Online Consumer Review: Word-of-Mouth as a New Element of Marketing

Communication Mix
Author(s): Yubo Chen and Jinhong Xie
Source: Management Science , Mar., 2008, Vol. 54, No. 3 (Mar., 2008), pp. 477-491
Published by: INFORMS

Stable URL: https://www.jstor.org/stable/20122400

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Management
Science

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
MANAGEMENT SCIENCE MSH
Vol. 54, No. 3, March 2008, pp. 477-491 DOI i0.1287/mnsc.l070.0810
ISSN 0025-19091 EissN 1526-55011081540310477 ? 2008 INFORMS

Online Consumer Review: Word-of-Mouth as a

New Element of Marketing Communication Mix
Yubo Chen
Eller College of Management, University of Arizona, Tucson, Arizona 85721,
y ubochen@eller. arizona.edu

Jinhong Xie
Warrington College of Business, University of Florida, Gainesville, Florida 32611,
[email protected]

As a new type that

phenomenon of isword-of-mouth
playing an increasingly information, onlinepurchase
important role in consumers7 consumer
decisions.product
This paper review is an emerging market
argues that online consumer review, a type of product information created by users based on personal usage
experience, can serve as a new element in the marketing communications mix and work as free "sales assistants"
to help consumers identify the products that best match their idiosyncratic usage conditions.
This paper develops a normative model to address several important strategic issues related to consumer
reviews. First, we show when and how the seller should adjust its own marketing communication strategy in
response to consumer reviews. Our results reveal that if the review information is sufficiently informative, the
two types of product information, i.e., the seller-created product attribute information and buyer-created review
information, will interact with each other. For example, when the product cost is low and/or there are sufficient
expert (more sophisticated) product users, the two types of information are complements, and the seller's best
response is to increase the amount of product attribute information conveyed via its marketing communications
after the reviews become available. However, when the product cost is high and there are sufficient novice (less
sophisticated) product users, the two types of information are substitutes, and the seller's best response is to
reduce the amount of product attribute information it offers, even if it is cost-free to provide such information.
We also derive precise conditions under which the seller can increase its profit by adopting a proactive strategy,
i.e., adjusting its marketing strategies even before consumer reviews become available. Second, we identify
product/market conditions under which the seller benefits from facilitating such buyer-created information
(e.g., by allowing consumers to post user-based product reviews on the seller's website). Finally, we illustrate
the importance of the timing of the introduction of consumer reviews available as a strategic variable and show
that delaying the availability of consumer reviews for a given product can be beneficial if the number of expert
(more sophisticated) product users is relatively large and cost of the product is low.
Key words: online consumer review; word-of-mouth; product review information; marketing communications;
social interactions
History: Accepted by Jagmohan S. Raju, marketing; received June 21, 2005. This paper was with the authors
111 months for 2 revisions.

1. Introduction have adopted a similar strategy. These online sellers

The Internet and information technology provide a invite users of their products to post personal prod
new opportunity for consumers to share their product uct evaluations on the sellers' websites or provide
evaluations online (Avery et al. 1999). Amazon.com their customers with consumer review information
began offering consumers an option to post their offered by some third-party sources such as Epin
comments on products on its website in 1995. Cur ions.com. Online consumer reviews are common for
rently, Amazon.com has about 10 million consumer many product categories such as books, electronics,
reviews on all its product categories, and these games, videos, music, beverages, and wine.
reviews are regarded as one of the most popular Recent evidence suggests that consumer reviews
and successful features of Amazon (New York Times have become very important for consumer purchase
2004). In recent years, an increasing number of online decisions and product sales. A study by Forrester Re
sellers (e.g., BevMo.com, BN.com, cduniverse.com, search finds that half of those who visited the retailer
circuitcity.com, GameStop.com, computer4sure.com, sites with consumer postings reported that consumer
c-source.com, half.com, goodguys.com, wine.com) reviews are important or extremely important in their
477

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
478 Management Science 54(3), pp. 477^91, ?2008 INFORMS

buying decisions (Los Angeles Times 1999). Based on (Wernerfeit 1994a) to help consumers to identify prod
the data from Amazon.com and BN.com, Chevalier ucts that best match their needs.
and Mayzlin (2006) find that online book reviews To examine such a matching function of online con
have a significant impact on book sales. Liu (2006) sumer reviews, we first present an empirical study to
shows that consumer reviews at the Yahoo Movies illustrate how this emerging information source is dif
website has a significant effect on box office revenue. ferent from other types of product information, such
However, not all online sellers supply consumer as third-party product reviews. We then develop a
reviews on their websites. For example, Chen and Xie normative model to address several specific questions
(2004) examine three product categories: MP3 play regarding a firm's strategic decisions vis-?-vis con
sumer reviews.
ers, PDAs, and video games. They identify a list of 68
online sellers from the referral list of the leading shop Our empirical study suggests that, different from
ping agent mySimon.com in June 18, 2003, and find third-party product reviews that emphasize the per
that 46 out of 68 online sellers did not offer consumer formance of a product based on its technical spec
reviews. ifications, consumer reviews tend to examine the
Online consumer review is a new product infor performance of a product from the perspective of
mation channel with growing popularity and impor its ability to match the consumers' own usage situa
tance. It has generated considerable attention in tions. Our strategic analysis reveals several important
practitioners and popular presses. Sellers face vari findings. First, we show that the two types of infor
ous important strategic decisions regarding consumer mation?consumer reviews and seller-created product
review information. For example, when consumer attribute information?can be complements or substi
tutes. Such interaction exists when the review infor
reviews appear, should a seller adjust its own commu
nication strategy to best respond to such a consumer
mation is sufficiently informative. The direction of
created information channel, and how? Under what the interaction (i.e., complementary or substitutive)
is determined by the characteristics of the product
conditions does the seller benefit from facilitating the
creation and dissemination of such user-based review and market. When the product cost is low and /or
there are sufficient expert (more sophisticated) prod
information by allowing consumers to post their com
uct users, the two types of product information are
ments on its own website (e.g., Amazon.com)? To
better understand the fundamental role of this new complements. In this case, the seller should increase the
amount of its own product attribute information con
information channel in the marketplace and its strate
veyed to potential customers when consumer reviews
gic implications to online marketers, more academic
become available. When the product cost is high
research is urgently needed.
and there are sufficient novice (less sophisticated)
Several recent studies have begun to examine online
product users, the two types of product informa
consumer-created information from the perspective tion are substitutes. Here, the seller should decrease its
of information credibility. Consumer-created informa
product attribute information supply when consumer
tion is likely to be more credible than seller-created reviews become available. In addition, we show that
information because credibility of information is often if the seller can anticipate the availability of consumer
positively related to the trustworthiness of the infor reviews, it is possible to adopt a proactive strategy
mation source (Wilson and Sherrell 1993). Dellarocas by adjusting its marketing strategies even before con
(2003) reviews the relationship between online con sumer reviews become available. Second, our analysis
sumer feedback information and an unknown seller's
reveals that allowing consumers to post user-based
reputation. Mayzlin (2006) studies the credibility of product reviews on the seller's website can increase
the promotional messages in online chat rooms and or decrease profit depending on product/market con
the implication of such new information channels on ditions. We show that it is detrimental to a seller to
sellers' profitability Furthermore, some recent stud supply consumer reviews unless such information is
ies (Fay and Xie 2008, Xie and Gestner 2007) suggest sufficiently informative. We also find that supplying
that consumer-created information allows the seller to
online consumer reviews is more likely to be ben
implement some marketing strategies that may not be eficial to the seller when there are sufficient novice
credible otherwise (e.g., probabilistic selling, service consumers (e.g., for technology-intensive products).
cancellation). These studies have advanced our under Finally, our results reveal that if it is possible for the
standing of consumer-created information. An impor seller to decide when to offer consumer reviews at the
tant but underexplored aspect of consumer reviews individual product level, it may not always be opti
is their degree of relevance to consumers. We argue mal to offer them at a very early stage of new prod
that online consumer reviews can be deployed as a uct introduction, even if such reviews are available.
new element in the marketing communications mix Delaying the availability of consumer reviews for a
and work as an online seller's free "sales assistants" given product can be beneficial if the number of the

