Effect of Product Type and Recommendation Approach on
Consumers’ Intention to Purchase Recommended Products
Yi-Cheng Ku
1
, Chih-Hung Chan
1
and Chin-Sheng Yang
2
1
Department of Computer Science and Information Management, Providence University, Taichung City, Taiwan, Taiwan
2
Department of Information Management, College of Informatics, Yuan Ze University, Chung-Li, Taiwan, Taiwan
Keywords: Recommender Systems, Product Type, Self-monitoring, Purchase Intention, Laboratory Experiment.
Abstract: Many e-stores offer product recommendation service to increase sales and customers’ satisfaction.
However, the performance of recommendation system will be influenced by the consumers’ judgments on
recommended products. The purpose of this study is to investigate the effect of product type and
recommendation approach on consumers’ intention to purchase recommended products. A laboratory
experiment was conducted to collect empirical data and ANCOVA was adopted to test the research
hypotheses. We found that there are significant interaction effects between recommendation approach and
recommended product type on consumers’ willingness to buy the recommended products and their
disconfirmation of the recommended products, while the consumers’ self-monitoring degree was a covariate
variable. Before experiencing the products, consumers’ willingness to buy search goods recommended by
the top-N approach is significantly higher than those recommended by the collaborative-filtering
recommendation approach. On the other hand, after experiencing the products, consumers’ disconfirmation
of the experience goods recommended by the top-N approach is also significantly higher than those
recommended by the collaborative-filtering recommendation approach. Besides, among the products
recommended by the top-N approach, the disconfirmation of experience goods is significantly higher than
that of search goods. The results of this study provide valuable implications for researchers and
practitioners.
1 INTRODUCTION
With the prevalence of the Internet, consumers begin
to explore the newly found field of e-shopping.
However, there is a wide variety of product
information on the Internet, and consumers have to
spend a lot of effort searching for the information
they need. In order to reduce their time and cost
spent on looking for information, e-stores offer
recommendation mechanism to help consumers to
filter out useful information (Zhang et al., 2011). For
example, Amazon.com set a precedent for adopting
service-oriented strategies by innovative
technologies, such as cross-selling and Today’s
recommendation, to recommend appropriate goods
to their customers based on their purchasing history
or browsing behavior. The techniques of
recommender system can be roughly separated into
personalized and non-personalized recommendation
techniques. For example, Top-N (i.e., Top N best-
selling product) is one of the non-personalized
recommendation techniques, and the recom-
mendation approach is based on the most popular
products. On the other hand, personalized
recommendation is based on each customer’s
personal interests. For example, collaborative-
filtering approach predicts a target consumer’s
preferences based on a group of consumers whose
interests are similar to the target consumer and then
recommend to him the products that he/she might
like.
The outputs of the recommendation system are
goods, but the properties of goods are different.
Physical goods can roughly be separated into search
goods and experience goods (Nelson, 1970). Search
goods mean that consumers are able to acquire
complete or most information of the product’s main
properties before buying it, while experience goods’
main properties can only be evaluated after that was
consumed or its information acquisition is much
more difficult than directly experiencing it. Since the
purpose of recommending products to consumers
with recommendation systems is to increase sales,
our research is aimed at exploring the influence of
475
Ku Y., Chan C. and Yang C..
Effect of Product Type and Recommendation Approach on Consumers’ Intention to Purchase Recommended Products.
DOI: 10.5220/0004442804750480
In Proceedings of the 15th International Conference on Enterprise Information Systems (ICEIS-2013), pages 475-480
ISBN: 978-989-8565-60-0
Copyright
c
2013 SCITEPRESS (Science and Technology Publications, Lda.)
recommendation systems and product types on
consumers’ willingness to buy a product. Besides,
consumers may have a kind of expectation before
buying a product, and there may be a gap between
their expectation and their perceived performance of
the product after purchasing it (Venkatesh and
Goyal, 2010), and this is called “disconfirmation” or
“experience gap” in this study. If consumers
experience significant disconfirmation after buying
the recommended product, it will affect their
satisfaction of the recommendation mechanism and
further influence their re-purchase willingness. As a
result, the other purpose of our research is to explore
the effect of recommendation system and product
type on consumers’ disconfirmation after the
consumers have experienced the recommended
product.
