CONSUMER-TO-CONSUMER TRUST IN E-COMMERCE
Are there Rules for Writing Helpful Product Reviews
Georg Peters, Matthias Damm
University of Applied Sciences – München, Department of Computer Science and Mathematics
Lothstrasse 34, 80335 Munich, Germany
Richard Weber
Universidad de Chile, Departamento de Ingeniería Industrial
Republica 701, Santiago, Chile
Keywords: Consumer-to-Consumer Trust, Online Shopping, Product Reviews.
Abstract: Since the emergence of the Internet online shopping has rapidly grown and replaced parts of traditional
face-to-face shopping in real shops in cities and shopping centres. The role of the sales assistant has been
supplemented or even taken over by online information like buyers guides, product reviews or product
related discussion groups. For example, Amazon offers its customers the possibility to write product
reviews which will be published on the product site. However, a potential buyer is confronted with a similar
problem as in physical shops: Can I trust the recommendation of the sales assistant in a physical shop
respectively can I trust the recommendations - the product reviews of former buyers published by the
Internet shop. So, at Amazon's readers of the product reviews can classify a review as helpful or not. In our
paper we analyse if there are relationships between the formal structure of a product review and the degree
readers classify a review as helpful. We present the results of a case study on the Germany's Amazon shop
and derive "writing rule" for good product reviews out of our analysis.
1 INTRODUCTION
Amazon is regarded as a pioneer of the Internet
revolution. It was founded as a Internet based
bookshop in 1994. Nowadays it has an extensive
product range covering kitchen ware, watches, CD
& DVD, computer hard- as well as software, cloth
besides many others products. Amazon also
functions as Internet-based shop platform for third-
party retailers and manages, as neutral body, the
payment process between customers and retailers.
In the year 2006 Amazon generated revenues of
USD 10.71 billions which makes it the world's
leading Internet retailer and shopping platform
(Flynn 2007).
On its shopping sites Amazon provides its
customers a C2C communication platform where
product reviews can be exchanged. The product
reviews of mostly former buyers shall help potential
buyers to make their decisions which product to
choose. So these reviews may influence the level of
trust a potential buyer feels for a product.
However, since the reviewers are normally
anonymous or unfamiliar to the potential buyer
she/he faces another challenge: Can I trust the
reviewers and their comments on a certain product.
In this context Amazon provides the possibility
to rate a product review as helpful. So product
reviews that are mostly considered as helpful can be
regarded as good reviews.
The objective of the paper is to investigate if
writing rule for good product reviews, product
reviews that are considered as helpful by readers,
can be determined.
The paper is organized as follows. In Section 2
we give a short overview on relevant foundations of
trust. In Section 3 we describe our experimental
setup and limitations of the study. In the following
Section we present the result of our analysis. The
paper concludes with a summary in Section 5.
61
Peters G., Damm M. and Weber R. (2008).
CONSUMER-TO-CONSUMER TRUST IN E-COMMERCE - Are there Rules for Writing Helpful Product Reviews.
In Proceedings of the Tenth International Conference on Enterprise Information Systems - SAIC, pages 61-66
DOI: 10.5220/0001675200610066
Copyright
c
 SciTePress

2 TRUST IN E-COMMERCE
Besides the reputation of a product or a company
and reviews in neutral consumer magazines, like
Stiftung Warentest in Germany or Which in the UK,
recommendations by former buyers influence the
decision which product will be bought by a
consumer.
The non-commercial communication between
consumers about goods and services is known as
word of mouth or electronic word of mouth if it is
primarily electronically based (Arndt 1967,
Westbrook, 1987).
Word of mouth plays an important role in the
decision process of a potential buyer (Duhan 1997,
Bruner 2007). It has a higher credibility for a
customer than information provided by the
producing company directly Blickart et al., 2001.