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 479

expert users is relatively large and cost of the product 2.1. Online Consumer Review as a New Element
is low. in the Marketing Communications Mix
From a theoretical perspective, this paper is mostly As consumer-created information, online consumer
related to Lewis and Sappington (1994), which pro review is likely to be more relevant to consumers
poses a model to show when it is optimal for the seller than seller-created information. Seller-created product
to provide partial- versus full-attribute information to information is more likely to be product oriented,
consumers. In their model, there is only one infor because it often describes product attributes in
mation channel between the seller and consumers terms of technical specifications and measures prod
(i.e., from the seller to consumers). Unlike Lewis and uct performance by technical standards. In con
Sappington (1994), we allow an additional informa trast, the consumer-created product information is,
tion channel (i.e., from consumers to consumers), and by definition, user oriented. It often describes prod
examine a seller's information decision in a setting of uct attributes in terms of usage situations and mea
dual channels. sures product performance from a user's perspective
Substantially, this paper augments the traditional (Bickart and Schindler 2001). Consumers have dif
marketing communications literature. To date, very ferent information-processing capabilities in inferring
few studies have examined a firm's strategic decisions benefits from product attribute information due to dif
regarding information content for its marketing com ferent levels of expertise (Alba and Hutchinson 1987).
munications. Wernerfelt (1994b) and Simester (1995) For this reason, seller-created product information
investigate when and how firms should include price may be more useful to more sophisticated consumers
information in their advertising. Godes (2003) stud (i.e., experts). Consumer-created product information,
ies the implications of the value-creating versus per however, can help less-sophisticated consumers (i.e.,
suasive personal selling format. Chen and Xie (2005) novices) in finding their best-matched products. As a
examine a firm's advertising format strategy in the result, consumer reviews can be deployed as a new
presence of third-party product reviews, and find that element in the marketing communications mix and
can work as an online seller's free "sales assistants"
using review-endorsed advertising (i.e., advertise
ments containing third-party award logos) to broad (Wernerfelt 1994a) to help consumers to identify prod
ucts that best match their needs.
cast its success can hurt the winning product of a
product review. In this paper, we study a firm's infor Consumer reviews are important for unsophisti
mation content strategy by investigating how much cated consumers (i.e., novices), who may hesitate to
and what type of product information a seller should purchase if only seller-created product information is
available. However, this sales assistant does not come
provide to its customers.
The remainder of this paper is organized as fol without cost. By allowing consumers to post their
lows: Section 2 illustrates the marketing role of online
own product evaluations, the seller creates a new
consumer reviews and how they differ from other information channel for consumers, which thereby
types of product information; ?3 presents our model eliminates the seller's capability to control the supply
setup; ?4 examines how the seller should best respond of product information. In this paper, we study when
the seller should facilitate consumer reviews and how
to consumer reviews (i.e., the optimal information
content decision); ?5 studies conditions under which it adjusts its own communication strategy in response
to consumer review information.
the seller should initiate or facilitate such consumer
review information itself (i.e., the optimal consumer 2.2. Online Consumer Review vs. Traditional
review supply decision); and ?6 concludes the paper (Offline) Word-of-Mouth
and discusses some strategic implications and direc
tions for future research. Online consumer reviews, as consumer-created prod
uct information, can be viewed as a special type of
WOM (e.g., Godes and Mayzlin 2004). Different from
2. Online Consumer Review: the traditional WOM, the influence of which is typ
ically limited to a local social network (e.g., Brown
An Emerging Source of and Reingen 1987, Biyalogorsky et al. 2001, Shi 2003),
Product Information the impact of online consumer reviews can reach far
As an emerging source of product information, beyond the local community, because consumers all
what fundamental role can online consumer review over the world can access a review via the Internet.
play in the marketplace? How does online con In addition, in general, traditional WOM is not a
sumer review differ from other product information, direct decision variable for the seller. However, the
such as seller-created product information, traditional recent development of information technology allows
word-of-mouth (WOM), and third-party product a seller to effectively initiate and broadcast consumer
reviews? online reviews via its own website. A seller can also

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
480 Management Science 54(3), pp. 477-491, ?2008 INFORMS

license consumer reviews from intermediaries (such on a scale of 0 through 10 based on his/her evalua
as Epinions.com), and decide when to offer them on tions on four key aspects of cameras: features, perfor
its website (e.g., c-source.com). Given the widespread mance, image quality, and design.
impact of consumer reviews, our paper investigates (2) Consumer review data from Amazon.com, the
how firms should adjust their marketing communi pioneer and top provider for online consumer prod
cation strategy to respond to this emerging source uct reviews. When posting reviews for a camera at
of WOM information. Our paper complements WOM Amazon.com, consumers are asked to give a star rat
literature by also examining the new and potentially ing (from 1 to 5) and write a paragraph describing
powerful opportunity for a seller to help consumers their experiences and rationale for their ratings. Based
create and disseminate their personal opinions about on the different consumer postings, Amazon.com
the seller's products. We look at the benefits and gives an average customer rating for each model.
drawbacks of encouraging or discouraging this spe (3) Product attribute data from CNET.com. We col
cial type of WOM information and provide insight lect data on the three most important digital cam
into firms' decisions on when and how to provide era attributes suggested by Consumer Reports: image
consumer reviews. resolution (megapixels), optical zoom, and shooting
speed.
2.3. Online Consumer Review vs. Third-Party (4) Other control variables. As the control variables,
Product Review we also collect data on the product launch date from
Another information source closely related to online CNET.com and the number of consumer review post
consumer review is product review from third parties ings at Amazon.com.
(e.g., CNET.com, caranddriver.com, PC Magazine, PC Our sample includes all 120 digital camera models
World). As discussed in Chen and Xie (2005), third reviewed by CNET.com from June 2004 to September
party product reviews provide product information 2005. Table 1 (Part A) presents the descriptive statis
usually based on lab testing or expert evaluations. tics of our samples. As shown in this table, for each
Third-party product reviews tend to focus on product
model in our sample, the average number of con
sumer reviews posted at Amazon.com is 23, and the
attribute information (e.g., performance, features, and
average product length of life (the difference between
reliability) because such information is easier to quan
the product launch date and our data collection date)
tify and measure. As a result, third-party review rat
is 338 days. Among 120 cameras, 90 models have
ings are likely to be correlated with the performance
complete data on third-party product review, prod
of these attributes. Different from third-party reviews,
uct attributes, and the two other control variables;
online consumer reviews are posted by users based
and 87 models have complete information on con
on their personal experiences, which can be highly
sumer reviews, product attributes, and the other con
affected by their taste preferences as well as their trol variables.
personal usage situations. For this reason, consumer
Using the CNET.com editor's review ratings and
reviews are more likely to focus on whether and how
Amazon.com average consumer ratings as the depen
a product matches a specific individual's preference
dent variables, we run two separate regressions to see
and usage condition.
if the two types of product reviews?third-party
To illustrate this difference, we conduct a prelim review and consumer review?have a similar relation
inary empirical study. We chose the digital camera
ship to product attribute information. As shown in
for our study because it is an ideal product category Table 1 (Part B), F statistic is significant for the third
to study online consumer and third-party product party review model but not for the consumer review
reviews, for the following reasons: (1) according to the model. Also, the rating of third-party review is signif
Consumer Electronic Association's annual ownership icantly affected by optical zoom and shooting speed,1
study (Raymond 2006), the digital camera has become but none of the three product attributes affect the rat
one of the top five most popular consumer electronic ing of the consumer review. Furthermore, for the 120
products; and (2) since 2000 the Internet has been the
most popular channel for consumers to buy digital
1 The coefficient of image resolution is not significant. One main
cameras (Photo Marketing Association International reason, as suggested by Consumer Reports, is that the image resolu
2001). We have collected the following data for our tion (megapixels) is the major category variable for a digital camera.
empirical study: Therefore, reviewers tend to rate different cameras within a cate
(1) Third-party product review data from CNET.com, gory such as five megapixels instead of comparing a two-megapixel
model with a five-megapixel model. The positive coefficient for
the leading third-party professional review website
product life length shows that, given the same attribute level, a
for consumer technology products. When reviewing model can get higher ratings if it was launched into the market
digital cameras, a CNET.com editor presents detailed earlier, which is consistent with the fast-evolving characteristic of
product attribute information, and rates the camera the technology-driven digital camera market.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 481