This paper is organized in the following way: we
are going to explore related research literature in the
next section. The research hypotheses and research
methods are proposed in Section 3, and the findings
and discussions are described in Section 4. We make
a brief conclusion and offer suggestions for future
research in the last section.
2 LITERATURE REVIEW
2.1 Recommendation System
The most widely used recommendation approaches
are non-personal recommendation, attribute-based
recommendation, item-to-item correlation, and
people-to-people correlation (Schafer et al., 1999).
Non-personal recommendation is mainly based on
the Top-N products in the current market, and it
does not put users’ personal interests into
consideration, so every user will receive the same
recommendation information. Attribute-based
recommendation, also known as content-based
approach, is based on the attributes of the products.
Users have to specifically “tell” the system their
needs so that similarity analysis of their needs and
products’ attributes will be learned and then the
products that meet users’ needs will be
recommended. Item-to-item correlation is based on
the correlation between different products. For
example, when a customer purchases coffee beans,
usually he will also buy some sugar and coffee-mate.
People-to-people correlation, also known as
collaborative filtering approach, is based on the
correlation between users, using the preference
similarity of similar users to find un-experienced
products or services that users might be interested in.
For consumers, the process of recommendation
systems is a “black box,” because they don’t
understand the internal algorithm of recommender
systems. As a result, consumers can only judge from
the recommended information when they are
evaluating a recommendation (Wang and Benbasat,
2007). Recommended information means all the
contents provided by e-stores when they recommend
a product to consumers, and it includes
recommendation messages generated by
recommendation system (e.g. the explanation of the
recommendation) and the messages generated from
non-recommendation system (e.g. peer reviews).
The recommended information is also the main
information that affects consumers’ buying decision.
2.2 Effect of Product Type on Online
Shopping
Products can be classified on the basis of their
features. For example, Nelson (1970) used product
features to separate products into search goods and
experience goods. Search goods mean that through
collecting information, consumers are able to find
out a product’s quality or features before buying it
(e.g. books), while experience goods’ main
properties can only be evaluated after buying or
consuming it (e.g. wine). The difference in product
types not only affects consumers’ message
processing but also influences their on-line shopping
behavior (Mudambi and Schuff, 2010). Hence,
different product types will lead to the difference in
the difficulty degree of evaluating an on-line
product’s quality.
Previous research found that consumers are more
willing to buy search goods than experience goods
on-line. For example, Chiang and Dholakia (2003)
did a survey and found that consumers are more
willing to buy books (search goods) than perfume
(experience goods) online. Gupta et al., (2004) also
found that consumers are more willing to buy books
and tickets (search goods) than wine and stereos
(experience goods) online. As a result, product type
is an important factor that affects consumers’
purchase decision (Im and Hars, 2001); (Wang and
Lin, 2003). Therefore, we also put the influence of
product type into consideration when we are
evaluating the effect of recommendation systems.
2.3 Expectancy-disconfirmation
Theory
Disconfirmation is the result of the comparison
between expectation and performance. Based on the
ICEIS2013-15thInternationalConferenceonEnterpriseInformationSystems
476
difference between expectation and performance,
there are three kinds of situation: when expectation
equals performance, there is no disconfirmation;
when expectation is smaller than performance, it is
positive disconfirmation; when expectation is greater
than performance, it’s negative disconfirmation
(Anderson, 1973). Before purchasing a product,
consumers will establish a comparison standard
based on past purchase experience. After the
purchase, they will compare the performance of the
product with the standard and generate positive or
negative disconfirmation, which affects their
satisfaction (Cadotte et al., 1987). If there is a gap
between a product’s performance and a customer’s
expectation, the customer will change his cognition
of the product and exaggerate the gap. As a result, if
the performance of a product is lower than
expectation, customers tend to give it an even lower
rating.