Customers regard word of mouth reliable and
serious (Pollach, 2006). So it is seems not surprising
that Chevalier et al. (2006) reported that a positive
word to mouth on a product can have a direct
influence on its turnover.
Principally electronic word to mouth
communities can be classified according to the entity
that controls the platform (Henning-Thurau, 2004):
 Producer controlled. These are, for example,
chat sites on producer websites or closely
affiliated partners.
 Customer controlled. Examples are fan sites or
the opposite, boycott sites.
 Independently controlled. These sites provide
neutral services. This means they do not have
any connections to the producers nor they are
organised by customers. Examples are Ciao!
(www.ciao.de/) or dooyoo (www.dooyoo.de/).
Obviously trust between customers exchanging
opinions about goods and services plays an
important role in such electronic communities.
According to Henning-Thurau et al. (2002) different
forms of trust on electronic word to mouth platforms
can be defined. They are as follows:
 Judgement and emotion based trust. Based on
her/his judgment or emotion the consumer
decides which content she/he trusts.
 Incentive based trust. The consumer assumes
that providing wrong or misleading information
is not profitable for the information source.
Therefore she/he trusts every information
 Sanction based trust. The consumer is the
opinion that wrong or misleading information
will be disclosed by other consumers.
 Network based trust. A consumer regards
information as trustable when other consumers
trust the information also.
 Control based trust. The consumer trusts the
organiser of the platform and her/his control
function.
 Basic trust. The consumer trusts every
information she/he gets.
In the contex of our analysis network based trust
is of particular interest. A high degree of helful
classifications shows that many people in the
network trust a particular product review. Therefore,
the chance that new readers also trust the review is
higher than in the case of a low helpful indicator.
Please note that there are two overlapping effects
to be consider. (1) The product review might be
good and will be classified as helpful. (2) However,
an average product review might be consider also as
mostly helpful when the helpful indicator is high
(network based trust), for example due to
manipulation by the writer of the review.
3 EXPERIMENTAL SETUP AND
LIMITATIONS OF THE STUDY
3.1 Experimental Setup
Amazon offers a service for customers to write
product reviews that will be published on its
websites along with the product information. As
discussed in the previous Section these product
reviews can be an important information source for
potential buyers since they should ideally reflect real
and unbiased customer experience. An example of a
product review is given in Figure 1.
However, obviously these product reviews vary
in their qualities. An indicator for the quality of a
product review is also provide by Amazon.
Customers can classify product reviews as helpful or
not (see Figure 1).
Therefore the main objective of our research is to
analyse if there are rules how to write a helpful
product review.
ICEIS 2008 - International Conference on Enterprise Information Systems
62

Figure 1: A Product Review.
Amazon offers a very wide product range.
Therefore we had to limit our research to a few
product categories and arbitrarily selected the
following ones.
Product Categories:
 Business and other specialized books.
 Digital cameras.
 DVD (movies etc.).
 Mobile phones.
 Music (CD).
We analysed the following features of the
product reviews and if they influence the judgement
on the "helpfulness" of the reviews:
Text Features:
 Number of review stars.
 Usage of the personal pronouns.
 Length of the product reviews.
 Formal structure of the product reviews (list
character style).
 Readability index of the product review.
In total we analysed up to 60,000 product
reviews of the Amazon's online shopping platform in
Germany. If not explicitly mentioned all results
presented in Section 4 are p-significant.
3.2 Limitations of the Study
Manipulated Product Reviews. Product reviews as
well as the classification of a product reviews as
helpful can be conducted anonymously. Therefore
product reviews can be manipulated and bias.
Product reviews written by consumers are
regarded as important for the success of a product
(Chevalier et al. 2006). Therefore companies
producing or selling a product have strong interests
in good reviews on their products and vice versa bad
reviews on products of competitors. So they might
think of placing manipulated product reviews while
not disclosing their relationship to and interest in the
product.
Customer activists may also have an interest to
denigrate without disclosing their real intentions and
affiliations.