Table 1 Third-Party Professional Review vs. Online Consumer Review looking for something that took the picture NOW!
A: Descriptive statistics As opposed to 3 seconds later. Being a parent, this
was very important to me... this camera is a real
N Min. Max. Mean Std. deviation
treasure"; "The 10X zoom makes it easy to see images
Third-party review 120 4.8 8.4 6.8 0.67 a long way away. I am able to capture the beauty
rating (CNET.com) of deer and other outside landscapes and animals
Consumer review 115 3.0 5.0 4.2 0.44 at great quality... ." Differently, focusing mainly on
rating (Amazon.com)
120 1.9 13.5 5.3 1.82 the attribute information, a typical paragraph in the
Image resolution
(megapixels)
third-party review from CNET for the same model is
Optical zoom (X) 111 0.0 12.0 4.1 2.62 "shutter lag was moderate at 0.7 second under high
Shooting speed (fps) 98 0.0 5.0 2.1 0.92 contrast lighting but a languorous 1.8 seconds under
Product life length 120 71.0 739.0 337.9
156.61 more challenging low-contrast lighting, even with aid
(days)
No. of Amazon.com 120 0.0 134.0 23.2 23.97 from the focus-assist lamp." Based on this qualitative
consumer postings
inquiry, it is clear that consumers evaluate their pur
chased product based on whether it fits their indi
B: Regression analysis vidual preferences and performs well in their specific
Dependent variable: Dependent variable: situations, which is quite different from third-party
Third-party review Consumer review reviews provided by professionals emphasizing prod
rating (CNET.com) rating (Amazon.com) uct technical specifications and performances.
Standardized Standardized
Independent variables coefficient t statistic coefficient t statistic
3. Model Setup
Image resolution 0.082 0.831 0.112 0.968 In this section we specify key assumptions and setup
Optical zoom 0.356*' 3.770 0.088 0.799 for our model. Key notations are summarized in the
Shooting speed 0.264*' 2.669 0.017 0.145 appendix.
Product life length 0.194* 1.910 -0.089 -0.756
No. of Amazon.com 0.010 0.091 3.1. Seller and Consumers
consumer reviews
No. of observations 90 87
We consider a monopoly seller2 carrying a multi
/?-squared 0.247 0.032 attribute product. Let c denote the marginal cost of
F-statistic 6.986*' 0.747 the product.
We allow consumer heterogeneity in two dimen
Note. The specification also includes an intercept.
sions: preference and expertise. First, we allow con
*p < 0.10, **p < 0.05, ***p < 0.01.
sumers to differ in their preferences toward the
seller's product. For a given product, some consumers
models tested, the correlation between the consumer will find that the product matches their preference
review ratings and third-party product review ratings better than others. Specifically, consider a product
is only 0.267 (p < 0.01), which suggests that these two with two attributes, ax and a2. For any given con
types of reviews may not offer the same information. sumer, there is an equal chance that a given attribute
A closer look at the data finds that although matches her preference, which is known to both
many camera models get low ratings from a third the seller and buyers. Consumer preferences for the
party, they get high ratings from consumers. For two attributes are independent. For example, a video
instance, Kodak Easy Share Z 740 gets 6.4 out 10 game often has two key attributes: (1) genre, which
points rated by experts at CNET.com, but gains 4.5 specifies the type of the game (i.e., role-playing game
out 5 stars based on more than 100 consumer post or strategy game); and (2) plot difficulty, which
ings at Amazon.com. Whereas the professional review indicates the challenge level to the players. Indepen
focuses on its lukewarm performance, most consumer dence in preference implies that a consumer's pref
reviews praise how this camera matches their dif erence for the type of the game is not necessarily
fering usage conditions. For instance, in the Kodak related to her preference for the difficulty of the game
Easy Share Z 740 example, just briefly skimming "the
most recent 20 postings" at Amazon.com, we can 2 The seller's monopoly position mainly results from consumers7
identify more than 10 different usage situations, vary loyalty and limited search. Recent studies have demonstrated
ing from outdoor landscape shots, animal and bird online consumers' loyalty and limited search for online sellers. For
shots, kids' sports, New Year's Eve celebrations, long example, Johnson et al. (2004) present empirical evidence that con
sumer online search is very limited during the shopping process.
distance shooting, Christmas gifts, overseas vacation
On average, consumers visit 1.2 book sites and 1.3 compact disc
trips, sharing photos with friends, family trips to sites in each category. The monopoly model can help us understand
Disney World, even crime-scene photography. Here the fundamental impact of the new information channel, i.e., online
are words in some typical consumer reviews: "I was consumer review on firm marketing strategies (e.g., Shugan 2002).

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
482 Management Science 54(3), pp. 477-491, ?2008 INFORMS

(i.e., a role-playing game lover may prefer difficult 3.2. Information Structure
or easy games). Hence, consumers can be catego We allow a two-sided information asymmetry be
rized into four types according to their preference tween the seller and consumers. The seller has pri
matching situations with the product: fully matched vate product information, but has no information on
type TMM (matching on both attributes), partially consumer characteristics. Consumers know their own
matched types TM^ and T^M (matching on attribute tastes and expertise levels, but have no information
a1 or a2), and fully unmatched type T^m (matching on on product attributes.
neither attribute). Let vF ,vp, v? denote consumer val There are two possible information sources for con
uations for fully matched, partially matched, and fully sumers: (1) seller-created product attribute informa
unmatched consumers when consumers have com tion, (2) consumer review information (if available).
plete information on product attributes, respectively. Due to the seller's concern for its reputation, we
When consumers are not fully aware of information assume the seller-created product attribute informa
on product attributes, they form their valuation based tion is accurate and truthful. Without loss of gener
on the expected value. For example, in the case with ality, we assume the seller's information supply cost
out any product information, all consumers have the is zero, considering the significantly reduced costs of
same willingness to pay, v = (vF + 2vp + z;?)/4. Without collecting and distributing information via the Inter
loss of generality, we assume v? = 0. net (Avery et al. 1999).
Second, we allow consumers to differ in their exper The information structure is determined by (1) how
much attribute information the seller provides via its
tise and knowledge about the product. We consider
own communication to consumers, and (2) whether
two consumer segments: an expert segment (?) and
product reviews created by the current users are avail
a novice segment (N). Let S denote the segment,
where S = E,N. Alba and Hutchinson (1987, p. 428, able to consumers. We call the former information
content strategy and the latter consumer review sup
Proposition 4.7) argue that, due to their difference
ply strategy. The information content strategy is com
in causal inference capability, "experts are more able
pletely determined by the seller. Specifically, the seller
than novices to infer intended product benefits from
can choose to adopt a full-information strategy (i.e.,
technical information and to infer likely technical
providing information on both attributes), a partial
causes of claimed benefits." This proposition implies
information strategy (i.e., providing information only
that experts are more likely than novices to cor on one of the two attributes), or a no-information
rectly map their usage situations with the product strategy. To focus on more realistic and interesting
attributes based on the attribute information offered
cases (i.e., full- and partial-information strategy), we
by the seller. Because consumer-created information assume that in the absence of consumer review infor
is more user oriented than seller-created informa
mation, the no-information strategy cannot be optimal.
tion, it is likely that all consumers can benefit from Specifically, we assume that the seller cannot make
such information. Accordingly, we assume that, sim a positive profit when refusing to offer any prod
ilar to experts, novices can identify a matching or uct attribute information to consumers.3 Let I = IF ,IP
mismatching product by learning from the experi denote the information content decision, where IF and
ence of existing users through consumer reviews. Ip present the case when the seller adopts full- and
However, different from experts who can find the partial-information content strategies, respectively.
matched product solely based on seller-created infor Let P(0) denote the probability that a consumeras
mation, novice consumers are unable to match prod true status is 6 for an attribute, where 6 = M, M
uct attributes with their preferences in the absence of (i.e., the attribute can be either a match or mismatch).
consumer reviews. In the absence of any information about the attribute,
To distinguish consumer heterogeneity in the pref there is an equal chance (i.e., 50%) that the attribute
erence and expertise dimensions discussed above, is a match/mismatch for a consumer. Hence, P(M) =
hereafter we refer to consumers with a different pref P(M) = 1/2.
erence as a different consumer type and consumers Let y denote the informativeness (accuracy and
with a different expertise level as a different consumer content) of consumer review information, where
segment. As described earlier, there are four types of 0 < y < 1. A higher degree of informativeness is asso
consumers with different preference-matching situ ciated with better information such that the con
ations (i.e., T = TMM, TMm, T?Mr tmm) an? two seg sumer review information is perfectly informative
ments of consumers with different expertise levels when y = l, and completely uninformative when
(i.e., S = E,N). The preference dimension and the 7 = 0. Let P(0 | s) denote the conditional probability
expertise dimension are orthogonal, i.e., for both
expert and novice segments, there are four types of con 3 This assumption implies that v < c, where v is buyer expected
sumers with different preference-matching situations. value in the absence of any product information.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 483