If there is a gap between product performance
and customers’ expectation, irrespective of the size
and direction of the gap, customers will lower their
ratings for the product and adopt a general negative
attitude toward it, thus reducing their satisfaction.
Only when product performance equals customers’
expectation customers will feel satisfied. As a result,
in the process of purchasing recommended products,
consumers will compare their expectation before
experiencing the product and perceived performance
after actually experiencing the product or service in
order to judge if there is difference between
expectation and performance. Therefore, we infer
that if a customer finds that the performance of the
product recommended by the system is very
different from his expectation, then his/her
satisfaction of the recommendation system will be
reduced.
3 RESEARCH METHODS
The purpose of this research is to understand
whether recommendation approach and product type
will affect consumers’ purchase intention and
experience gap when they are shopping online. This
study focuses on the comparison between Top-N and
personalized collaborative-filtering approach (CF)
when discussing the effect of recommendation
approach. As for exploring the effect of product type,
we aim at comparing search goods with experience
goods. When evaluating consumers’ willingness of
purchasing a certain product, besides estimating the
degree of their purchase intention after they read
recommendation information, we also evaluate their
disconfirmation after they actually used the product;
that it, we are going to evaluate the absolute value of
the difference in purchase intention before and after
they experience the product. The following sub-
sections will propose the hypotheses and
experimental design of this study.
3.1 Research Hypotheses
When consumers are choosing products, they will
search for related information based on their own
experience and outer environment. After the process
of comparing and judging, consumers’ purchase
behavior will be formed. Purchase intention means a
consumer’s subjective purchase tendency of a
certain product, and it has been proved to be an
important indicator that can predict purchase
behavior (Fishbein and Ajzen, 1975). Purchase
intention can be defined as the possibility of a
customer buying a certain product, and the higher
the purchase intension is, the greater the purchase
possibility becomes (Schiffman and Kanuk, 1999).
Previous research found that the number of
consumers who referred to recommendation was
twice more than that of those who didn’t refer to
recommendation (Senecal and Nantel, 2004), which
indicated that consumers’ purchase decision would
be affected by recommendation. Since personalized
recommendation approach is based on consumers’
personal interests, while non-personalized
recommendation approach is on the basis of people’s
common preferences, our research suggests that
consumers’ purchase intention of the products
recommended by CF is higher than that of the
products recommended by Top-N, and the products
recommended by CF can meet consumers’ needs
better than the products recommended by Top-N.
Therefore, we propose hypothesis H1 and H2:
H1: Consumers’ purchase intention of the products
recommended by CF is higher than that of the
products recommended by Top-N.
H2: Consumers’ disconfirmation of the products
recommended by CF is lower than that of the
products recommended by Top-N.
On the other hand, product type will also affect
consumers’ cognition of the recommended products,
because they are able to find out more about the
quality of search goods than experience goods
before purchasing them, and consumers are more
willing to buy search goods than experience goods
on-line (Chiang and Dholakia, 2003); (Gupta et al.,
2004). Thus, our research proposed hypothesis H3
and H4:
H3: Consumers’ purchase intention of search
EffectofProductTypeandRecommendationApproachonConsumers'IntentiontoPurchaseRecommendedProducts
477
goods recommended by recommendation
systems is higher than that of experience
goods.
H4: Consumers’ disconfirmation of search goods
recommended by recommendation systems is
lower than that of experience goods.
Since different recommendation approaches adopt
different recommendation basis, when
recommendation approaches recommend different
types of products, there might be different
recommendation effect. For example, Aggarwal and
Vaidyanathan (2005) found that consumers rate the
recommendation result that rule-based recommender
agents use on search goods higher than that of they
use on experience goods. Thus, our research
proposed hypothesis H5 and H6:
H5: Recommendation approach and product type
have significant interaction effect on
consumers’ purchase intention of
recommended products.
H6: Recommendation approach and product type
have significant interaction effect on
consumers’ disconfirmation of recommended
products.
