more and more on the Internet
to pre-search for homes,
1
very little research has examined how people search for
them on the Internet. With the tremendous wealth of information about listed
homes online, potential homebuyers must weigh the marginal search cost (time)
of looking through thousands of additional properties against the marginal benefit
of possibly finding the home that may be slightly better for them.
2
If the number
of available homes on the Internet was small, people would spend more time
looking at each property. But with so many homes available, it is generally
accepted that people spend very little time on a particular property. As such, it is
important to understand, through the use of quantifiable data, how homebuyers
seeking a new home search the net.
Researchers often refer to the difference between ‘stated preference’ and
‘revealed preference. A stated preference is what people say they prefer or say
they would do in a certain situation, whereas a revealed preference is what they
212
Seiler, Madhavan, and Liechty
actually do. Even a person who is completely honest might experience a
divergence between a stated and a revealed preference as many actions occur on
a sub-conscious level. Therefore, in order to accurately assess homebuyer
preferences, it is important to focus on physiological measures that go beyond
simple verbalizations from the user. The purpose of this study is to quantify
previously unknown data and thereby advance the field of residential real estate
brokerage. Specifically, ocular tracking technology is used to record the exact
scanning pattern employed by the homebuyer when searching the web for homes
for sale, the number of locations on the screen where they fixate, and the time
spent on each fixation point, which serve a s indicators of homebuyer revealed
preferences during the home search process.
Most behavioral experiments use a convenience sample of student participants.
This study makes an additional contribution in that both students and actual
homebuyers currently searching for a primary residence are examined. A
comparison of the sub-sample results provides a direct test of the extent to which
substitution should be deemed acceptable in behavioral experiments between
convenient and actual subject samples. The results vary in several key areas and
when possible, subject sample data should be directly collected.
In examining the sequence of each component of the webpage that is viewed, the
photo is overwhelmingly viewed first, followed by the property description
section, and lastly, the real estate agent open remarks section. Actual homebuyers
pay somewhat more attention to these later two sections when viewing the opening
page of the home tour than do students. This may be because students are not
actually in the market to buy a home.
Finally, this paper contributes to the literature on charm pricing, the marketing
technique where agents do not round off the listing price to the nearest $1,000 or
even $10,000. Instead, prices are listed at slightly lower than round numbers (e.g.,
$299,900 vs. $300,000). Rationally, this $100 should have almost no impact on
homebuyer opinion of value, yet Allen and Dare (2004) have shown it to be an
effective pricing strategy. In opposition to Allen and Dare (2004), the results of
the current study support the findings of Palmon, Smith, and Sopranzetti (2004)
in that charm pricing is found to work against the seller. When coupled with the
finding that homebuyers spend the greatest amount of time looking at the first
home that results from their search, the results do not support the use of charm
pricing strategies.
3
Literature Review
Eye movements are arguably the most frequent of all human movements. Large
scanning movements called saccades typically occur 34 times every second; the
search pattern followed by the eyes as they saccade from one point to another is
called the scanpath; and, the amount of time spent at each point is the fixation
time. Eye tracking as a methodology is based on Just and Carpenter’s (1976) ‘eye-
Real Estate Homebuyer Internet Search Behavior
213
JRER
Vol. 34
No. 2–2012
mind’ hypothesis: the location of a person’s gaze directly corresponds to the most
immediate thought in a person’s mind. In usability testing and advertising, eye
tracking is useful because it can be used to measure behavior that would be
difficult to obtain through other more overt measures (Karn, Ellis, and Juliano,
2000). Due to their close relation to attentional mechanisms, saccades can provide
insight into cognitive processes such as picture comprehension, memory, mental
imagery, and decision making. Thus, eye movement research has historically been
of great interest in the fields of neuroscience and psychiatry, as well as
ergonomics, advertising, and industrial design.
Eye tracking research that is most relevant to this study has been found in the
following areas: (1) reading print and online material, and (2) searching and
scanning webpages. Eye tracking studies of reading behaviors have yielded
interesting and useful findings about how people visually interact with documents.
Much of the usability-related research in this area centers on how variations in
textual and graphical presentation affect behavior (Brysbaert and Vitu, 1998).
When readers encounter cognitively complex material, the rate at which they read
slows considerably as shown by increases in eye fixation times and number of
regressions (backtracking) and decreases in saccade lengths (Liversedge, Paterson,
and Pickering, 1998). Furthermore, variations in type sizes affect the normal range
of eye fixation durations, saccade movements, and regressions (Tinker, 1963). In
studies of online reading, researchers have found that excessive use of color
4
decreased reading speeds by as much as 30% to 40% (Krull and Rubens, 1987),
although more recent studies have seen an increase in retrieval times due to
participants being more familiar with computers (Krull, Sundararajan, Sharp, and
Potts, 2004). As with reading, eye tracking research has provided insight into how
people scan and search for information online. This has been done in using two
methods: (1) in targeted-search studies, people are asked to identify specific
information or perform a predefined task, and (2) in free-scan studies, people are
asked to view a screen or series of screens without any predefined goal.
In general, it has been found that people prefer text over graphics as entry points
into websites (Boaz, Cuneo, Kreps, and Watson, 2002). Also, eye movements
roughly followed the ‘Z’ pattern of design. People’s eyes first travel to the upper
left corner (typically where the website identifying logo is placed), across the
page to the right corner, and then continue scanning the page in small ‘z’ patterns
progressing down the page (Goldberg, Stimson, Lewenstein, Scott, and
Wichansky, 2002). Interestingly, scanning did not end at the bottom right corner,
but instead continued up the right column of the page.
In addition, small type encouraged focused reading behavior while large type
promoted light scanning. When scanning, people are looking for words or phrases
that catch their attention. Images of at least 210 X 230 pixels received more eye
traffic than smaller images, and people frequently clicked on the 210 X 230 pixel
images. Interestingly, scanpath analysis revealed that users do not necessarily
follow the same scanpath for every type of website. Instead, there appear to be
universal scanpaths that people develop based on the function, genre, and design
of a website (Josephson and Holmes, 2002).
214
Seiler, Madhavan, and Liechty
Data and Design
To understand how homebuyers search for homes on the Internet, it is necessary
to perform tests on actual homebuyers. Twenty
5
currently searching or recent
(within the last three months) homebuyers from the local area who are in the
market for a home in the price range of $285,000$350,000 were invited to
participate in this study.
6
Twenty-five university students serve as the control
group.
7
These 45 individuals were shown 10 homes, each containing 6
photographs. The result is a cross-sectional dataset containing 450 (45 people
10 homes) completed home tours and a total analysis of 2,700 (450 6) photos.
8
To our knowledge, this constitutes the largest sample collected in a study using
an ocular tracking methodology.
Participants were seated in front of a computer equipped with a 17-inch CRT
monitor (optimal for ocular tracking due to its refresh rate). Below the monitor
was a desk-mounted, unobtrusive ocular tracking hardware/software (Eyelink
1000) device.
9
When using the device, the participant positions his chin on a
padded shelf with his forehead resting against a padded frame.
10
Once calibrated
for that particular participant, a PC-based remote camera then records all eye
movements. The task involves participants taking 10 different home tours on what
effectively is the Internet.
11
The website intentionally reflects the appearance of
those currently in existence. Specifically, the opening page has three primary
components. The first is an enlarged curb appeal photograph that listing agents
hope will catch the eye of the viewer. Underneath this large photo are five
thumbnail pictures of the remaining rooms in the home. The second section is
located in the upper right-hand portion of the page and includes property statistics
such as square footage, number of bedroom/bathrooms, and so forth. The third
and final section of the opening page shares real estate agent remarks, which is
an opportunity for the agent to help sell the property. Exhibit 1 shows the layout
of the website.
Once the opening page is viewed, the participant can search the six component
pictures of the home (curb appeal, main living area, kitchen, master bedroom,
master bathroom, and view/backyard). The participant is allowed an unlimited
time to view each picture and may move forward in the search at any time simply
by clicking a button. After the six photos of the first home are shown, a series of
short questions is asked, which allow for the measurement of the user’s ratings.
The data were analyzed via content analysis, a technique described later in this
paper. After answering the questions, the second home is shown, followed by a
few questions about that home, and so forth. After five homes have been viewed,
the participant is allowed to take a short break and then continue to complete the
second half of the tours. At the end of the visual portion of the experiment, the
participant is asked to share basic demographic data (such as gender, age, marital
status, ethnicity, income, etc.).
For validity of design, these general survey questions were alternated between the
beginning and the end of the visual portion of the experiment (to prevent such
Real Estate Homebuyer Internet Search Behavior
215
JRER
Vol. 34
No. 2–2012
Exhibit 1 Sample View of the Website Design with an Ocular Tracking Variable Analysis Superimposed
This exhibit shows the first page the participant sees. It includes an indication of the interest areas, as well as
various superimposed ocular tracking measures.
biases as cognitive anchoring). Other design mechanisms implemented to prevent
respondent bias include randomizing the order of the 10 homes that are shown
(to prevent order effects), as well as randomizing the order of the five pictures
beneath each home. Note that the curb appeal design mechanisms were put in
place to minimize the potential for introducing bias, while at the same time
keeping the search process as realistic as possible.
12
Because of the expensive
equipment used and the need to avoid outside distractions, it was necessary to
bring actual homebuyers to the university study site. For this reason, they were
compensated with a $50 gas card redeemable at a nearby gas station.