The same applies to the helpful classification of a
product review. A possible manipulation would be
to classify positive reviews as helpful. According to
our discussion on network based trust, a reader could
be influenced by a high helpful indicator. She/he
then would consider the positive product review as
more helpful than neutral and negative ones.
While Merschmann (2007) addressed the
possibility of manipulated product reviews the
authors are not aware of any study regarding
manipulated helpful indictors. However the later
cannot be excluded and therefore may fudge the
results of our analysis.
Since the reviews and in particular the
judgements on the helpfulness of the reviews are
anonymous the authors of the study had no
possibility to support their arguments by, e.g. semi-
structured interviews with customers to disclose and
better understand their motivations. Therefore any
interpretations given in this study can only give
limited evidence of the real structures.
Amazon's Function as Editorial Office. Amazon
functions as editorial office. Before a product review
will be published on its websites Amazon checks the
review. The authors do not have any information
about the Amazon's editorial process, for example,
how they select product reviews for publications and
how they may shorten or change them.
So the published product reviews may not be
representative and therefore also could distort the
results of our analysis.
Interdependencies between the features. Finally,
the possible interdependencies between the features
were not subject of this study and will be addressed
in further research.
CONSUMER-TO-CONSUMER TRUST IN E-COMMERCE - Are there Rules for Writing Helpful Product Reviews
63

4 RESULTS OF THE ANALYSIS
4.1 Amazon's Star System
As depicted in Figure 1 product reviews consist of
text and summarizing stars. The star system ranges
form one star for poor to five stars for excellent
products.
Taken the stars as indicators an average product
review is rather positive than negative. In every
examined product category five stars make up for
more than 50% of the evaluations (see Table 1).
Table 1: Distribution of Stars Reviews [%].
Category
* ** *** **** *****
Books 9.19 6.03 7.21 16.84 60.73
Cameras 4.63 4.92 8.26 22.90 59.30
DVD 12.88 7.77 10.20 17.38 51.78
Mobiles 7.73 7.17 10.03 24.90 50.13
Music 7.00 7.59 11.14 22.27 67.36
There is also a trend that positive product
reviews are regarded as more helpful for the readers
than negative (see
Table 2). A reasons for this could
be that readers are seeking for support for their
decisions to buy a product. Therefore they might
find product reviews more helpful that are positive
than negative ones.
Table 2: Helpful Reviews [%].
Category
* ** *** **** *****
Books 59 70 76 78 78
Cameras 47 62 66 85 62
DVD 43 40 46 63 63
Mobiles 41 53 61 79 74
Music 36 42 48 63 70
Furthermore balanced but positive reviews,
having four stars, seemed to be more helpful than
extreme ones. A reason for that could be that
balanced but positive reviews might be considered
as more neutral and unbiased than (extreme) five
stars product reviews.
4.2 Usage of the Personal Pronouns
The use of personal pronouns like "ich", "mich",
"wir" or "unsere" ("I", "me", "we"" of "ours") and
their implications on the helpfulness of the product
reviews are investigated in this Section.
The results are depicted in
Table 3. Obviously in
longer product reviews there is a greater chance of
the occurrence of a high number of personal
pronouns. Therefore, to diminish the influence of the
length of a product review (see Section 4.3), only
reviews with one to three personal pronouns are
considered and compared to reviews containing no
personal pronouns. The product reviews in the
categories DVD and music are potentially (even)
more subjective than in the other categories.
Categories like cameras and mobiles are more driven
by technical facts than by personal favours as in the
categories DVD and music. The observed results
support the above argumentation.
Table 3: Usage of Personal Pronouns [%].
Category No Personal
Pronouns
1-3 Personal
Pronouns
Books 72 75
Cameras 70 78
DVD 52 53
Mobile phones 58 66
Music 58 58
For the categories DVD and music it seems to be
irrelevant for the helpfulness of a product review if
personal pronouns are used or not. The readers of
these reviews may consider them as subjective
anyhow whether personal pronouns are used or not.