that a buyer's true valuation is 6 after receiving a sig post their product reviews on its website, some third
nal s from consumer reviews, where s = m, in (e.g., parties (e.g., consumer social network or infomediary
the signal can be either a "match" or "mismatch"). websites) may decide to create such user-generated
Intuitively, P(# | s) depends on the degree of informa information for a given product at any time. As a
tiveness of consumer reviews such that consumers are result, some sellers may find themselves facing unex
more likely to correctly identify their true status (i.e., pected consumer reviews (i.e., the case of the seller
match or mismatch with the given attribute) based on as an observer of online WOM in Godes et al. 2005).
consumer reviews when the reviews is more informa In ?4.2, we examine this case and show how the
tive. Formally, seller in this situation can best respond to unexpected
consumer reviews by adjusting its own information
d?(M | m)/dy > 0, dP(M \ m)/dy > 0,
(1) content strategy once reviews become available. We
d?(M | m)/dy < 0, dP(M \ m)/dy < 0. call this defensive response to consumer reviewers.
In ?4.3, we examine the case where the seller antici
For mathematical tractability, we adopt the same spec pates the availability of consumer reviews (e.g., if the
ification of P(0 | s) suggested in Lewis and Sappington seller allows consumers to post their reviews on its
(1994):
own website), and derive the seller's proactive strategy
toward consumer reviews.
' P(M | m) = P(M | m) = q(y) = (1/2 + 7/2)
(2)
P(M I m) = P(M I m) = q(y) = (1/2 - 7/2). 3.3. Model Timing
We consider two periods, t = 1,2. In each period,
This specification satisfies (1) and implies: one unit of consumers (with different preferences and
(a) When reviews are completely uninformative expertise levels) arrives at the market, makes a pur
(7 = 0), consumers' true status is independent of the chase decision, and then exits. Let 17 denote the size
signal received from consumer reviews, and such of the experts over two periods, and r\t denote the
reviews are useless (i.e., P(M | m) = P(M) = 1/2). fraction of experts in period t (i.e., 1 ? i)t is the frac
(b) When reviews are completely informative tion of novice consumers), where r\t = \tr\, 1 > Kx > 0,
(7 = 1), consumers can perfectly identify their true 1 > A2 > 0 and Ya=\ Af = l.5 Period 1 arrivals can learn
status based on consumer review information (i.e., about product attributes only from the seller-created
P(M\m) = P(M\m) = l). information. Period 2 arrivals can learn about prod
(c) When reviews are partially informative uct attributes from an additional information source,
(0 < 7 < 1), consumers benefit from consumer review consumer review information, if such reviews become
information but are unable to perfectly identify their available in period 2. In each period t, the seller
true status based on it (i.e., P(M) < P(M | m) < 1, adjusts its information content strategy J and price Pt
P(M) < P(M I m) < 1,0 < P(M I m) < P(M), and to determine which different segments and types of
0<P(M|m)<P(M).4 consumers to serve.
Note that (a)-(c) imply that the buyer revises her
expected valuation upward after receiving a "match"
signal, but downward after receiving a "mismatch" 4. Information Response to
signal. The more informative the review information, Consumer Reviews
the more the buyer adjusts her valuation based on the
consumer reviews. In this section, we examine the seller's best response
to consumer reviews. We first derive the seller's opti
It is important to notice that although the seller
has full control over the information content decision,
mal strategy in the absence of consumer review in
?4.1. We then study the seller's defensive response
the availability of consumer reviews may not be com
and proactive response to consumer reviews in ??4.2
pletely the seller's decision. For example, even if the
and 4.3, respectively.
seller determines not to offer its users the option to
4.1. Benchmark: In the Absence of
4 As shown in Appendix A.2, the specification in (2) can be de Consumer Reviews
rived as a posterior probability by specifying P(s | 6): P(m | M) =
P(m | M) = (1/2 + y/2) and P(m | M) = P(m | M) = (1/2 - y/2), We first derive conditions under which it is opti
where P(s | 0) is the conditional probability that a buyer obtains a mal for the seller to supply partial (full) attribute in
signal s from consumer reviews given that a buyer's true status is 6. formation to consumers in the absence of consumer
Such a specification of P(s | 6) implies that (a) the probability that reviews.
the buyer gets the correct signal from consumer reviews increases
with y and approaches to 1 when y = 1, and (b) the probability that
the buyer gets an incorrect signal from consumer reviews decreases 5 We assume ^ > 0 (a nonzero number of experts arriving in
with y and approaches to 0 when y ? 1. period 1) to ensure the availability of consumer postings.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
484 Management Science 54(3), pp. 477-491, ?2008 INFORMS

Partial-Information Strategy. Under this strategy, the margin-driven strategy, because by offering informa
seller provides information on only one attribute, for tion on both attributes, the seller is able to charge a
instance ax. Experts are certain about their match or high price, vF, to the fully informed/fully matched
mismatch on the informed attribute (ax), but remain experts, although other consumers will be priced
uncertain about the uninformed attribute (a2). With out of the market at this high price. On the other
out any information on the uninformed attribute, they hand, partial information is a volume-driven strat
perceive an equal probability that the uninformed egy because when the seller offers information on
attribute is a match or mismatch. Hence, the valuation only one attribute, both fully matched and some par
is (vp + vF)/2 for experts whose taste matches the tially matched experts (i.e., experts whose preference
informed attribute (i.e., TMm and Tmm) aRd (vp + v?)/2 matches the informed attribute ax) perceive the same
for experts whose taste mismatches the informed probability to be the fully matched type. The seller
attribute (i.e., T^M and 7^), respectively. Due to their can sell to both segments by charging a sufficiently
inability to process the seller-created information, all low price, (vp + vF)/2, to compensate for the con
novice consumers perceive an equal probability to be sumers' uncertainty. Lemma 1 shows that it is opti
one of four possible types and their expected valu mal to offer full information and only serve the fully
ation is v = (vF + 2vp + i??)/4. The seller maximizes matched experts (i.e., pursue a margin-driven strat
its profits by setting its price. In order to differentiate egy) when the cost is sufficiently high (c > vp), but
the benchmark case from that in the presence of con offer partial information and serve more types (i.e.,
sumer reviews to be examined below, we use a "hat" pursue a volume-driven strategy) otherwise.
for all the variables in the former. The optimal profit Lemma 1 is consistent with Lewis and Sappington
under the partial-information strategy in the absence (1994), although neither the buyer's heterogeneity
of consumer reviews is II(F)*. in expertise nor the availability of consumer review
information is considered in their model. In the rest of
Full-Information Strategy. Under this strategy, the
seller provides information on both product at this section, we examine how the seller should adjust
tributes. Expert consumers are fully informed. Their its information content strategy given in Lemma 1 in
valuations is vF for fully matched (T^), vp for par response to consumer reviews. Then, in the next sec
tially matched (TM^, T^M), and v? for mismatched tion, we examine the seller's decision on the supply
of consumer reviews?both issues that have not been
(Tmm) experts, respectively. The expected valuation
for novices is v. The optimal profit under full-infor previously explored.
mation strategy in the absence of consumer reviews
4.2. The Optimal Defensive Information
is ?(iFy.
Response to Consumer Reviews
Let I* denote the optimal information content We now consider the case where the seller faces unex
strategy in the absence of consumer reviews. Compar
pected reviews, and show how the seller can opti
ing the profits under the two strategies, partial infor
mally adjust its own information content strategies
mation and full information, leads to Lemma 1 (see
in response to consumer reviews. In the presence of
proofs of all lemmas and propositions in the appendix
consumer reviews, period 1 arrivals learn about the
and the online technical appendix, which is provided
product only from seller-created information, hence,
in the e-companion).6
having the same valuation as in the benchmark case.
Lemma 1 (Benchmark). In the absence of consumer Accordingly, the seller's period 1 strategy is the same
reviews, the seller's optimal information content and pric as that given in Lemma 1. However, different from
ing strategy is the benchmark case, period 2 arrivals can now learn
about the product not only from the seller, but also
offering full information from existing buyers via consumer reviews. Their val
(i.e.,r = IF),P; = vF ifc>vp; uations in period 2 may be different from that in
period 1, as shown below.
offering partial information Supplying Partial Attribute Information (I = Ip)
(i.e., 7* = Ip), P; = (vp + vF)/2 otherwise. Expert Consumers. In period 2, consumer reviews
with informativeness y become available. Hence,
Lemma 1 reveals that, in the absence of consumer period 2 experts can use the signal from con
reviews, either a full- or a partial-information strat sumer reviews to update their matching probabil
egy can be optimal. Note that full information is a ity on the uninformed attribute (a2). For example,
for those experts who have a match with informed
6 An electronic companion to this paper is available as part of attribute (ax) and receive the match signal on unin
the online version that can be found at http://mansci.journal. formed attribute (a2), their expected valuation is
informs.org/. q(y)vF + q(y)vp. Similarly, we can derive the expected