13
Student
participants were from psychology classes and were given experiment
participation credit, a practice very common in the field.
Methodology
Since no known study of this type has ever been conducted within a real estate
setting, the analysis begins with a series of univariate tests and design verification
procedures. Afterwards, a direct test of whether or not ocular tracking variables
can be used to predict the willingness of a homebuyer to pay for a home is
considered based on several variations of the following equation:
216
Seiler, Madhavan, and Liechty
Home Value ƒ{Home Characteristics, Qualitative Variables,
Ocular Tracking Variables, Demographic
Characteristics, Miscellaneous, error}. (1)
Where the specific variables are defined as follows:
Dependent Variables
% List Price
The participant’s opinion of fair market value
minus list price, divided by list price;
Market Value Opinion
Participant’s opinion of the fair market value of
the home;
Worth More Dummy
A dummy variable equal to 1 if the participant
estimates the fair market value of the home to be
greater than the listing price; 0 otherwise;
Worth Less Dummy
A dummy variable equal to 1 if the participant
estimates the fair market value of the home to be
less than the listing price; 0 otherwise; and
Overall Home Rating
Participant’s overall rating of the home on a scale
from 1 (not at all favorable) to 9 (extremely
favorable).
Home Characteristics
Actual Rating of the Rooms
Participant’s ex post rating of the specific room
(photo) for the curb appeal, kitchen, main l iving
area, master bedroom, master bathroom, and
backyard/view on a scale from 1 (worst ever) to
9 (best ever); and
Importance of the Rooms
Participant’s ex ante importance rating of the
specific room (photo) for the curb appeal, kitchen,
main living area, master bedroom, master
bathroom, and backyard/view on a scale from 1
(not at all important) to 9 (very important).
Qualitative Variables
Total # of Words
Total number of words the participant used to
describe the home after the tour;
# Positive Words
Number of positive words the participant used to
describe the home after the tour;
% Positive Words
Percentage of positive words (to total words) the
participant used to describe the home after the
tour;
# Negative Words
Number of negative words the participant used to
describe the home after the tour; and
Real Estate Homebuyer Internet Search Behavior
217
JRER
Vol. 34
No. 2–2012
% Negative Words Percentage of negative words (to total words) the
participant used to describe the home after the
tour.
Ocular Tracking Variables
Total Dwell Time Total time spent looking at each picture (in
milliseconds);
Fixation Duration Average time spent at each fixation point within
each picture (in seconds);
# of Fixations per Picture Number of points on the screen where the
participant’s eye stopped (i.e., fixated);
Saccade Count The number of times the participant’s eye jumped
from one fixation point to the other; and
Saccade Amplitude Average distance between fixation locations
(computer screen distance).
Demographic Characteristics
Gender Male
0; Female 1;
Age Age in years;
Income Annual income level (0
$0$20,000; 1
$20,001$40,000; 2 $40,001$60,000; 3
$60,001$80,000; 4 $80,001$100,000; 5
$100,001$120,000; 6 Over $120,000);
College Degree 1
College degree; 0 No college degree;
Homeowner 1
Has purchased a home before; 0 Otherwise;
Homes Purchased Number of homes purchased in lifetime;
Married 1
Married; 0 Single; and
White 1
White; 0 Non-White.
Miscellaneous
Consumption Motive 1
Consumption motive only; 0 Otherwise;
Investment Motive 1
Investment motive only; 0 Otherwise;
%ofSearchonWeb Home search time spent on the web (as a
percentage of total home search time);
Familiarity with Market Familiarity with the real estate market in the local
area on a scale from 1 (not at all familiar) to 9
(extremely familiar); and
Charm Pricing Dummy 1
Charm pricing was used in the list price; 0
Otherwise.
Dependent Variables
Agarwal (2007) and Seiler, Seiler, Lane, and Harrison (2012) document that
people are not proficient at identifying the true value of the home in which they
live. If this is the case, then asking them to accurately place a fair market value
on a home they have only visited through a virtual tour would seem more
218
Seiler, Madhavan, and Liechty
daunting. Still, in economics, finance, and real estate, the concern is with the
amount people are willing to pay for a home more so than just how a person rates
the overall quality of the home. While a measure of homebuyer willingness to
pay is preferred to a simple rating of the home, there are numerous difficulties
with collecting this data through a controlled experiment. For this reason, both
measures are collected and a series of regressions are performed accordingly.
Additionally, we further segment the analysis based on those who believe the
home they viewed is worth more (or less) than the listing price offered by the
selling agent via a series of logistic regressions.
Home Characteristics. Ex ante homebuyer stated preferences are collected in
terms of which rooms inside the home they believe to be the most important in
the final assessment of whether or not they will like the home. These importance
measures are then correlated with the ex post ratings of how much the homebuyer
liked the rooms, as well as the overall home rating. Real estate agents often use
the phrase ‘buyers are liars’ to describe the frustration of showing clients exactly
what they ask to see only to eventually have them buy a property whose attributes
do not even remotely resemble the original ‘must have’ list. While it is not the
central focus of the investigation to fully pursue this research question, we do
take a cursory glance at the issue.
Qualitative Variables. Buying a home can be a very difficult and time-consuming
process. Not only is the purchase likely the largest an individual will ever make,
but the selection of a home is one based on both economics and emotion. After
completing each tour, participants are asked to share the words that come to mind
when thinking about the home they just toured. The list of words may at times
seem to represent a stream of unconsciousness. However, using textural data
analysis techniques such as those in Li (2008), Dempsey, Harrison, Luchtenberg,
and Seiler (2012), it is possible to classify these words as being positive, negative,
or neutral. From this process, we are able to create and include in the quantitative
analysis a number of variables that were previously considered too qualitative.
Ocular Tracking Variables. One of the key contributions of this study is to quantify
variables that have previously been impossible to capture. Through the use of
modern technology, we are able to record with incredible precision five key ocular
tracking variables. The first is the total dwell time (in milliseconds) that each
participant spends looking at each of the 2,700 photos in the sample. Past studies
have shown that an increase in total dwell time is an indication of viewer interest.
Dwell time is hypothesized to have a positive association with interest in the
photo. A second captured measure is the number of fixation points (referred to as
fixation count) when viewing each of the photos. Fixation points represent
locations where the viewer has slowed down to take a closer look at a photo.
Fixation duration is the average time spent at each fixation point (typically in
milliseconds). A greater number of fixation points is hypothesized to be an
indicator of either greater interest or more effort expended in viewing the image
since evidence indicates that intense cognitive processing occurs during a fixation
(Rayner, 1998).
Real Estate Homebuyer Internet Search Behavior
219
JRER
Vol. 34
No. 2–2012
Saccade count is the number of times a participant’s eye jumped from one fixation
point to another. A related measure is saccade amplitude. This variable measures
the average distance between fixation locations. Although there are several
different interpretations of a saccade in the ocular tracking literature, it is generally
accepted that when a saccade occurs, information is being suppressed (Rayner,
1998). Thus, the greater the number and length (amplitude) of saccades, the less
time an individual was focusing on or actively interpreting the information in front
of them. This could mean that greater saccade counts and longer saccade
amplitudes imply: (1) a lower level of interest in the visual scene, or (2) a higher
level of familiarity with the task/context and low need to search very thoroughly.
Each of the five ocular tracking variables is collected for all 2,700 photos viewed.
Demographic Characteristics. Because people of all demographic profiles search
for homes on the Internet, it is not the focus of this study to examine the scanpath
differences by demographic characteristics. Still, the data does allow for the
inclusion of these as control variables. It is also interesting to see how the results
change when moving from actual homebuyers to a convenience sample containing
only students. As previously stated, the overwhelming majority of experiments
are conducted on student samples due to ease of access. The question is the degree
to which students provide a sufficient proxy for a sample of actual homebuyers.
Miscellaneous. Flachaire, Hollard, and Luchini (2003) used a novel approach to
measure a variable they termed ‘conformist versus non-conformist’ in a study of
a participant’s willingness to pay for admission to a park. The process involves
respondents answering a simple open-ended question: ‘When you think of (the
park), what words come to mind?’ Then, after a series of iterative steps similar
to a factor analysis, a dummy variable is formed that allows for a quantification
of a previously strictly qualitative concept. This approach is used to see whether
or not the manner in which a homebuyer views a home (as a consumption good
vs. an investment) influences a willingness to pay for the property.
The percentage of search time on the web is also collected with the notion that
more experienced searchers might view webpages differently than those who are
new to the process. For example, experienced searchers likely do not need nearly
as much time to view an Internet real estate listing because they have seen so
many already and presumably know what they are looking for. The next
independent variable is familiarity with the local residential real estate market.
Increased familiarity should also influence the search process in a manner similar
to experience on the web.
Charm pricing is an employed marketing technique whereby the list price of a
residence is set just below a round number (e.g., $299,900 vs. $300,000). While
several studies, such as Miller and Sklarz (1987), Kang and Gardner (1989),
Asabere, Huffman, and Mehdian, (1993), Knight, Sirmans, and Turnbull (1994),
and Benjamin and Chinloy (2000), have examined the relationship between list
price and sales price in a macro sense, only two studies have directly examined
charm pricing in residential real estate. Allen and Dare (2004) examined
220
Seiler, Madhavan, and Liechty
transactions in south Florida and found charm pricing in listed properties to lead
to higher transaction prices. In direct contrast to their study, Palmon, Smith, and
Sopranzetti (2004) found the opposite result when examining price behavior in
Texas. Because the literature remains sparse in the charm pricing area, this variable
is included as a secondary contribution of this study.