Surprisingly, in the remaining categories
subjectively written product reviews seemed to be
more helpful for the readers than reviews that are
written without the usage of personal pronouns.
4.3 Length of Product Reviews
The average length of the product reviews varies
between 1059 letters for digital cameras and 805
letters for books (see Table 4).
Table 4: Average Length of Reviews.
Category Average Text Length
Books 870
Digital cameras 1059
DVD 1045
Mobile phones 1052
Music 805
For our analysis we took more than 35,000
product reviews that were assessed by at least five
customers and calculated the correlation between the
length of a product review and the support
(percentage of reviews who regarded the review as
helpful) that the review has got. The results are
presented in
Table 5.
ICEIS 2008 - International Conference on Enterprise Information Systems
64

Table 5: Correlation - Length of Text vs. Support.
Category Correlation
Books 0.196
Digital cameras 0.395
DVD 0.265
Mobile phones 0.387
Music 0.267
Although the correlations are not very strong
longer product reviews seem to be regarded as more
helpful by the customers than shorter ones.
Surprisingly the book category has the lowest
correlation of the analysed categories. In contrast to
that technical equipments, like digital cameras and
mobile phones, have the highest correlations. Music
and DVD have almost similar correlations and are
place in between technical equipment and the books.
A possible explanation for the observed results
might be that technical equipment is more complex
and needs more information in comparison to
"simpler" products like books. Music and DVD are
related product categories which is also reflected in
very similar correlations.
4.4 List Character Style
In this Section we investigate whether list character
styled reviews or reviews written in prose are more
helpful for customers.
We defined list character styled reviews as
follows:
 A text line has between two and ten words (one
word is excluded to avoid headlines, more then
ten words are excluded to avoid normal text).
 At least three text lines as defined above are in
the review.
List character styled reviews only count for a
small number of all reviews. They are more popular
for technical products like digital cameras and
mobiles than for the remaining categories (see Table
6).
Table 6: List Character Style Reviews.
Category List Character Styled
Reviews [%]
Books 0.64
Digital cameras 5.12
DVD 2.62
Mobile phones 8.66
Music 1.84
As depicted in Table 7 list character styled
reviews seem to be more helpful for the customer
than prose styled product reviews.
Table 7: Helpfulness of Prose and List Style Reviews [%].
Category Prose Styled List Styled
Books 75 65
Cameras 81 86
DVD 54 58
Mobile phones 68 78
Music 59 59
A reason for that could be that list character
styled product reviews might be easier to be read
and are more fact oriented than prose styled reviews.
4.5 Readability Index
In literature several indices are suggested to formally
measure the readability of prose, e.g. the Flesh
Reading Ease (Flesh 1994) or the Wiener Sachtext
Formel (Bamberger et al. 1984).
In our analysis we use the LIX index (Ohrt 1980)
which is defined for the German language as
follows:
LIX = SL + LW
with
 SL: average number of words in a sentence and
 LW: percentage of words with more than six
letters.
According to Ohrt (1980) a classification as
depicted in
Table 8 has been suggested for the
German language.
Table 8: LIX Classification.
LIX Readability Examples
20-25 very easy
30-35 easy books for children
40-45 average fiction
50-65 difficult non-fiction
60-70 very difficult textbooks
For our product categories we obtained
correlations between the LIX and the helpfulness of
a product review as shown in
Table 9.
Table 9: LIX and Helpfulness.
Category Correlation
Books 0.10
Cameras 0.18
DVD 0.18
Mobile phones 0.17
Music 0.20
The observed correlations are weak in every
product category. However, again technical
equipment show similar correlations while books
CONSUMER-TO-CONSUMER TRUST IN E-COMMERCE - Are there Rules for Writing Helpful Product Reviews
65

have the weakest relationship between LIX and the
helpfulness of a product review.