The process of recommendation systems
recommending products to consumers can be viewed
as a process of persuasion. Since every consumer
has his own personality, the effect of the persuasion
made by recommendation systems on consumers’
cognition will be different. However, pervious
research didn’t put much emphasis on the influence
of consumers’ different personality on
recommendation effect. People generally have
different ability to adjust their behavior in order to
adapt to the changes in outer environment. The self-
monitoring concept which was investigated in
previous studies deals with the phenomena of
expressive controls (Snyder, 1974). People with high
self-monitoring will adjust their own behavior based
on the social cues gathered from other people and
social context, while low self-monitoring people’s
behavior is controlled by their own internal states
(i.e., beliefs, attitudes, and dispositions) instead of
environmental cues (Snyder, 1974).
High self-monitoring people are very sensitive to
the outer environment, and they will adjust their
behavior so as to adapt themselves to the external
expectation, while low self-monitoring people aren’t
concerned about outer situation and they tend to
maintain a consistent behavior model, ignoring the
needs of the situation (Snyder and Williams, 1982).
As a result, the differences in consumers’ self-
monitoring degree may affect their willingness to
purchase recommended products, because
recommendation is provided by recommender
systems (outer environment). Since consumers’ self-
monitoring degree may affect their purchase
intention of recommended products, consumers’
self-monitoring degree was adopted as the covariate
to explore its effect when we analyze the influence
of recommendation approach and product type on
consumers’ purchase intention and disconfirmation.
3.2 Experiment Design
The purpose of this research is to explore the effect
of recommendation approach and product type on
consumers’ purchase intention and disconfirmation.
Thus, a two-factor experiment was conducted to test
the hypotheses. Both recommendation approach and
product type are the independent variables.
Recommendation approach has two levels, i.e., Top-
N and CF. Product type also has two levels, i.e.,
experience goods (healthy drink) and search goods
(magazine). Purchase intention and disconfirmation
are the dependent variables while self-monitoring
degree is the covariate. Three items adapted from
previous studies were used to measure the purchase
intention (Dodds et al., 1991); (Grewal et al., 1998)
and 18-item proposed by Snyder and Gangestad
(1986) were used to measure the self-monitoring.
The voluntary subjects were recruited from a
medium university in Taiwan. In order to offer
recommended products to the subjects in accordance
with their preferences, all subjects had to fill out
questionnaires about their product preferences when
they registered the experimental activity. There were
108 subjects who joined the experiment.
The laboratory experiment was conducted in a
lab and the experimental process is as follows: Step
1: the subjects got a paperback questionnaire. They
logged in the experimental website and browsed the
experiment website after the instructor’s
introduction. Then the subjects were randomly
assigned to Top-N group or CF group by the
experimental website. Step 2: the experimental
website would adopt different recommendation
approach to offer recommended products for
subjects based on the group they belonged to, but the
number of products was fixed (e.g. three bottles of
healthy drink for experience goods and three
magazines for search goods). The recommendation
order was random in order to avoid order effect.
After reading the recommendation messages, the
subjects would choose the product he liked most and
put it into shopping cart to complete his shopping
mission. Step 3: subjects had to fill out their
purchase intention degree for each recommended
product after shopping, and the evaluation of
ICEIS2013-15thInternationalConferenceonEnterpriseInformationSystems
478
purchase intention was calculated by the average of
three items. Step 4: After completing the first part of
the questionnaire, subjects were arranged to try the
recommended products outside the experiment lab.
Step 5: After trying the recommended products,
subjects finished the second part of post-purchase
intention in the questionnaire. Step 6: Each subject
got US$3 as reward.
4 DISCUSSIONS
4.1 Sample and Descriptive Statistics
Among the 108 subjects that participated in our
experiment, there were more females than males. As
for educational background, college and graduate
students both accounted for a large proportion, 57%
and 42% respectively. The subjects’ Internet
experience was mostly in the range of 6 to 10 years.
As for daily Internet using time, it was mostly with 5
hours, and about half of the subjects had e-shopping
experience. Besides, the mode of the number of
magazines read by every subject per month was less
than 2, and the healthy drink drunk by every subject
per week was mostly less than 1.