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 485

valuations for other types of period 2 experts (see (c) MAINTAIN the same level of attribute information,
Appendix A.2). otherwise.
Novice Consumers. In period 2, novice consumers
update their valuation based on consumer review in Mathematically, AI = I* - ?*
formation. For example, with the help of consumer
review information, the novices who receive the match
7f-Ip>0 if (i) c<vp, (ii) ye[y,y), or
signals on both attributes have as their expected valu y e [y, 1) and 77 > N;
ation q(y)q(y)v? + 2q(y)q(y)vp + q(y)q(y)vF. Similarly,
we can find the expected valuations for other types of Ip-iF<0 (3)
novices (see Appendix A.2). if (i) c>?p, (ii) ye[y,y"), (ni) t?<N;
Given the valuations of different consumer types/
segments, the seller sets price P2 to maximize its profit 0 otherwise.
in period 2, n2(7p).
Supplying Full Attribute Information (I = IF) Equation (3) implies that the seller can best respond
Expert Consumers. If the seller adopts the full-in to the availability of consumer reviews by revers
formation strategy, experts in period 2 are fully in ing its information content strategy, i.e., switching
formed. Their valuation is vF,vp,v? for types TMM, from supplying full to partial information or vice
TMM and TMM, and TMM, respectively. versa. This result is intriguing because it implies
Novice Consumers. Although the seller provides full that the two types of information, consumer cre
information, the expected valuations of novices are ated and seller created, can be either complements
the same as in the case where the seller provides par (i.e., consumer reviews increase the seller's incen
tial information, provided that novice consumers are tive to supply attribute information) or substitutes
unable to process seller-created information. (i.e., consumer reviews decrease the seller's incen
Given the valuations of different consumer types/ tive to supply attribute information). Proposition 1
segments, the seller sets price P2 to maximize its profit reveals that the existence and direction of the inter
in period 2, U2(IF)7 action are determined by three product/market fac
We now examine WHEN and HOW the seller tors: (1) product cost, (2) informativeness of consumer
should vary its own information content strategy in reviews, and (3) the size of different segments of con
response to consumer reviews. Let I* denote the opti sumers (experts or novices).
mal information content strategies in the presence of For low-cost products, it is optimal to offer par
consumer reviews and AI denote the difference in tial information and sell to both fully and some
the optimal amount of information content with and partially matched experts at a lower price in the
without consumer reviews, A7 = 7* ? 7*. The follow absence of consumer reviews (see Lemma 1). How
ing proposition states the seller's optimal information ever, such a volume-driven strategy is no longer
response to consumer reviews. optimal when consumers can learn about the prod
uct from consumer reviews with sufficient informa
Proposition 1 (Defensive Information Response
tiveness (y e [ y, y)). This is because such review
to Consumer Reviews). Facing unexpected consumer
information significantly decreases the valuation of
reviews, the seller can improve its profit by adjusting its
the experts receiving the mismatch signal on the unin
own information content strategy once reviews become
formed attribute, even as it also increases the valua
available. Specifically, compared with the case without con
tion of those experts receiving the match signal on the
sumer reviews, in the presence of consumer reviews, it is
uninformed attribute. As a result, to maintain such
optimal for the seller to
a volume-driven strategy, the seller has to reduce its
(a) INCREASE attribute information via its own com
price greatly. In contrast, switching to a margin-driven
munication if two conditions hold: (i) the product cost
(i.e., full-information) strategy, under which the seller
is low, and (ii) either the review informativeness is in a
midrange, or the review informativeness is extremely high
sells only to the fully matched experts at a high
price, is more profitable. Note that when consumer
and there are sufficient expert consumers.
review information is highly informative (y e [y, l)),8
(b) DECREASE attribute information via its own com
munication if three conditions hold: (i) the product cost is
the valuation of novices who receive match signals
high, (ii) the review informativeness is in a midrange, and
on both attributes increases significantly, so that it
(iii) there are sufficient novice consumers.
becomes profitable to sell to these novice consumers
if their segment is sufficiently large. If the novice seg
ment is sufficiently small (or equivalently, the expert
7 Theoretically, the seller can also decide to supply no attribute
information in the presence of consumer reviews. However, it is
straightforward that this strategy is a dominated strategy compared 8 When y = l, the seller is indifferent between providing full, par
with the partial-information strategy. tial, or no information given a zero supply cost.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
486 Management Science 54(3), pp. 477-491, ?2008 INFORMS