Results
Panel A of Exhibit 2 displays the total dwell times by the order in which the
home was shown. Recall that the home order was semi-randomized to prevent
order effects bias in the results. What is evident from the exhibit is that the first
home shown to participants took longer to view than those that follow. A learning
curve such as this is expected with any new experience. So while someone in the
market for a n ew home has likely searched for homes on the Internet before, this
is the first time they have seen this particular website. The curve levels off almost
immediately as the dwell time to view homes in positions 2 through 10 are rarely
significantly different from each other. Since most search engines for residential
real estate show results sorting prices from low to high, sellers might want to list
their properties at the very bottom of the pricing interval. This would cause their
home be shown first, which is associated with a longer dwell time by the potential
buyer.
14
Panel B of Exhibit 2 considers the order effects of rooms within the homes. Recall
that the curb appeal photo was always shown first to the participant as this is the
common industry practice. The photos shown beneath it randomly shuffle through
the remaining rooms, which is also consistent with both cyberspace and physical
tours of a home. ANOVA and post hoc tests reveal very few significant differences
in dwell times for any of the rooms, suggesting that participants maintained a
fairly consistent speed while scrolling through the website.
Summary statistics for the dependent variables in Equation (1) are listed in Panel
A of Exhibit 3. As expected, both homebuyers and students estimate the true
market value of the home to be less than the list price. This is consistent with
both observational intuition and cognitive anchoring hypotheses. Both groups
estimate prices to be over-stated by roughly 5%6%. Interestingly, homebuyers
report a higher market price for the homes, but a lower overall rating when
compared to students. This conflicting result is often argued to exist in studies
that are published in psychology versus real estate/finance journals and was the
impetus for collecting both metrics in this investigation. There are no statistically
significant differences in Panel A, however.
Panel B of Exhibit 3 displays the ex post or actual ratings for each photo of the
home followed by the ex ante stated level of importance when evaluating a
residence. While four of the six importance ratings are statistically significantly
different between the two groups of sample participants, only one (a different
room) was significant in terms of ex post evaluation. After the home was toured,
Real Estate Homebuyer Internet Search Behavior
221
JRER
Vol. 34
No. 2–2012
Exhibit 2 Total Dwell Time versus Order in Which Home and Room Were Shown
Overall
Curb
Appeal Kitchen
Main Living
Area
Master
Bedroom
Master
Bathroom
Backyard/
View
Panel A: Total dwell time by home
Home Pos 1 64,590*** 21,653 8,677 9,801** 9,701*** 8,996** 9,483**
Home Pos 2 56,963 20,311 6,872 8,086 7,454* 7,812 7,208**
Home Pos 3 57,909* 22,012 8,794 9,073** 8,984*** 8,245 8,274
Home Pos 4 55,982 19,203 8,402 7,508 8,449*** 7,238* 9,223*
Home Pos 5 56,231 19,757 7,418 8,692 8,001 8,576 8,127
Home Pos 6 55,369 20,234 7,451 7,709 7,706 7,676 6,974**
Home Pos 7 56,886 20,401 8,797 7,210* 7,532 7,955 8,635
Home Pos 8 50,437* 17,701 7,517 8,062 6,672*** 7,964 8,169
Home Pos 9 54,700 18,762 7,179 7,483 6,704*** 6,926** 8,804
Home Pos 10 55,827 20,445 8,075 6,977** 6,891*** 7,653 7,394*
ANOVA 1.30 0.42 1.35 1.44 2.32** 0.68 1.10
F-statistic 0.232 0.924 0.208 0.169 0.015 0.727 0.360
222
Seiler, Madhavan, and Liechty
Exhibit 2 (continued)
Total Dwell Time versus Order in Which Home and Room Were Shown
Overall
Curb
Appeal Kitchen
Main Living
Area
Master
Bedroom
Master
Bathroom
Backyard/
View
Panel B: Total dwell time by order room was shown
Rm. Order 1 20,052
Rm. Order 2 8,103 9,639* 7,650 9,680 9,300
Rm. Order 3 8,692* 7,551 7,784 7,877 8,193
Rm. Order 4 8,171 7,425* 8,176 7,697 8,849
Rm. Order 5 7,118* 7,374 7,425 7,401 8,280
Rm. Order 6 7,221 8,284 7,884 7,192 7,069
ANOVA 1.93 0.83 0.78 0.48 0.35
F-statistic 0.104 0.509 0.537 0.753 0.845
Notes: Panel A shows the dwell times for the homes that appear in the first through tenth positions. Panel B displays total dwell times for the order in which
the room was shown. The curb appeal photo was always shown first. The home order and all other rooms were randomized to avoid biasing the results.
Overall column significance levels are based on ANOVA tests. Significance indicators within specific cells are based on post hoc tests. Specific tests were
selected after a Levene statistic was computed in order to make the correct assumption regarding homogeneity of variance.
*Significant at the 10% level.
**Significant at the 5% level.
***Significant at the 1% level.
Real Estate Homebuyer Internet Search Behavior
223
JRER
Vol. 34
No. 2–2012
Exhibit 3 Descriptive Statistics by Sample Participants
Full Sample Home Seeker Students
Min. Max. Mean Min. Max. Mean Min. Max. Mean
Panel A: Dependent variable
% List Price 0.54 0.83 5.07% 0.42 0.83 4.22% 0.54 0.25 5.75%
Market Value Opinion 150,000 550,000 301,164 185,000 550,000 303,776 150,000 400,000 299,074
Worth More Dummy 010.20010.16010.22
Worth Less Dummy 010.56010.58010.55
Overall Home Rating 196.52196.45396.57
Panel B: Home characteristics
Actual Rating
Curb Appeal 196.46196.40296.50
Kitchen 196.26296.14196.36
Living Room 196.13*195.97296.26
M. Bedroom 195.95295.97295.94
M. Bathroom 196.16296.19296.14
Backyard /View 196.19196.26196.13
Importance
Curb Appeal 296.78*297.35396.32
Kitchen 397.79**397.62697.92
Living Room 497.69597.65497.72
M. Bedroom 497.60**497.45597.72
M. Bathroom 397.29497.35397.24
Backyard /View 297.24**297.05497.40
224
Seiler, Madhavan, and Liechty
Exhibit 3 (continued)
Descriptive Statistics by Sample Participants
Full Sample Home Seeker Students
Min. Max. Mean Min. Max. Mean Min. Max. Mean
Panel C: Qualitative variables
Total # of words 1 8 3.24** 1 8 3.39 1 7 3.12
# Positive Words 071.56071.61061.53
% Positive Words 0 1 48.07 0 1 45.53 0 1 50.10
# Negative Words 050.93040.88050.97
% Negative Words 0 1 30.89 0 1 28.26 0 1 32.99
Panel D: Ocular tracking variables
Total Dwell Time
Overall 6,082 168,749 56,488 27,720 168,749 57,164 6,082 141,379 55,945
Curb Appeal 5,658 94,300 20,052** 5,658 45,200 18,533 5,907 94,300 21,266
Kitchen 2,898 33,131 7,917*** 2,929 18,152 7,229 2,898 33,131 8,458
Living Room 1,685 45,098 8,058*** 1,685 34,942 7,362 2,728 45,098 8,612
M. Bedroom 1,361 42,189 7,817*** 2,451 28,323 6,891 1,361 42,189 8,550
M. Bathroom 2,963 42,535 7,902*** 2,963 34,569 6,861 3,240 42,535 8,733
Backyard /View 2,788 65,715 8,228*** 2,981 33,411 7,443 2,788 65,715 8,845
Fixation Duration
Overall 109 729 263.76 109 436 259.67 112 729 267.02
Curb Appeal 49 489 253.40 49 439 249.66 90 489 256.39
Kitchen 48 566 253.07 48 503 252.76 70 566 253.32
Living Room 51 553 259.93 72 553 259.11 51 485 260.58
M. Bedroom 61 582 251.57 61 582 250.42 62 474 252.48
Real Estate Homebuyer Internet Search Behavior
225
JRER
Vol. 34
No. 2–2012
Exhibit 3 (continued)
Descriptive Statistics by Sample Participants
Full Sample Home Seeker Students
Min. Max. Mean Min. Max. Mean Min. Max. Mean
Panel D: Ocular tracking variables
M. Bathroom 39 778 251.80 56 778 251.89 39 433 251.73
Backyard /View 60 533 253.53 63 533 253.82 60 502 253.30
# of Fixations per Picture
Overall 14 514 156.59*** 18 514 166.22 14 349 148.89
Curb Appeal 3 265 60.74 3 161 59.05 16 265 62.10
Kitchen 4 111 24.40** 4 64 22.83 6 111 25.64
Living Room 5 117 24.39* 7 104 22.91 5 117 25.56
M. Bedroom 2 119 24.05** 2 119 22.15 3 117 25.56
M. Bathroom 3 113 24.53*** 3 113 21.73 4 111 26.76
Backyard /View 3 163 24.77* 3 113 23.22 5 163 25.99
Saccade Count
Overall 2 513 154.55** 2 513 163.29 14 349 147.55
Curb Appeal 2 264 60.10 2 161 58.21 15 264 61.61
Kitchen 3 111 24.31** 3 64 22.65 6 111 25.61
Living Room 2 116 24.16** 2 104 22.49 4 116 25.49
M. Bedroom 2 120 23.94*** 3 120 21.98 2 117 58.48
M. Bathroom 3 113 24.40*** 3 113 21.54 4 111 26.66
Backyard /View 2 163 24.67* 2 113 23.06 5 163 25.94
Saccade Amplitude
Overall 2 11 3.13*** 2 11 3.37 2 5 2.94
Curb Appeal 1 11 2.78 1 11 2.75 1 6 2.80
Kitchen 1 11 2.98 1 11 2.98 2 7 2.97
226
Seiler, Madhavan, and Liechty
Exhibit 3 (continued)
Descriptive Statistics by Sample Participants
Full Sample Home Seeker Students
Min. Max. Mean Min. Max. Mean Min. Max. Mean
Panel D: Ocular tracking variables
Living Room 1 13 2.95 1 13 3.02 1 6 2.90
M. Bedroom 1 10 3.11 1 10 3.05 2 7 3.16
M. Bathroom 1 15 3.09 1 15 3.13 2 6 3.05
Backyard /View 2 21 2.93 1 21 2.98 1 6 2.90
Panel E: Demographic Variables
Males 0 1 40%*** 0 1 50% 0 1 32%
Age 18 54 29.60*** 22 54 37.35 18 43 23.40
Income 061.69***062.85050.76
College Degree 0136%***0175%014%
Homeowner Dummy 0 1 53%*** 0 1 90% 0 1 24%
Homes Purchased 081.24***082.40020.32
White Dummy 0 1 73%*** 0 1 85% 0 1 64%
Married Dummy 0 1 36%*** 0 1 55% 0 1 20%
Real Estate Homebuyer Internet Search Behavior
227
JRER
Vol. 34
No. 2–2012
Exhibit 3 (continued)
Descriptive Statistics by Sample Participants
Full Sample Home Seeker Students
Min. Max. Mean Min. Max. Mean Min. Max. Mean
Panel F: Miscellaneous Variables
Consumption Motive 0 1 62%*** 0 1 40% 0 1 80%
Investment Motive 0116%***0135%010%
%ofSearchonWeb 0 100 59%*** 2 100 71% 0 100 48%
Familiarity w /Market 194.22***195.10173.52
Charm Pricing 0 1 40.0% 0 1 40.0% 0 1 40.0%
Notes: Significance is based on independent sample t-tests. Specific tests were selected after a Levene statistic was computed in order to make the correct
assumption regarding homogeneity of variance.