5 CONCLUSION
In this paper we investigate if there are rules how to
write helpful product review. In a case study we
analysed five product categories of the Germany's
Amazon shop.
In our analysis we showed that there is no simple
guidebook how to write good product reviews but
some tendencies that longer rather than shorter
product reviews are more helpful. Also using lists
seems to be promising to write good product
reviews. To explicitly express a personal opinion by
the usage of personal pronouns makes a product
review more helpful for the readers. And last but not
least positive reviews are regarded as more helpful
than negative ones. However the correlations
between these effects are rather weak.
In our future work we will analyse
interdependency of the text features, extend our
research to other country sites of Amazon and - if
possible - try to support our analysis by semi-
structure reader interviews.
ACKNOWLEDGEMENT
Support by the Scientific Millenium Institute
"Complex Engineering Systems'' is greatly
acknowledged; see www.sistemasdeingenieria.cl.
REFRERENCES
Arndt, J. 1967. Roll of Product-Related Conversation in
the Diffuse of a New Product. Journal of Marketing
Research 4, 291-295.
Bamberger, R., Vanecek, E. 1984. Lesen - Verstehen -
Lernen - Schreiben. Die Schwierigkeitsstufen in
deutscher Sprache. Jugend und Volk. Wien.
Bickart, B., Schindler, R. 2001. Internet Forums as
Influential Sources of Consumer Information, Journal
of Interactive Marketing 15, 31-40.
Bruner, R. 2007. DoubleClick Touchpoints IV - Europe:
How French, British, and German Consumers See the
Role of Digital Media in Their Purchase Decisions.
(http://emea.doubleclick.com/UK/downloads/pdfs/Tou
chpoints_IV_uk.pdf, retrieved 26.10.07).
Chevalier, J., Mayzlin, D. 2006. The Effect of Word of
Mouth on Sales: Online Book Reviews. Journal of
Marketing Research XLIII, 345-354.
Duhan, D., Johanson, S., Wilcox, J., Harrell, G. 1997.
Influences on consumer use of word-of-mouth
recommendations sources. Journal of the Academy of
Marketing Science 25, 283-295.
Flesh, R. 1994. Art of Readable Writing. Macmillan
General Reference.
Flynn, L. 2007. Amazon's revenue rises but profit drops.
International Herald Tribune Online, 02.02.2007
(http://www.iht.com/articles/2007/02/02/business/ama
zon.php, retrieved 26.10.07).
Hennig-Thurau, T., Hansen, U., Eifler, V., Bornemann, D.
2002. Vertrauen in Kundenartikulationen auf
virtuellen Meinungsplattformen. Bruhn, M., Stauss, B.
(eds.), Dienstleistungsmanagement Jahrbuch 2002 –
Electronic Services 3, Wiesbaden: Gabler, 461-487.
Henning-Thurau, T. 2004. Warum Kunden anderen
Kunden im Internet zuhören. Jahrbuch der Absatz-
und Verbrauchsforschung 50, 52-75.
Merschmann, H. 2007. Rezensions-Mißbrauch - Guerilla-
Marketing bei Amazon. Spiegel Online, 15.04.2007
(http://www.spiegel.de/netzwelt/web/0,1518,476359,0
0.html, retrieved 26.10.2007).
Ohert, C. 1980. SL+LW=Lix. IDV-Rundbrief 26, 15-25.
Pollach, I. 2006. Electronic Word of Mouth: A Genre
Analysis of Product Reviews on Consumer Opinion
Web Sites, Proceedings of the 39th Hawaii
International Conference on System Sciences. Los
Alamitos: IEEE Computer Society Press.
Westbrook, R. 1987. Product/Consumption-Based
Affective Responses and Postpurchase Processes.
Journal of Marketing Research 24, 258-270.
ICEIS 2008 - International Conference on Enterprise Information Systems
66