The average and deviation of the two groups of
subjects’ purchase intention and disconfirmation of
the three search goods and three experience goods
recommended by the experiment system are shown
in Table 1 and Table 2.
The higher the grades are,
the stronger purchase intention and
disconfirmation are.
Table 1: The subjects’ average purchase intention.
Product
type
Recommendation approach
Top-N CF
Experience
goods
4.666
(0.912)
4.716
(1.029)
Search
goods
4.827
(1.073)
4.418
(0.934)
Note: mean (standard deviation)
Table 2: The subjects’ average disconfirmation.
Product
type
Recommendation approach
Top-N CF
Experience
goods
1.189
(0.696)
0.838
(0.644)
Search
goods
0.702
(0.465)
0.762
(0.61)
Note: mean (standard deviation)
4.2 ANCOVA Analyses
Two repeated measures ANCOVA were adopted to
test the hypotheses proposed by this study. The
results of first ANCOVA show that the interaction
effect of recommendation approach and product type
on the purchase intention is significant (p < 0.05),
but the covariate variable, self-monitoring, is not
significant. Hence, H5 is significantly supported.
Since the interaction effect is significant, the simple
main effect of recommendation approach was tested
at a given level of product type, and vice versa. As
Table 1 shows, the average purchased intention
toward experiences goods is higher than search
goods when the products were recommended by CF
approach, but the average purchased intention
toward search goods is higher than experience goods
when the products were recommended by Top-N
approach. However, the differences are non-
significant. On the other hand, the effect of
recommendation approach is significant in search
goods (p < 0.05), but not in experience goods. This
finding reveals that Top-N outperformed CF
approach when the search goods were recommended.
This result contradicts the hypothesis H1, so the
hypothesis H1 and H3 are not supported.
The results of second ANCOVA show that
subjects’ disconfirmation is affected by the
interaction of recommendation approach and product
type (p < 0.05), so the hypothesis H6 is significantly
supported. In addition, the self-monitoring variable
is significant (p<0.05). We conducted further simple
main effect analysis. As Table 2 shows, the subjects’
disconfirmation of the experience goods
recommended by Top-N approach is significantly
higher than that recommended by CF approach, but
the search goods is non-significant. Hence, H2 is
partially supported. Furthermore, consumers’
disconfirmation of the experience goods is
significantly higher than the search goods when the
products were recommended by Top-N approach.
The findings revealed that H4 is partially supported.
5 CONCLUSIONS
According to the experiment results of this study, the
purchase intention will not be affected by product
type when the recommendation approach is fixed.
However, before consumers consumed a product,
their purchase intention of the search goods
recommended by the Top-N recommendation
approach is significantly higher than that
recommended by CF approach. This result implies
that consumers in general have certain impression
and understanding of search goods, so they are able
to evaluate the products before actually using it, and
EffectofProductTypeandRecommendationApproachonConsumers'IntentiontoPurchaseRecommendedProducts
479
they tend to accept the products preferred by the
great majority instead of the products recommended
by the collaborative recommendation. Our research
suggests that online shops should offer Top-N
recommendation approach for search goods in order
to enhance consumers’ purchase intention.
In addition, the subjects’ disconfirmation of the
experience goods recommended by the Top-N
approach is significantly higher than that
recommended by CF approach. Our findings reveal
that the products recommended by the CF approach
have higher personalization degree than that of the
Top-N approach. On the other hand, the subjects’
disconfirmation of the experience goods is
significantly higher than search goods when the
products were recommended by the Top-N approach.
Therefore, this study suggests that online e-stores
should recommend experience goods by CF
approach in order to reduce consumers’
disconfirmation after purchasing the products.
As for the influence of consumers’ personality,
self-monitoring is not a significant covariate variable
when measuring consumers’ purchase intention.
However, self-monitoring is a significant covariate
variable when measuring consumers’ dis-
confirmation. Thus, our research suggests that
consumers’ personality can affect the performance
of recommendation systems, and we are going to do
further analysis and exploration of this issue.
ACKNOWLEDGEMENTS
This work was supported in part by the National
Science Council of the Republic of China under the
grant NSC 95-2416-H-126-014.