*Significant at the 10% level.
**Significant at the 5% level.
***Significant at the 1% level.
228
Seiler, Madhavan, and Liechty
participants were asked what words come to mind when thinking of the home just
toured. Using content analysis (Berelson, 1952), these terms were coded based on
three concepts: positivity, negativity, and neutrality. Using responses to the survey
questions given to each participant in the beginning (or ending, for the other half
of the sample) of the study, each word was then further coded based on a relational
analysis, meaning that the underlining concept of each word coded was cross-
referenced with the participant’s verbal description of each property immediately
after it was viewed. Coding participant data in this manner is a typical method
for trying to understand the mental models participants construct (Carley and
Palmquist, 1992; Flachaire, Hollard, and Luchini, 2003). While homebuyers use
significantly more words to describe the home, both groups are remarkably similar
in both their number and percentage of positive and negative words.
Significant differences between homebuyers and students only begin to emerge
when considering the ocular tracking variables presented in Panel D of Exhibit 3.
Specifically, homebuyers take significantly less time to view each room when
compared to students. This result is fully anticipated given that they have already
been searching for homes and therefore better know what they are looking for
when perusing a website. A similar pattern of significant differences between the
homebuyers and students is seen in the number of fixations per picture and saccade
count. Consistent with expectations, experienced homebuyers fixated less and
demonstrated fewer saccades indicating that experience and familiarity with the
context compensated for the need to search thoroughly. However, this could also
be an indication of low interest in the website display (Rayner, 1998), which poses
an additional challenge when designing for the experienced buyer. Without the
use of ocular tracking technology, these challenges would never be known.
Understandably, the demographic profiles between homebuyers and students differ
considerably. Homebuyers in the sample are significantly older, earn more income,
have a higher level of education, have owned more homes in the past, and are
more likely to be married. They also have a significantly greater percentage of
search time on the web and a greater familiarity with the local residential real
estate market. Finally, homebuyers are more likely to consider the investment
component of a home when making a purchase. In fact, not one of the 25 students
associated strictly investment-related words when buying a home compared to
16% of homebuyers. Similarly, the consumption only motive was mentioned in
80% of student cases as opposed to just 62% of homebuyers. Clearly, a homebuyer
is thinking more in terms of both consumption and investment when searching
for a home.
In sum, the demographic profile between homebuyers and students is significantly
different across almost every measure, and while the groups do not seem to agree
on what they are looking for in a home (ex ante), the ex post results reported thus
far are strikingly similar with the exception of three of the ve ocular tracking
variables.
Exhibit 4 displays the sequence in which participants viewed the three sections
of each webpage. Panel A reveals the analysis associated with the curb appeal
Real Estate Homebuyer Internet Search Behavior
229
JRER
Vol. 34
No. 2–2012
Exhibit 4 Viewing Order for the Curb Appeal and Main Living Areas
Viewing Sequence Photo
Quantitative Property
Description Agent Remarks
Panel A: Curb Appeal Sequential Analysis
Viewed First 95.1% 2.4% 2.4%
Viewed Second 2.4% 75.6% 7.3%
Viewed Third 2.4% 7.3% 48.8%
Not Viewed 0.0% 14.6% 41.5%
Panel B: Main Living Area Sequential Analysis
Viewed First 87.8% 7.3% 4.9%
Viewed Second 12.2% 29.3% 2.4%
Viewed Third 0.0% 2.4% 4.9%
Not Viewed 0.0% 61.0% 87.8%
photo, which is always seen first by the participant. Overwhelmingly (95.1%),
participants first look at this picture. After focusing on the photos, participants
next turn to the quantitative property description section where the number of
bedrooms, bathrooms, square footage, etc. is displayed. Finally, the real estate
agent’s open remarks section is viewed. Very little emphasis is placed on this
section. In fact, over 40% of all participants (20% of homebuyers; 62% of
students) do not even look at the real estate agent remarks when viewing the first
page of the home tour. Participants comply by viewing the remainder of the home,
but in an actual setting, one has to wonder if the agent remarks would ever get
read if the home searcher does not like the initial photo of the home.
Panel B of Exhibit 4 reports the sequence of viewing for a later page in the home
tour. Recall that after the curb appeal, the remaining rooms are rotated to appear
in any order. As such, we sample the main living area and generate viewing
patterns for these rooms. Similar to the curb appeal, participants clearly focus on
the photo of the room as opposed to reading the quantitative property description,
and finally the agent’s remarks. This is understandable since these latter two areas
have been present on every page, whereas the photo is new on each subsequent
page.
When parsing the data, homebuyers are much more likely to view all three sections
of the webpage whereas students focus much more on the photos. This is
understandable as homebuyers are actually in the market to buy these properties,
and as such, need to collect as much information as possible to reach an informed
decision. Gender did not play a role in determining search sequence.
230
Seiler, Madhavan, and Liechty
Panel A of Exhibit 5 correlates dwell time with the ex ante stated preference of
what is important to participants. Correlation coefficients are either insignificantly
different from zero or significantly negative. The interpretation is that a person
does not have to look at a photo for a long time to gather from it the information
necessary to reach a conclusion. In fact, the opposite is true. People seem to be
more efficient at gathering information when the photo is of greater importance.
Panel B relates the ex post evaluations of the overall home rating to the individual
rating of each room. While all rooms are significantly positively correlated with
the overall rating, no room in particular stands out as being a better indicator of
the overall home evaluation. Finally, Panel C correlates the five ocular tracking
variables to decide which ones should not be included in the estimation of
Equation (1) due to multicollinearity concerns.
Exhibit 6 reports the results from six different variations of estimates to Equation
(1).
15
In the first three regressions, the dependent variable is the participant’s
estimate of the home’s value as a percentage of the list price. Recall that price is
a measure more palatable to real estate economists. The fourth through sixth
regressions have the overall home rating on a scale from 1 to 9 as the dependent
variable, which is a measure used more in psychology studies. Unreported
correlations indicate that multicollinearity concerns prevent the inclusion of all
the demographic variables. Therefore, in the first regression, a dummy variable
for participant type was used to capture differences between homebuyers and
students. Consistent with the results in Exhibit 3, homebuyers and students are
heterogeneous between groups, but extremely homogeneous within each group.
As such, only low correlated variables are included in each subsequent regression.
As hypothesized, the percentage of negative words is significantly negative in all
three regressions.
16
Total dwell time and fixation duration are significantly positive
and driven primarily by the student sample, while the statistically negative
coefficient on saccade amplitude is driven primarily by the homebuyer sample.
Homebuyers with a higher income perceive prices to be significantly lower than
lower income buyers, presumably because they are in the market for higher priced
(and therefore, nicer) homes. The same relationship is true for experienced
homebuyers. They appear not to be as easily impressed as less seasoned buyers.
Consistent with univariate tests, the negative coefficient on the participant type
dummy variable indicates that students perceive the value of homes within the
sample to be lower than the perception of value by homebuyers. Finally, the
significantly negative coefficient on charm pricing is consistent with Palmon,
Smith, and Sopranzetti (2004) in that the technique works in the opposite direction
as intended.