segment is sufficiently large, 17 > N), switching to the period 1 determines the number of potential review
full-information strategy to only serve full-matched ers in period 2, which can be positively related to
experts is more attractive than retaining the partial the informativeness of consumer review information,
information strategy. i.e., dy/dD1 > 0.10 Note that the seller can control Dx
For high-cost products, in the absence of consumer via its period 1 information content strategy (i.e., the
reviews, a margin-driven strategy (offering full infor margin-driven strategy characterized by offering full
mation and selling only to fully matched experts at information and high prices or the volume-driven strat
a high price) is more profitable than a volume-driven egy characterized by offering partial-information and
strategy (offering partial information and selling both low prices). Let T = y(D1 = iq^, which is the level of
review informativeness reached under the maximum
to fully and some partially matched experts at a low
price) (see Lemma 1). However, in the presence of period 1 demand (i.e., when all period 1 arrivals buy
consumer reviews, the volume-driven strategy can in period 1 under positive prices). We refer T to the
be more attractive than the margin-driven strategy. review informativeness potential.
When consumer review information is sufficiently When the seller anticipates the availability of con
informative (y > y), the valuation of the novices who sumer reviews in period 2, the seller can act proac
receive match signals on both attributes increases suf tively to adjust its strategies even before reviews
ficiently, so that it becomes profitable for the seller to become available, which can affect review informa
reduce its price to sell to these consumers in addi tiveness. Proposition 2 states when and how the seller
tion to the perfectly matched experts if the number of can benefit from this proactive strategy.
novices is sufficiently high (equivalently, the number Proposition 2 (Proactive Strategy Towards
of experts is sufficiently small, 77 < N). As a side bene Consumer Reviews). When the seller anticipates the
fit of this low-price, volume-driven strategy, the seller availability of consumer reviewers, it is optimal to
can profit from switching to the partial-information (a) Adjust its period 1 strategy (i.e., prices and informa
strategy by gaining extra demand from some partially tion content supply) if the review informativeness potential
matched experts. is sufficiently high. Specifically, iff T >T, then compared
Note that switching to the partial-information strat with Lemma 1, it is optimal to change period 1 strategy by
egy will not help the seller if the consumer review
information is extremely informative (y > y"). This decreasing price if c <vp,
is because, under the partial-information strategy, the
, switching to patial-information strategy
valuation of expert consumers who receive the mis
or remaining with full-information
match signal on the uninformed attribute is nega
tively related to the review informativeness (i.e., the
strategy but reduce price if c>vp.
more informative the reviews, the more likely for (b) Adopt the same period 2 response given in Proposi
these consumers to realize their mismatch). As a tion 1.
result, when the review informativeness is extremely
high (y > y"), the valuation of these experts becomes When the seller adopts a proactive response strategy
too low to retain the advantage of the partial and adjusts its marketing strategy before consumer
information strategy. reviews become available, it will offer a lower price
and generates a higher demand in period 1. Although
4.3. Proactive Response to Online such a response reduces period 1 profits, it can sig
Consumer Reviews nificantly increase period 2 profits because a larger
We now examine the case where the seller can proac number of buyers in period 1 leads to a higher level
tively respond to consumer reviews. This is possible of informativeness in the consumer reviews in period
when the seller anticipates the availability of con 2. Proposition 4 reveals that whether a seller should
sumer reviewers in the second period (e.g., if the adopt such proactive strategy depends on the review
seller allows consumers to post their reviews on its (WOM) informativeness potential, T. When the review
own website). Different from the last section, where
the review informativeness y is exogenously given, behavior. An implicit assumption behind this type of model is that
the seller may now influence y by controlling the the discount rate of early arrivals is very high, which is consistent
with the behavior of early consumers in many new product mar
number of consumers who purchase in period 1
kets (Moore 1991). If the discount rate of early arrivals is very low
(e.g., Da).9 This is because the number of buyers in and reaches zero (Af is endogenous), some period 1 experts might
wait to buy in period 2. In this case, the required price to induce
9 An alternative way to model a firm's dynamic behavior is to period 1 purchase would be lower.
consider consumers' strategic waiting behavior explicitly and treat 10 We do not specifically model consumer review posting behavior
Af as endogenous. Similar to some previous research (e.g., Lazear here. Admati and Pfleiderer (2004) have studied consumer posting
1986, Raju et al. 1990), our model does not explicitly consider such behavior and review informativeness.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 487

informativeness potential is very high, i.e., T > T, the reviews if it is possible to control when consumer
reviews become available.
profit decrease in period 1 from the proactive response
can be compensated for by a profit increase in period
2, the proactive response strategy improves seller's 5.1. Supply of Consumer Reviews
To derive the conditions under which it is optimal
profit, and vice versa. Proposition 4 also specifies how
to offer consumer reviews, we compare the profits in
a seller responds proactively when T > T. As stated the absence of consumer review information and in
earlier (see Lemma 1), when cost is low (c < vp), the
the presence of consumer reviews (under the optimal
optimal period 1 strategy for a seller is to adopt the
volume-driven strategy, under which the seller offers
strategic response), II* and II*. Proposition 3 states
conditions required for the seller to benefit from sup
partial information and sells to both fully and par
plying consumer reviews.
tially matched experts at a low price. Proposition 4
shows that when T > T, a seller will adopt a proac Proposition 3 (Consumer Review Supply Deci
tive response strategy and deviate from such a strat sion). The seller will not benefit from offering consumer
egy by further reducing its price in period 1. When reviews unless (i) the review informativeness potential is
cost is high (c > vp), the passive response strategy for sufficiently high (T > y), and (ii) the size of the expert
the seller is to maintain the margin-driven strategy segment is sufficiently small (17 < N').
in period 1 as in the case without consumer reviews, Proposition 3 reveals that the consumer review
under which the seller offers full information and sells
supply decision depends on two crucial factors: the
only to fully matched experts at a high price. However, review informativeness potential and the size of the
when T > T, the seller will deviate from such a strat
expert segment. As a new source of product infor
egy and adopts the volume-driven strategy by either mation, consumer review information is a double
switching to the partial-information strategy or simply edged sword that can benefit or hurt the seller. On
further reducing its price. the one hand, such information can increase the
An important implication of Proposition 2 is that profit from the novice segment because consumer
optimal response to consumer reviews may differ reviews can serve as a free sales assistant to novice
depending on the review (WOM) informativeness consumers. When the review informativeness poten
potential, which largely relies on product characteris tial is sufficiently high (T > y), consumer reviews
tics (Bone 1992). A seller acting proactively and inten can significantly increase product valuation for some
tionally increasing the number of buyers at an early novices (e.g., those who receive match signals on
stage of a product's life cycle may NOT be benefi both attributes). As a result, selling to these con
cial for ALL products. For instance, vanity products sumers is not profitable in the absence of consumer
are more likely to generate "buzz" (i.e., the number reviews, but can be profitable in their presence. On
of review postings), but embarrassing products are the other hand, consumer reviews can decrease the
less likely to do so; products targeted to younger con profit from the expert segment because such review
sumers (e.g., MP3 players) are more likely to generate information takes away the seller's control over the
buzz than products targeted to older consumers (e.g., information content available to experts. The seller's
hearing aids); original products are more likely to decision as to whether or not to provide online con
increase WOM than less-original products (Moldovan sumer reviews depends on the trade-off between its
et al. 2006). A proactive response strategy to reviews profit gain from novices and loss from experts. When
will improve profit only if the increased demand in the size of the expert segment is small (17 < N')
the early period can significantly increase the number (e.g., for technology-driven products), the seller will
of review postings, and thus increase the informative obtain a sufficient profit gain from the novices while
ness of consumer reviews in the later period. incurring very limited profit loss from the experts.
As a result, offering consumer reviews is advanta
geous. However, when the review informativeness
5, Decision on Supply of potential is extremely small (T < y), or the size of
Consumer Reviews the expert consumer is too large (17 > N'), the profit
In the last section, we focus on the seller's best infor gain obtained from the novices segment is insuffi
mation content response to consumer reviews. In this cient to compensate the profit loss from the expert
section, we examine another important strategic deci segment, and the seller can be worse off by offering
sion: whether the seller should initiate or facilitate consumer reviews. This might partially explain why
consumer review information itself, such as offering many lesser-known online stores do not offer their
its existing consumers the opportunity to post their shoppers the option to post reviews on their websites.
product reviews on its own website. We first examine Consumer reviews cannot be very informative unless
conditions under which such a strategy is beneficial, there is a sufficient number of review postings. For
and then discuss the optimal timing to offer consumer less-popular sellers with a small volume of transac

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
488 Management Science 54(3), pp. 477-491, ?2008 INFORMS