REFERENCES
Aggarwal, P., Vaidyanathan, R., 2005. Perceived
effectiveness of recommendation agent routines:
search vs. experience goods. International Journal of
Internet Marketing and Advertising, 2(1-2), pp.38-55.
Anderson, R. E., 1973. Consumer dissatisfaction: The
effect of disconfirmed expectancy on perceived
product performance. Journal of Marketing Research,
10 (1), pp.38-44.
Cadotte, E. R., Woodruff, R. B., Jenkins, R. L., 1987.
Expectations and norms in models of consumer
satisfaction. Journal of Marketing Research, 24(3),
pp.305-314.
Chiang, K. P., Dholakia, R. R., 2003. Factors driving
consumer intention to shop online: an empirical
investigation. Journal of Consumer Psychology, 13(1-2),
pp.177-183.
Dodds, W. B., Monroe, K. B., and Grewal, D. 1991.
Effects of price, brand, and store information on
buyers’ product evaluations. Journal of Marketing
Research, 28(3), pp.307-319.
Fishbein, M., Ajzen, I., 1975. Belief, attitude, intention
and behavior: an introduction to theory and research.
MA: Addison-Wesley.
Grewal, D., Krishnan, R., Baker, J., and Borin, N. 1998.
The effect of store name, brand name and price
discounts on consumers’ evaluations and purchase
intentions. Journal of Retailing, 74(3), pp. 331-352.
Gupta, B., Lai, F. C., Pal, D., Sarkar, J., and Yu, C. M.,
2004. Where to locate in a circular city? International
Journal of Industry Organization, 22(6), pp.759-782.
Im, I., Hars, A., 2001. Finding information just for you:
knowledge reuse using collaborative filtering systems.
In Proceedings of the Twenty-Second International
Conference on Information Systems, pp.349-360.
Mudambi, S. M., Schuff, D., 2010. What makes a helpful
online review? A study of customer reviews on
amazon.com. MIS Quarterly, 34(1), pp.185-200.
Nelson, P., 1970. Information and consumer behaviour.
Journal of Political Economy. 78(2), pp.311-329.
Schafer, J.B., Konstan, J., and Riedl, J., 1999.
Recommender systems in e-commerce. In Proceedings
of the first ACM conference on Electronic Commerce,
pp.158-166.
Schiffman, L. G., Kanuk, L. L., 1999. Consumer behavior,
7th Edition, Pearson Education.
Senecal, S., Nantel, J., 2004. The influence of online
product recommendations on consumers’ online
choices. Journal of Retailing, 80(2), pp.159-169.
Snyder M., 1974. Self-monitoring of expressive behavior.
Journal of Personality and Social Psychology, 30(4),
pp.526 – 537.
Snyder, R. A. and Williams, R. R., 1982. Self theory: an
integrative theory of work motivation. Journal of
Occupational Psychology, 55(4), pp.257-267.
Snyder, M., Gangestad, S., 1986. On the nature of self-
monitoring: matters of assessment, matters of validity.
Journal of Personality & Social Psychology
51(1), pp.
125-139.
Venkatesh, V., Goyal, S., 2010. Expectation
disconfirmation and technology adoption: polynomial
modeling and response surface analysis. MIS
Quarterly, 34(2), pp.281-303.
Wang J. C., Lin, J. P., 2003. Are personalization systems
really personal? – Effects of conformity in reducing
information overload, In Proceedings of the 36th
Hawaii International Conference on System Sciences.
Wang, W., Benbasat, I., 2007. Recommendation agents for
electronic commerce: effects of explanation facilities
on trusting beliefs. Journal of Management
Information Systems, 23(4), pp.217–246.
Zhang, T., Agarwal, R., and Lucas, H. C. Jr., 2011. The
value of IT-enabled retailer learning: personalized
product recommendations and customer store loyalty
in electronic markets. MIS Quarterly, 35(4), pp.859-
881.
ICEIS2013-15thInternationalConferenceonEnterpriseInformationSystems
480