When comparing the results from the percentage of list price regressions to the
results from the overall home rating regressions, the dependent variable does seem
to make a difference. The sign changes for three of the variables (total dwell time,
consumption motive, and participant type). In two of the cases, the variable
becomes insignificant, but in the case of the consumption motive, the sign changes
direction and significance. While it is not possible to know which dependent
Real Estate Homebuyer Internet Search Behavior
231
JRER
Vol. 34
No. 2–2012
Exhibit 5 Selective Correlation Coefficients by Sample Participants
Full Sample (N 450) Home Seekers (N 200) Students (N 250)
Panel A: Importance and dwell time
Curb Appeal 0.004 0.046 0.073
Kitchen 0.085* 0.154** 0.076
Living Room 0.036 0.140** 0.021
M. Bedroom 0.197*** 0.282*** 0.188***
M. Bathroom 0.148*** 0.231*** 0.104
Backyard /View 0.147*** 0.218*** 0.135**
Panel B: Overall and room ratings
Curb Appeal Kitchen Living Room M. Bedroom M. Bathroom Backyard/View
Kitchen 0.401***
Living Room 0.408*** 0.595***
M. Bedroom 0.484*** 0.525*** 0.631***
M. Bathroom 0.493*** 0.490*** 0.547*** 0.611***
Backyard /View 0.295*** 0.369*** 0.440*** 0.450*** 0.384***
Overall 0.512*** 0.513*** 0.535*** 0.537*** 0.490*** 0.509***
232
Seiler, Madhavan, and Liechty
Exhibit 5 (continued)
Selective Correlation Coefficients by Sample Participants
Panel C: Ocular Tracking Variables: Curb Appeal
1
Total Dwell Time Fixation Duration # of Fixations Saccade Count
Fixation Duration 0.216***
# of Fixations 0.929*** 0.067
Saccade Count 0.922*** 0.059 0.996***
Saccade Amplitude 0.144*** 0.218*** 0.196*** 0.191***
Notes: Correlation coefficients are qualitatively similar no matter which room is considered.
*Significant at the 10% level.
**Significant at the 5% level.
***Significant at the 1% level.
Real Estate Homebuyer Internet Search Behavior
233
JRER
Vol. 34
No. 2–2012
Exhibit 6 Regression Results by Sample Participants
Percentage of List Price Overall Home Rating
Full Homebuyers Students Full Homebuyers Students
% Negative Words 0.094*** 0.103*** 0.079*** 1.545*** 1.842*** 1.235***
(0.017) (0.025) (0.022) (0.165) (0.251) (0.213)
Total Dwell Time 5.85
7
** 2.28
7
1.01
6
*** 4.29
6
9.93
6
** 1.99
8
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Fixation Duration 0.000** 0.000 0.000*** 0.001 0.000 0.001
(0.000) (0.000) (0.000) (0.001) (0.002) (0.001)
Saccade Amplitude 0.010* 0.014** 0.018 0.006 0.056 0.005
(0.006) (0.007) (0.013) (0.058) (0.069) (0.128)
Income 0.009** 0.021
(0.005) (0.046)
Homes Purchased 0.015*** 0.182***
(0.004) (0.044)
Consumption Motive 0.011 0.318*
(0.018) (0.182)
234
Seiler, Madhavan, and Liechty
Exhibit 6 (continued)
Regression Results by Sample Participants
Percentage of List Price Overall Home Rating
Full Homebuyers Students Full Hombuyers Students
Charm Pricing 0.112*** 0.074**
(0.025) (0.032)
Participant Type 0.017 0.163
(0.011) (0.111)
F-Statistic 10.575*** 6.114*** 8.596*** 13.314*** 9.179*** 7.724***
Adj. R-Squared 0.232 0.296 0.285 0.280 0.402 0.261
Notes: This exhibit reports the regression results from six separate regressions based on Equation 1 after removing independent variables that are too highly
correlated. The dependent variable is the participant’s opinion of value as a percentage of list price. Independent variables include the percentageofwords
used to describe the home, that are negative. Independent variables include the percentage of words used to describe the home which are negative; total
dwell time; fixation duration; saccade amplitude; participant income; total homes purchased in lifetime; a dummy variable for the consumption motive where
1 participants who expressed words for home ownership that only relate to the consumption component of owning a home, and 0 otherwise; a dummy
variable for charm pricing where 1 charm pricing was used, and 0 otherwise; and participant type where 0 homebuyers, and 1 students. The nine
(n 1) ‘home number’ fixed effects variables are suppressed for the sake of brevity, but are available from the authors upon request. For the full sample,
N 450; for homebuyers, N 200; for students, N 250.
*Significant at the 10% level.
**Significant at the 5% level.
***Significant at the 1% level.
Real Estate Homebuyer Internet Search Behavior
235
JRER
Vol. 34
No. 2–2012
Exhibit 7 Logistic Regression Analysis for the Dichotomous Dependent Variables Indicating Whether the Home is Worth More or Less than the List Price
Worth More Dummy Worth Less Dummy
Full Homebuyers Students Full Homebuyers Students
% Positive Words 1.292*** 0.156 2.399***
(0.410) (0.641) (0.613)
% Negative Words 1.582*** 2.144*** 1.270***
(0.355) (0.636) (0.469)
Total Dwell Time 0.000* 0.000 0.000*** 0.000 0.000 0.000
(0.000) (0.000) (0.002) (0.000) (0.000) (0.000)
Fixation Duration 0.005*** 0.001 0.007*** 0.006*** 0.001 0.008***
(0.002) (0.004) (0.002) (0.002) (0.004) (0.002)
Saccade Amplitude 0.150 0.226 0.156 0.050 0.153 0.252
(0.153) (0.204) (0.342) (0.116) (0.160) (0.275)
Income 0.048 0.008
(0.137) (0.106)
Homes Purchased 0.278* 0.459***
(0.162) (0.122)
Consumption Motive 0.735 0.928**
(0.521) (0.435)
Charm Pricing 2.435*** 1.921** 3.462*** 1.997*** 2.338*** 2.140***
(0.688) (0.916) (1.156) (0.496) (0.815) (0.686)
Participant Type 0.303 0.180
(0.278) (0.225)
236
Seiler, Madhavan, and Liechty
Exhibit 7 (continued)
Logistic Regression Analysis for the Dichotomous Dependent Variables Indicating Whether the Home is Worth More or Less than the List Price
Worth More Dummy Worth Less Dummy
Full Homebuyers Students Full Homebuyers Students
Overall Model X
2
89.60*** 37.53*** 79.45*** 107.65*** 71.41*** 75.53***
2 Log Likelihood 355.76 140.16 186.00 502.92 195.69 267.53
Cox & Snell R
2
0.182 0.174 0.273 0.215 0.305 0.262
Nagelkerke R
2
0.288 0.292 0.417 0.288 0.410 0.350
% Correctly Classified 81.6% 86.2% 82.3% 70.31% 77.0% 74.7%
Notes: This exhibit reports the regression results from six separate regressions based on Equation 1 after removing independent variables that are too highly
correlated. Independent variables include the percentage of words used to describe the home which are negative; total dwell time; fixation duration; saccade
amplitude; participant income; total homes purchased in lifetime; a dummy variable for the consumption motive where 1 participants who expressed
words for home ownership that only relate to the consumption component of owning a home, and 0 otherwise; a dummy variable for charm pricing where
1 charm pricing was used, and 0 otherwise; and participant type where 0 homebuyers, and 1 students. The dependent variable in the first three
regressions is a dummy variable where 1 the participant’s opinion of value is greater than the list price, 0 otherwise. The nine (n 1) home number’
fixed effects variables are suppressed for the sake of brevity, but are available from the authors upon request. For the full sample, N 450; for
homebuyers, N 200; for students, N 250.
*Significant at the 10% level.
**Significant at the 5% level.
***Significant at the 1% level.
Real Estate Homebuyer Internet Search Behavior
237
JRER
Vol. 34
No. 2–2012
variable is a better choice on which to base the analysis, it is simply our purpose
to identify the fact that the two measures, at times, yield significantly different
results. This is a topic we suggest future studies delve into further.
Exhibit 7 reports six separate logistic regression results from two variations of
Equation (1). Otherwise similar to the specifications in the prior exhibit, the new
dependent variable for the first three columns is a dichotomous variable, set equal
to one if the participant believes the home is worth more than the list price and
zero otherwise. In the last three logistic regressions, the dependent variable is set
equal to one if the participant believes the home is worth less than the list price.
For the full sample logistic regression in the first column, as expected, the
percentage of positive words carries a significantly positive coefficient as does
total dwell time and fixation duration. These results are clearly driven by the
student portion, where the signs and significance levels are consistent.
Alternatively, for the homebuyer sub-sample, only the numbers of homes
purchased in the past is significant.
In the three logistic regressions attempting to model the causes of participants
who believe a home is worth less than its list price, the percentage of negative
words used to describe it replaces the percentage of positive words. Consistent
with expectations, the coefficient is positive and significant. The signs for fixation
duration and saccade amplitude have switched when moving from the worth more
to worth less regressions, as expected. Fixation duration not only switched signs,
but also retained its statistical significance. This means that when a person likes
the home, they fixate for greater periods of time. Conversely, when they do not
like it, they experience shorter fixation periods. In both of the full sample logistic
regressions, the variable, participant type, is not significant.