tions, it might be difficult to attract sufficient number full-information) strategy is optimal in the absence
of review postings on their websites. As suggested of consumer reviews in the former (latter) case and
by our model, the seller can be worse off from low consumer reviews reduce the profit from the expert
informative review information. As a result, it can be segment in the former, but not in the latter, case.
unprofitable for lesser-known sellers to facilitate such
consumer-created information.
6. Conclusion, Implications,
5.2. Timing of Offering Consumer Reviews and Discussions
We have shown that under some conditions (see Recent developments in information technology have
Proposition 3), the seller benefits from initiating con significantly increased online sellers' information
sumer reviews. However, when offering consumer capacity. In this paper, we investigate an emerging
reviews is profitable, is it always optimal to provide research area: online consumer reviews and their
such information immediately after product launch?
We now address this issue. implications on a firm's marketing strategies. Specif
ically, we study the marketing function of consumer
To allow for flexibility in timing, we extend our reviews and address three information decisions of
previous two-period model to a three-period model.
an online seller: (1) its best marketing communica
Let Af denote the percentage of experts who enter the
tion response to consumer reviews, (2) its decision to
market in the beginning of period t, t = 1,2,3 (i.e.,
actively facilitate the creation and dissemination of
Y?t=\ K ? !) Without loss of generality, we assume that
consumer reviews using its website as a medium, and
the review supply cost is zero, and the review informa
(3) the best timing for supplying consumer reviews.
tiveness is sufficiently high and reaches its potential
Compared with seller-created attribute information,
in the end of period 1. We examine, if the seller has consumer reviews are more user oriented and have
control over the time to provide consumer reviews at
an advantage in helping consumers to find products
the product level, whether it is best to offer reviews at
matching their preferences. This information is partic
the beginning of period 2 (i.e., right after the reviews
ularly important for unsophisticated consumers (i.e.,
become available) or the beginning of period 3 (i.e.,
novices), who will be less likely to buy the seller's
delaying the timing of offering). Proposition 4 pro
product if only seller-created product attribute infor
vides conditions under which it is profitable to delay
mation is available. However, this sales assistant does
the timing of offering consumer reviews.
not come without cost. By allowing consumers to
Proposition 4 (Timing Decision on Consumer post their own product evaluations, the seller cre
Review Offering). When consumer reviews improve ates a new information channel for consumers, which
seller profit, the seller is better off by delaying the timing eliminates the seller's capability to control the sup
of offering consumer reviews if k2>k and c <vp. ply of product information (e.g., providing full vs.
Proposition 4 reveals that when the seller benefits partial information to consumers). We provide sev
from offering consumer reviews, it is not always opti eral important strategic implications for online sellers'
decisions related to consumer reviews:
mal to provide them as early as they become avail
able. Specifically, for low-cost products (c < vp), if a Consumer reviews, as a form of independent
sufficient percentage of experts enter the market in product information, may play different marketing
period 2 (?2 > ?), then it is optimal for the seller to roles from that by third-party product reviews. The
postpone the supply of consumer review information. seller needs to develop a unique strategic response to
Because offering consumer reviews weakly decreases consumer review information. This research provides
the seller's profit from the expert segment but weakly some guides for such strategic response.
increase its profit from the novice segment, the opti The seller's optimal response to consumer
mal timing of offering consumer reviews depends on reviews may differ for different types of products.
the distribution of expert arrivals. When there are The seller would increase the product attribute infor
too many experts (or too few novices) in period 2 mation in response to consumer reviews for low-cost
(?2 > ?), the seller can benefit from delaying the tim products, but decrease for high-cost products.
ing of supplying consumer reviews (i.e., offering con For "exciting" products (e.g., vanity products,
sumer reviews in period 3 rather than in period 2) original products) or products targeted to "talkative"
because it is more profitable to give up the potential segments (e.g., college students), the seller can adopt
gain from the novices in order to avoid the potential a proactive response strategy, under which the seller
loss from the experts in this period. Finally, delay adjusts its marketing strategies even before consumer
ing supplying consumer reviews can improve profit reviews become available.
for low-cost products (c < vp) but not for high-cost When deciding whether to provide the platform
products (c > vp), because the partial-information (the for consumers to post their product reviews on the

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 489

seller's website, the seller needs to consider the rela the product-matching complexity. Because consumer
tive size of expert and novice segments. Offering con reviews are "created by users for users," increas
sumer reviews can benefit products with a sufficient ing in the degree of matching complexity implies
number of novice consumers (e.g., for technology that the review information will naturally contain
driven products), but can hurt the seller if the seg more possible usage situations (e.g., see the digi
ment of expert consumers is relatively large. tal camera examples discussed in ?2). An important
Less-popular online stores need to be cautious line of future research will be to empirically inves
when facing the decision of offering consumers the tigate firms' consumer review supply decisions and
option to post reviews on their websites. The seller to assess the impact of consumer reviews on a firm's
can be hurt by offering such consumer-created infor marketing strategy. When testing our model results,
mation when failing to attract a sufficient number of the future empirical studies can use the matching
consumer postings (i.e., failing to ensure a sufficient complexity between product attribute space and con
level of review informativeness). Such sellers may sumer usage condition space as a proxy variable for
want to consider licensing consumer reviews from review informativeness y.n
well-known third-party sources.
The timing of the introduction of consumer
review information can be an important strategic vari 7. Electronic Companion
able for a seller. When a seller is able to decide such An electronic companion to this paper is available as
timing at the individual product level, delaying the part of the online version that can be found at http://
availability of consumer reviews for a given product mansci.journal.informs.org/.
can be beneficial if the size of the expert segment is
relatively large and cost of the product is low. Acknowledgments
The authors thank Joseph Alba, David Sappington, Steven
Although this research improves our understand
Shugan, Barton Weitz, and participants in the Marketing
ing of online consumer review and its implications
Science Conference at Maryland (2003), the sixth Invita
for firm marketing strategies, many other interest tional Choice Symposium at CU-Boulder (2004), and the
ing questions remain unanswered and require fur second Product and Service Innovation Conference at Park
ther investigation. First, our empirical study is very City, Utah (2005), for their helpful comments. The authors
preliminary It is a first step to empirically under are grateful to the department editor, associate editor, and
stand the difference between consumer review and two reviewers for constructive and valuable suggestions.
third-party professional reviews. To completely study
this issue, future research may need to combine the Appendix
archival statistical data analysis, qualitative netno Summary of Notations
graphic inquiry (Kozinets 2002), and possible exper
imental testing to provide deeper insights. Second, c: Marginal cost of the seller's product
t: Time period in the model
we study a monopoly model and focus on the y: Consumer review informativeness
matching function of online consumer review. Future I: Product attribute information (I = IF ,IP,IF: full
research may study some other functions of online information; Ip: partial information)
consumer reviews and investigate its implications 6: Consumer true status of attribute matching sit
for firm competition. Third, even though the fun uation (6 = M,M,M: match; M: mismatch)
damental trade-offs examined in our model apply s: The signal from consumer reviews on attribute
to both direct-selling manufacturers and distributors, matching situation (s = m, m, m: "match," m:
it will be very interesting to study how the deci "mismatch")
sion on supplying consumer reviews and the sell T: Consumer type
ers' best response to consumer reviews differ across S: Consumer segment (S = E,N,E: expert; N:
novice)
different types of sellers (e.g., manufacturers versus
rjt: Fraction of expert consumers among all con
retailers). Fourth, future research may study from
sumers in period I
the perspective of consumer review intermediaries
kt: Fraction of experts among all experts in period t
such as Epinons.com, and examine their optimal mar v: Consumers' expected valuation in the absence
keting strategies. Finally, it is important to point of product information
out that it is possible for the seller to offer its Vj(I, y): Expected valuation of type T consumers in seg
own product-matching information that mimics con ment S given attribute information I and review
sumer reviews. Compared with consumer reviews, informativeness y
the seller-created product-matching information has a vF ,vp: Consumers' evaluations on their fully and par
cost disadvantage because the seller has to incur the tially matched products separately
cost of information creation and dissemination. It is
expected that such a cost disadvantage increases with 11 We thank one of the reviewers for this suggestion.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
 Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
490 Management Science 54(3), pp. 477-491, ?2008 INFORMS