Charm pricing results in Exhibit 7 are consistent with Exhibit 6. In the first three
columns, the charm pricing dummy carries a significantly negative signal,
indicating that its use resulted in fewer homebuyers evaluating the home as being
worth more than the list price. Consistently, in the last three columns, the
coefficients on charm pricing are all significantly negative, indicating a
symmetrical relationship. The conclusion is that charm pricing worked against its
intended purpose, consistent with the findings of Palmon, Smith, and Sopranzetti
(2004), but opposite the conclusions Allen and Dare (2004).
Conclusion
This study compared the results of actual homebuyers versus a convenience
sample of university student participants. While the demographic differences are
stark, the results typically are not. However, significant differences are observed
in a number of the ocular tracking variables: total dwell time, number of fixations,
and saccade count. We find that homebuyers and much more so students, focus
primarily on the photo of the home and secondarily on the quantitative description
of the property. Real estate agent remarks are substantially lower on the list of
238
Seiler, Madhavan, and Liechty
priorities. Just over 40% of participants did not view these remarks when looking
at the first page of the website. This speaks to the revealed importance of real
estate agent remarks to people searching for a home. In an examination of home
prices, charm pricing has the opposite effect as intended (a negative relationship
with price). The literature has been mixed on the success of this marketing
technique.
Since this is the first study to use ocular tracking technology in a real estate setting,
we focused on fundamental concepts and an exploratory understanding of how
people search for homes on the Internet. Future studies should extend the work
described here by testing the impact on various website designs. By performing
a similar analysis, after altering the layout of the page, real estate agents can begin
to learn which designs best grab the homebuyers’ attention. Moreover, in future
studies, it would be beneficial to simply observe homebuyers in their natural
element where they are free to continuously surf the Internet for homes without
having to pause periodically and answer questions about what they have just seen.
Still the initial steps taken in the current investigation are necessary groundwork
before more advanced investigations can be considered.
Endnotes
1
See Bond, Seiler, Seiler, and Blake (2000) and Benjamin, Chinloy, Jud, and Winkler
(2005) for the homebuyer market, as well as Hagen and Hansen (2010) for the rental
market.
2
See Gwin (2004).
3
The results from home searches typically yield properties listed in a price order from
lowest to highest. Since charm pricing results in a greater likelihood of being toward
the higher priced home order within the search range, it would seem that charm priced
homes would be seen much later by the homebuyer, who might be more prone to fatigue
and therefore, skim-reading.
4
The degree of color is typically quantified by dividing the portion of the page that is in
color by the total page size. The definition of ‘excessive’ varies from study to study.
5
The sample is split between 10 females and 10 males. Also, for married participants,
only one member (husband or wife) was allowed to participate to avoid duplication of
stated preferences.
6
This research was supported by three local real estate brokerage firms whose presidents
serve on the authors’ real estate center’s board. These firms kindly directed current
clients for inclusion consideration.
7
The overwhelming majority of these types of studies examine only students because it
is far easier to collect student data. We also collect student data simply to examine how
results vary between this convenience sample and those from the population of
homebuyers in whose behavior we are most interested.
8
All photographs used were from homes currently on the market in the Hampton Roads,
Virginia area. Original photographs were obtained directly from REIN, the company in
charge of posting photos for the local MLS.
Real Estate Homebuyer Internet Search Behavior
239
JRER
Vol. 34
No. 2–2012
9
Participants must not be color-blind and must have normal or corrected-to-normal vision,
since this is a primarily visual experiment with colored stimuli presented on a computer
screen.
10
Other than these two somewhat parallel bars, nothing stands between the participant and
the monitor.
11
Technically, the tours take place on a self-contained, originally created program that
allows the authors to completely control the process and avoid outside influences like
pop-up advertisements and connectivity loss. The result is a seamless, uninterrupted
home tour that allows tracking of ocular tracking-related variables of interest.
12
The entire process takes 3545 minutes to complete.
13
The institution who supported this research via a large grant did not allow the
distribution or handling of cash. The gas card was an acceptable alternative and thought
to be nearly as good an incentive as cash.
14
Clearly, this only makes sense if the market value of their home is extremely close to
the lower pricing level already.
15
Because each participant views 10 homes, nine (n 1) ‘home number’ dummy
variables are included in each regression to control for fixed effects.
16
When the percentage of positive words is included instead, the result is a consistently
positive and significant coefficient. As such, either variable (but not both due to their
significant negative correlation) can be used in the regressions.
References
Agarwal, S. The Impact of Homeowners’ Housing Wealth Estimation on Consumption and
Saving Decisions. Real Estate Economics, 2007, 35:2, 13554.
Allen, M. and W. Dare. The Effects of Charm Listing Prices on House Transaction Prices.
Real Estate Economics, 2004, 32:4, 695713.
Asabere, P., F. Huffman, and S. Mehdian. Mispricing and Optimal Time on the Market.
Journal of Real Estate Research, 1993, 8:1, 14956.
Benjamin, J. and P. Chinloy. Pricing, Exposure and Residential Listing Strategy. Journal
of Real Estate Research, 2000, 20:1/2, 6174.
Benjamin, J., P. Chinloy, G. Jud, and D. Winkler. Technology and Real Estate Brokerage
Firm Financial Performance. Journal of Real Estate Research, 2005, 27:4, 40926.
Berelson, B. Content Analysis in Communication Research. Chicago. IL: University of
Chicago, 1953.
Boaz, A., J. Cuneo, D. Kreps, and M. Watson. Using Eye Tracking Technology for Web
site Usability Analysis: The Application of ERICA to GEFANUC.COM. Proceedings of
the IEEE Systems and Information Design Symposium. Piscataway, NJ: Institute of
Electrical and Electronics Engineers, 2002, 15762.
Bond, M., M. Seiler, V. Seiler, and B. Blake. Uses of Websites For Effective Real Estate
Marketing. Journal of Real Estate Portfolio Management, 2000, 6:2, 20310.
Brysbaert, M. and F. Vitu. Word Skipping: Implications for Theories of Eye Movement
Control in Reading. In Eye Guidance in Reading and Scene Perception. G. Underwood
(ed.). Amsterdam, Netherlands: Elsevier, 1998, 12547.
Carley, K. and M. Palmquist. Extracting, Representing, and Analyzing Mental Models.
Social Forces, 1992, 70, 60136.
240
Seiler, Madhavan, and Liechty
Dempsey, S., D. Harrison, K. Luchtenberg, and M. Seiler. Financial Opacity and Firm
Performance: The Readability of REIT Annual Reports. Journal of Real Estate Finance
and Economics, 2012, forthcoming.
Doran, J., D. Peterson, and M. Price. Earnings Conference Call Content and Stock Return:
The Case of REITs. Journal of Real Estate Finance and Economics, 2011, forthcoming.
Flachaire, E., G. Hollard, and S. Luchini. A New Approach to Anchoring: Theory and
Empirical Evidence from a Contingent Valuation Survey. Working paper. University of
Paris, 2003.
Goldberg, J., M. Stimson, M. Lewenstein, N. Scott, and A. Wichansky. Eye Tracking in
Web Search Tasks: Design implications. Proceedings of the Eye Tracking Research and
Applications Symposium. NY, NY: Association of Computing Machinery Press, 2002, 29
36.
Gwin, C. International Comparison of Real Estate E-nformation on the Internet. Journal
of Real Estate Research, 2004, 26:1, 123.
Hagen, D. and J. Hansen. Rental Housing and the Natural Vacancy Rate. Journal of Real
Estate Research, 2010, 32:4, 41333.
Josephson, S. and M.E. Holmes. Visual Attention to Repeated Internet Images: Testing the
Scanpath Theory on the World Wide Web. Proceedings of the Eye Tracking Research and
Applications Symposium. NY, NY: Association of Computing Machinery Press, 2002,
43–9.
Just, M. and P. Carpenter. A Theory of Reading: From Eye Fixations to Comprehension.
Psychological Review, 1976, 87, 32954.
Kang, H. and M. Gardner. Selling Price and Marketing Time in the Residential Real Estate
Market. Journal of Real Estate Research, 1989, 4:1, 2135.
Karn, K., S. Ellis, and C. Juliano. The Hunt for Usability: Tracking Eye Movements,
SIGCHI Bulletin, 2000, November/December. http://www.acm.org/sigchi/bulletin/
2000.5/eye.html.
Knight, J., C.F. Sirmans, and G. Turnbull. List Price Information in Residential Appraisal
and Underwriting. Journal of Real Estate Research, 1998, 15:1, 5976.
Krull, R. and P. Rubens. Layout and Highlighting in On-line Information. In Empirical
Foundations of Information and Software Sciences IV. P. Zunde and J.C. Agrawal (eds.).
NY, NY: Plenum, 1987, 23744.
Krull, R., B. Sundararajan, M. Sharp, and L. Potts. User Eye Motion with a Handheld
Personal Digital Assistant. Proceedings of the International Professional Communication
Conference. Minneapolis: IEEE, 2004.
Li, F. Annual Report Readability, Current Earnings, and Earnings Persistence. Journal of
Accounting and Economics, 2008, 45:23, 22147.
Liversedge, S., K. Paterson, and M. Pickering. Eye Movements and Measures of Reading
Times. In Eye Guidance in Reading and Scene Perception. G. Underwood (ed.).
Amsterdam, Netherlands: Elsevier, 1998, 5575.
Miller, N. and M. Sklarz. Pricing Strategies and Residential Property Selling Prices. Journal
of Real Estate Research, 1987, 2:1, 3140.
Palmon, O., B. Smith, and B. Sopranzetti. Clustering in Real Estate Prices: Determinants
and Consequences. Journal of Real Estate Research, 2004, 26:2, 11536.
Rayner, K. Eye Movement and Information Processing: 20 Years of Research.
Psychological Bulletin, 1998, 124:3, 372422.
Real Estate Homebuyer Internet Search Behavior
241
JRER
Vol. 34
No. 2–2012
Seiler, M., V. Seiler, M. Lane, and D. Harrison. Familiarity Bias and Perceived Future
Home Price Movements. Journal of Behavioral Finance, 2012, forthcoming.
Tinker, M. The Legibility of Print. Ames, IA: Iowa State University, 1963.
We would like to thank Real Estate Information Network (REIN) for providing us with
the photographs used in this study. We would also like to thank the Old Dominion
University Research Foundation (ODURF) for their generous financial support of this
study.
Michael J. Seiler, Old Dominion University, Norfolk, VA 23529 or mseiler@odu.edu.
Poornima Madhavan, Old Dominion University, Norfolk, VA 23529 or pmadhava@
odu.edu.
Molly Liechty, Old Dominion University, Norfolk, VA 23529 or mdacliechty@
verizon.net.
... In terms of business, there is a benefit of mobile technologies adding a competitive advantage over other real estate firms (Mehmood, Zahoor and Ullah, 2019). The research on ocular tracking, is vital in understanding what people focused on when searching for a property online (Seiler, Madhavan and Liechty, 2011) and shows the advantages of providing multimedia resources on websites such videos and images as buyers tend to focus more on those resources. ...
... Interestingly, ocular tracking technology research that was conducted bySeiler et al. (2011), on the eye movement of property seekers as they were searching online; showed that property seekers focused first on an image (seeFigure 2.7) of the property before reading the information about it. The various categories the researchers examined included the dwell time (how long the users looked at a photo), and fixation time (time spent on each focal point)(Seiler, Madhavan and Liechty, 2011). ...
... Interestingly, ocular tracking technology research that was conducted bySeiler et al. (2011), on the eye movement of property seekers as they were searching online; showed that property seekers focused first on an image (seeFigure 2.7) of the property before reading the information about it. The various categories the researchers examined included the dwell time (how long the users looked at a photo), and fixation time (time spent on each focal point)(Seiler, Madhavan and Liechty, 2011). ...
... For Allen, Cadena et al. (2015) professional photos and 360° Tour make it easy for users to search for properties. Seiler, Madhavan et al. (2012) point out that during the search for the site, most of the users' time is spent visually browsing the properties. Photos provided by websites or real estate consultants about a property can be misleading. ...
Article
Full-text available
In early March, the newspapers reported the arrival of the Coronavirus in Brazil. The period of confinement directly affected our routines and transformed our consumption habits. Amidst the fear of contagion and uncertainty, people isolated themselves and postponed changes. Previous studies indicate that digital practices were accelerated after the onset of the COVID-19 pandemic, facilitating communication and improving the performance of various sectors of the economy. Thus, to better understand the behavior of Brazilian real estate market professionals in this period, the objective of this research was to verify whether the COVID-19 pandemic accelerated the process of perception of digitalization in real estate websites. To this end, five experts in the digital real estate market were interviewed, asked what are the main items that a quality real estate website should have to improve user experience. Based on the items selected by the experts, a questionnaire was developed and a question was constructed for each item. The questionnaire was sent online to 2144 realtors, asking about the importance of these items before and after the pandemic started. Using a 5-point Likert scale, and strict criteria for analysis, our survey yielded 423 valid responses. The results indicate that regardless of the groups analyzed, the hypotheses confirm that after the beginning of the COVID-19 pandemic, the perception of the importance of digital technology in Brazilian real estate websites was enhanced, increased. Managers took advantage of the pandemic and the availability of digital technology to offer specialized service to customers, thus creating a new competitive scenario. The tools associated with the visualization of the property and information such as ‘360°/Video Tour’, ‘Video Visit Broadcast’, ‘Chatbots’ and ‘FAQ List’ had the greatest increase in perception.
... To measure the estimated value of the home without the RTC, 1/3 of the participants were randomly assigned to this pool. Another 1/3 of the participants were randomly assigned to a second pool where the home tour was taken and at the end of which a short, purely legal description of the RTC 3 Studies that use experimental design in a real estate setting include Ikromov and Yavas (2012a,b), Seiler (2012), , , Seiler, Madhavan, and Liechty (2012), Seiler et al. (2008), and Yavas, Miceli, and Sirmans (2001).The aforementioned studies represent more direct applications of experimental design to applied research problems than studies outlined in well cited works in economics such as Gneezy and Potters (1997) and Harrison and List (2004). 4 Qualtrics software was chosen as the vehicle by which to represent the experiment to participants. ...
Article
This study uses a behavioral approach to measure the impact of a partial eminent domain land taking (which affects easement rights and causes proximity damage) on residential property values. Existing appraisal methods are not suited to handle this type of unique valuation impact determination, yet diminution must be determined by the courts to establish proper "just compensation". To accomplish the quantification of damages, an experimental design is implemented that allows the market to voice its opinion of valuation impact through a mechanism other than direct comparable sales. Moreover, this behavioral methodology can and should be used in a number of additional real estate valuation applications.
... In this work we propose and explore the usage of 'Eye-tracking' system to understand the decision making process in property valuations hence reducing the uncertainties and improving the corresponding AVMs. 'Eye-tracking' systems have been extensively used in fields such as psychology, marketing, human computer interaction and few instances in real estate (Seiler et al., 2012, Sun et al., 2021 but their use in improving urban economic models and property valuation is limited. Eye-tracking provides an excellent opportunity in developing a robust, objective and quant-itative method for capturing feedback in a co-design in addition to tradition qualitative methods. ...
Article
Full-text available
Building and using large scale, Automated Valuation Models (AVM) is one of the key multi-disciplinary pursuits in the study of cities and their economies. The methods used in building these AVMs such as ‘hedonic price modelling’ require a ‘co-design’ approach which needs significant collaboration and feedback between the modellers and the users of these models. The success of this collaborative approach depends crucially on our ability to capture the inputs and feedback from users without the bias and uncertainties present in traditional data collecting methods. In this paper, we explore and demonstrate the use of ‘eye-tracking’ technology in devising an objective methodology for collecting user feedback for co-design exercises. We employed a remote eye tracker in conjunction with traditional questionnaires to capture the decision making process of participants as buyers while selecting a property among a set of available options. We then compared the factors they reported to be important in their decision-making process to the factors they actually considered when looking at property listings. In our experiments, we found that pictures and maps captured more than 95% of the attention from buyers compared to the descriptive or statistical information showing the significance of the interface and medium of the valuation process. When responding to questionnaires, participants as property buyers reported that the attributes of a property such as number of beds, baths, quality of construction from pictures and location are equally important in selecting one over others. In contrast, when measured by an eye-tracker, we found that the participants gave significantly more attention to the quality of construction and location of the property compared to other factors. These preliminary results, though not definitive, demonstrate the value and usefulness of eye-tracking as a technique for capturing and measuring the factors that influence the desirability and in turn the price of a property. This methodology when controlled for characteristics of the participants, the properties and the medium of communication has the potential to help us to identifying and quantifying the relevance of parameters during property valuation and hence improve the accuracy and effectiveness of the corresponding hedonic price models.
... They provide suggestions based on their personal preferences and ratings for a specific film or series. Imagine if they did not need the previous viewing history to suggest what to watch next instead of relying on someone's behavioural data to infer his preferences [23,24]. Another example, Uber or any other cab aggregator has been utilising IoT to track drivers and passengers. ...
Article
Full-text available
Internet of Behaviours (IoB) aims to discuss how data are better understood and used to construct and promote new products from the view point of human psychology. The IoB can be used in a multitude of ways by public or private entities. This technology will become a compelling new marketing and distribution platform for companies and organisations worldwide. IoB platform allows developing an in-depth understanding of clients that every company needs. For example, IoB connects all cell phones in the app and can then see their errors and get visual recommendations on strengthening their swing and stroke. The interconnection of devices produces many new data points and extends from the Internet of things (IoT). Businesses collect information from customers through 'sharing' between connected devices, monitored through a single computer in real life. This paper is brief about the IoB and its role to fulfil customer requirements. This study also identifies and discusses IoB applications for better customer services. The aggregation of use and information from IoB devices gives helpful insight into consumers' comportments, desires, and tastes, something coined as a computer network. It encompasses many devices from phones to vehicles, exercise reloads, credit cards, to everything else literally connected to the Internet. Thus, The IoB's goal is to record, analyse, comprehend, and respond to all forms of human behaviour in a way that allows people to be tracked and interpreted using developing technology advancements and advances in machine learning algorithms.
... While the finance literature has experienced a recent surge on the topic of innovation, innovation is still an underexplored topic within the real estate literature. Although a few papers discuss innovation as it is related to real estate products (see Benjamin & Chinloy, 1995;Benjamin et al., 2002;Wallace, 2005;Seiler et al., 2012;Kuzina, 2015;Benefield et al., 2019, among others), the relation between local innovativeness and real estate prices or returns is yet to be explored. This paper seeks to address this gap in the literature. ...
Article
Full-text available
We examine the extent to which the quality of innovation created in different locations is related to subsequent changes in house prices in these metropolitan areas. Cities that foster a healthy quality of innovation are likely the home of many successful entrepreneurs and firms that provide high paying jobs. We hypothesize that the relation between innovation and changes in house prices is positive because, all else equal, locations with higher quality of innovation should not only be more desirable places to live, but also support higher rate of wealth and income growth, which allow for higher house price appreciation. We find results consistent with our hypothesis: there is a statistically and economically significant positive relation between innovation quality and subsequent house price appreciation. We find that this association runs from innovation quality to house price appreciation but not the reverse, and that the effect is stronger in areas with inelastic land supply.
... For example,Seiler et al. (2012b) and Seiler et al. (2012a) used an eye-tracking method to analyse exactly how buyers search for and take in information on the Internet.Arndt et al. (2013) also analysed the content of classied ads according to whether they were factual and veriable (e.g. new apartment) or not (e.g. ...
Thesis
Nous nous sommes intéressés dans cette thèse à étudier le viager immobilier sous plusieurs de ses aspects, notamment financiers. Cette thèse se divise en cinq chapitres, les quatre premiers étudiant le viager, et le dernier un aspect actuariel relatif à la longévité. Le premier chapitre se consacre à l’étude du comportement des acteurs du viager immobilier au travers d’interviews qualitatives. Le second étudie le processus de décision des vendeurs au travers d’un questionnaire. Le troisième consiste en une étude statistique de la préférence pour le bouquet ou la rente et des déterminants de la négociation, de l’extraction de richesse et de l’annuitisation. Le quatrième chapitre consiste en une étude de la dynamique spatiale du viager en France. Enfin, dans un dernier chapitre, les aspects actuariels de longévité sont abordés dont un modèle mathématique particulier pour de multiples populations.
... Most of the users' time is spent browsing through the various photos. Users will then review the basic description of the house (Seiler et al., 2012). By using a virtual tour, viewing the indoor space of the real estate can be possible without physically going to the actual place (Ching, 2018). ...
Article
Full-text available
Emerging the power of collective intelligence through crowdsourcing could create a clear path for visualizing real estate as well. One of the Crowdsourcing applications is describing the indoor space of a real estate. This paper aims to describe real estate in the context of spatial crowdsourcing. Qualitative and quantitative criteria were used in this study to describe the real estate space, topological relationships, directional relations, color, location, dimensions, and height as qualitative criteria. Quantitative criteria were selected as the dimensions and height. The proposed model was evaluated by two groups: those who had never seen the real estate and others that had already seen the same real estate. We implemented a website called SAMA 1 to evaluate the proposed model with crowdsourcing data using online collaborative tools. SAMA is using tools, such as a sketch plan, photo, text, virtual tour, and visual descriptions. To evaluate SAMA, we compared it with four representative commercial websites, and the impact of the tools was precisely examined. The obtained results demonstrate that the proposed model can be utilized for visual describing the indoor space of real estate in crowd-sourcing environments.
Article
Full-text available
The impact of color psychology on behavior has been clearly demonstrated for tangible products and advertising. However, the use of colors in combination has had little attention and has not been examined in the service context. This paper tests the small palette principal and the visual coherence perspective in an accommodation sharing setting. The studies are based on machine learning using large-scale data sets of US accommodation, with n = 1053 final images in study 1 and n = 1654 in study 2. Overall, the paper finds support for the small palette principal across the two studies. Study two tested the visual coherence perspective and found that significantly lower average distances between colors in perceptual space were found for listing images of properties with higher patronage. The implications are that in order to increase patronage, accommodation hosts should pay attention to the nature of color features of the listing images used for their properties.
Article
In this study, we use functional magnetic resonance imaging (fMRI) to understand how homeowners process non-financial information when considering strategic mortgage default. We find that borrowers initially attempt to inhibit their knee-jerk reaction to retaliate against a lender who has engaged in egregious lending practices when compared to a financially conservative lender. Moreover, when defaults are rare, borrowers are less likely to default because violating the social norm results in feelings of disgust. Finally, when a lender refuses a loan modification, the borrower is found to seek retribution. Interestingly, granting even a modest loan modification removes the desire of homeowners to seek retribution towards their lender no matter what their impression of the lender may be. The results carry a number of policy implications.
Article
Full-text available
Usability testing methods have not changed significantly since the origins of the practice. Usability studies typically address human performance at a readily observable task-level, including measures like time to complete a task, percentage of participants succeeding, type and number of errors, and subjective ratings of ease of use [3]. Certain types of questions are difficult to answer efficiently with these techniques. Imagine, for example, that we observe users spending longer than expected looking at a particular dialog of a software application or web page without making the appropriate selection to complete the task. Participants often have difficulty reporting their behavior and the experimenter is clueless about what went wrong. Is it because the user is overlooking the control? Is the user distracted by another element in the interface -- perhaps an animated graphic? Is the user seeing the control, but failing to comprehend its meaning? Different answers to these questions would clearly lead to different recommendations. If overlooking the control is a problem, increasing its salience is appropriate. If confusion of the control's function is a problem, changing the graphic or text label may be appropriate. If distraction is a problem, decreasing the salience of other stimuli may help. Without answers to these questions, design recommendations have to be implemented by trial and error. Recording the fixation pattern of the participant's eyes can offer additional information to help answer these questions. While this concept is not new, it has been confined primarily to military aircraft cockpit issues [2,4]. Only recently has eye tracking technology advanced to make it practical in the broader usability community. Usability studies of human-computer systems that have included eye tracking, e.g., [1] are beginning to show benefits of these techniques. However, important challenges remain.
Chapter
Sixty subjects were shown online information similar to commercially marketed microcomputer software. Information content was held constant across experimental groups, while the following layout features are manipulated: multicolor and monochrome highlighting; single and two-column arrangement of text; and short, medium and long text lines. Eye movement of some subjects was monitored by an eyecamera system; and keystroke errors were monitored for other subjects. Response speed was monitored for all subjects. Results show negligible performance gain for monochrome highlighting and response speed decrements for some kinds of multicolor highlighting. For example, multicolor highlighting disrupted eye movements characteristics of normal reading patterns. Columnar arrangement and line length had minimal effect on performance. Subject input errors were evenly balanced between syntactic errors and major logical errors. The factor most consistently producing gains in user performance was practice on successive trials. When repeated trials were combined with feedback to subjects about their errors, results show substantial reduction in syntactic input errors. The paper discusses implications of the findings for computer interface design.
Article
Executive Summary. The Internet is quickly becoming the place where customers first look to buy products and services. The real estate brokerage industry is no exception. Whether relocating across town or across the country, individuals are shopping more and more over the Internet. It is the fastest and least expensive information source available. As such, it is becoming increasingly more important that brokerage firms remain competitive by offering their properties on the Net. This study examines the efforts of residential real estate brokers to keep up with this dynamic environment by gathering listing information of existing real estate brokerage websites.
Chapter
This chapter illustrates the use of reading time measures. The time measures sum fixation durations to gain an understanding of the difficulty experienced while reading texts. A distinction is made between two approaches to sum the duration of fixations. The first approach is to sum the duration of fixations which neighbor each other in a specified region of space. The second approach is to sum fixations that occur in a sequence over a specified period of time. It is argued that both types of reading time measures are needed if the experimenter is to understand the time course of the influence of a linguistic variable on readers' processing of text. Primarily, a number of hypothetical eye movement records are discussed to illustrate the differential sensitivities of different eye movement measures. An eye movement experiment investigates how people process reduced relative clause sentences with and without the focus operator only to examine the utility of different reading time measures is recorded. The results show that measures summing temporally contiguous fixations can make an important contribution to the experimenters' understanding of the precise pattern of eye movements which occur when a problem is encountered in the text.
Article
When making decisions or talking to others, people use mental models of the world to evaluate choices and frame discussions. This article describes a methodology for representing mental models as maps, extracting these maps from texts, and analyzing and comparing these maps. The methodology employs a set of computer-based tools to analyze written and spoken texts. These tools support textual comparison both in terms of what concepts are present and in terms of what structures of information are present. The methodology supports both qualitative and quantitative comparisons of the resulting representations. This approach is illustrated using data drawn from a larger study of students learning to write where it is possible to compare mental models of the students with those of the instructor.
Article
Given its role as a tool for business and information sharing, efficient navigation of the Web has become imperative. In order to evaluate and improve a Web site, usability has emerged as an important field of study. In the business world, an intuitive Web interface mitigates the frustration of customers and arguably relates directly to increased sales. Despite this fact, many contemporary usability techniques remain subjective. Our team conducted an innovative analysis of the GE Fanuc Web Site. Using ERICA (Eye-Gaze Response Interface Computer Aid), we examined the paths test subjects used to accomplish tasks outlined in a testing protocol. GazeTracker™, ERICA's associated software, allowed the team to record and quantify user behavior patterns. Analysis of this data enabled us to make recommendations for design improvements to the content owners of gefanuc.com. GE Fanuc's relies on the Six Sigma design process, which requires quantitative data to ensure the quality of their products and services. Traditional usability techniques could not provide GE Fanuc with adequate quantitative data. Through this project, we used ERICA to collect this previously unavailable data. Our study found that the addition of ERICA to a usability study makes testing more precise and quantitative. We found it to be a valuable complement to traditional, qualitative usability analysis.