v?: Consumers' evaluation on their fully product. Hence, the expected valuation for them is vp/2.
unmatched product (v? is assumed to 0) This valuation is lower than v = (vF + 2i?p)/4 and c. The
Pt: Seller's price in period t seller charges a price Pt = (vp + vF)/2 and gains the demand
II, IIf: Seller's overall profit and profit in period t in from type TMM and TMM experts. Recalling that rjt is the
the absence of consumer reviews fraction of the experts among all consumers in period t, the
II, Ut: Seller's overall profit and profit in period t in seller's profit in period t is
the presence of consumer reviews
ns,IIs: Seller's profit from segment S in the absence ?t(Ipy = ?F(Ipy = ((vp + vF)/2 - c)Vl/2. (6)
and presence of consumer reviews respectively
The seller's overall profit is
A.l. Proof of Lemma 1
U(IPY = UE(Ipy = ((vp + vF)/2 - c)(Vl +12)/2
In the absence of consumer reviews, when the seller pro
vides full information, the valuations are zero for type
= ((vp + vF)/2-c)V/2. (7)
Tmm> vP f?r tyPes tmm and Tmm/ and vF for tyPe tmm expert
Comparing (5) with (7), we find ?(IF)* > fi(?p)* and fi* =
consumers. The seller can charge a premium price vF to ?(IF)* iff c > z;p. D
serve only TMM or charge a low price vp to serve TMM, TMM,
and TMm- Consistent with Lewis and Sappington (1994), we A.2. Sketchy Proof of Proposition 1
focus on the case in which the seller charges a premium price In the presence of consumer reviews, the seller's profit from
with full information, which implies vF > 3vp ? 2c. Notice the experts in period 1 is the same as in period 1 in the
that for any given consumer, there is an equal chance that absence of consumer reviews, i.e., ITJ = 11^.
a given attribute matches her preference. In other words, In period 2, consumer reviews are available. Consumers
each attribute matches the preferences of half the consumers. revise their expected valuations using different posteriors:
Hence, the size of each type of consumers is 1/4. Note that P(M | m) = P(M | m) = q(y) = (1/2 + y/2) and P(M | m) =
the fraction of experts in period t is j]t. In each period t, P(M | m) ? q(y) = (1/2 ? y/2). These posteriors can in
seller's overall profit (i.e., its profit from the experts) is fact be derived using Bayes' Theorem and by assuming
n((/fr = nf(/Fr = (^-c)V4. (4) P(M) = P(M) = 1/2, P(m | M) = P(m | M) = (1/2 + y/2), and
P(m | M) = P(ra | M) = (1/2 - y/2). Specifically,
The seller's overall profit is
P(m | M)P(M)
?(IFY = ?E(IFY = (vF - c)(Vl + r?2)/4 = (vF - c)t,/4. (5)
P(M | m) =
?(m | M)P(M) +P(m | M)P(M)
When the seller only provides information on one at (l/2 + y/2)/2
= (1/2 + y/2)
tribute?for instance, a1?the expected valuation for the (l/2 + y/2)/2 + (l/2-y/2)/2
product is (vp + vF)/2 for type TMM and Tmm experts who
have matched tastes in the attribute av For those consumers P(M | m) = ?(m 1 M)P(M)_
P(m | M)P(M) + p(m | M)P(M)
who have mismatched tastes in the attribute a1 (type TMm
and Tmm)/ they have equal probabilities to find match and (l/2-y/2)/2
(1/2-y/2)
mismatch in a2 and have a valuation of vp and 0 on the (l/2-y/2)/2 4-(l/2 + 7/2)/2
Table A1 Novice Consumer Expected Valuations in the Presence of Consumer Reviews

Probability that type T novice's true status is 0 given the

signals obtained form consumer reviews
Novice Consumer expected
consumer 6=MM 0 = MM o = mm valuation
type (T) (v = 0) (v = vp) (v = vp) VTN(y)

Tmm q_(y)g_(y) q_(y)q(yY q{y)Q_{y) Q{y)Q(y) (2vp+ vF)/4 + yvF/2

-y2(2vp-vF)/4
Lm Q(y)Q(y) g_(y)g_(y) q{y)Q(y) Q(y)Q(y) {2vp + vF)/4
+ y2(2i/p-i/f)/4

Tmm q(y)g_(y) q(y)q(y) q_(y)q_(y) g_(y)q(y) (2vp + vF)/4

+ y2(2vp-vF)/4
im q{y)q(y) q{y)g_(y) q_{y)q{y) q_(y)g_{y) (2vp + vF)/4-yvF/2
-y2(2vp-vF)/4
Notes. If the seller provides full information, in period 2, different from the partial-information content case, con
sumer reviews have no impact on experts due to their fully realized valuations from full product attribute infor
mation. The expected valuations of four types of novices are the same as in the partial information content case.
Based on the expected valuations for all types of experts and novices, we can derive seller's profit functions,
which vary with c,y,r}t, and /. By comparing the profits in the presence of full information and in the pres
ence of partial information, we can derive Proposition 1. The detailed derivations are shown in Online Technical
Appendix AA.1.
*?(y) = 1/2 + y/2,anda(y) = 1/2-y/2.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms
Chen and Xie: Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix
Management Science 54(3), pp. 477-491, ?2008 INFORMS 491

P(m | M)P(M)
P(M I m) = ? Chen, Y., J. Xie. 2005. Third-party product review and firm market
ing strategy. Marketing Sei. 23(2) 218-240.
P(m | M)P(M) + P(m | M)P(M)
Chevalier, J., D. Mayzlin. 2006. The effect of word of mouth on sales:
(l/2-y/2)/2
= (1/2 -y/2) Online book reviews. /. Marketing Res. 43(August) 345-354.
(l/2-y/2)/2 + (l/2 + y/2)/2 Dellarocas, C. 2003. The digitization of word of mouth: Promise

PM m) = P(m|M)P(M)
and challenges of online feedback mechanisms. Management
-_ v _'?'?1-+ Sei. 49(10) 1401-1424.
P(m | M)P(M) + P(m | M)P(M)
Fay, S., J. Xie. 2008. Probabilistic goods: A creative way of selling
(l/2 + y/2)/2 products and services. Marketing Sei. Forthcoming.
= (1/2 + y/2).
(l/2 + y/2)/2 + (l/2-y/2)/2 Godes, D. 2003. In the eye of the beholder: An analysis of the rel
ative value of a top sales rep across firms and products. Mar
If the seller only provides information on one attribute?for keting Sei. 21(1) 74-96.
instance, a1?then experts know if ax is a match from the Godes, D., D. Mayzlin. 2004. Using online conversations to study
seller-provided information, and form a probability about word of mouth communication. Marketing Sei. 23(4) 545-560.
whether a2 is a match based on the signal they obtained Godes, D., D. Mayzlin, Y. Chen, S. Das, C. Dellarocas, B. Pfeiffer,
from consumer reviews. Thus, experts can be classified into B. Libai, S. Sen, M. Shi, P. Verlegh. 2005. The firm's manage
four categories. Let TMs and TMs denote the experts whose ment of social interactions. Marketing Eett. 16(3/4) 415-428.
preference matches and the experts whose preference mis Johnson, E., W. Moe, P. Fader, S. Bellman, J. Lohse. 2004. On the
matches with ax, respectively. The signal they obtained from depth and dynamics of world wide web shopping behavior.
Management Sei. 50(3) 299-308.
consumer review about a2 is s = m,m. The expected valua
Kozinets, R. V. 2002. The field behind the screen: Using netnogra
tions for four types of experts are
phy for marketing research in online communities. /. Marketing
Res. 39(1) 61-72.
V?MJIP, y) = vFq(y) + vpq_(y) = (vp + vF)/2 + y(vF - vp)/2,
Lazear, E. 1986. Retail pricing and clearance sales. Amer. Econom.
VLMm
(Ip,y) = vpq(y) = vp(l + y)/2, Rev. 76(1) 14-32.
Lewis, T., D. Sappington. 1994. Supplying information to facilitate
VL, VP' y) = vF?M + A~(7) = ivp + vF)/2 - {vF - vp)yl2,price discrimination. Internat. Econom. Rev. 35(2) 309-327.
and Liu, Y 2006. Word of mouth for movies: Its dynamics and impact
VF__?p,y) = vpq_(y) = vp(l-y)l2<c. on box office revenue. /. Marketing 70(July) 74-89.
Eos Angeles Times. 1999. Everyone is a critic in cyberspace. (Decem
When consumer reviews are available, let Tss denote the ber 3) Al.
novices who receive signal s on each attribute, s = m,m. Mayzlin, D. 2006. Promotional chat on the Internet. Marketing Sei.
The novice consumers can be categorized into four types 25(2) 155-163.
depending on the signals (match versus mismatch) on two Moldovan, S., J. Goldenberg, A. Chattopadhyay. 2006. What drives
attributes they receive from consumer reviews: Tmm, Tmm, word-of-mouth? The roles of product originality and useful
Tmm, and Tmm. As shown in Table Al, we can derive the ness. Report 06-111, Marketing Science Institute, Cambridge,
MA.
expected valuations for four types of novices as
Moore, G. 1991. Crossing the Chasm: Marketing and Selling High
'KJyy ' (2vp + vF)/4 + yi?F/2 - y2(2z;p - vF)/A'
Tech Products to Mainstream Customers. HarperCollins Publish
ers, New York.
KJy) (2vp + vF)/4 + y2(2vp -vF)/A New York Times. 2004. Amazon glitch unmasks war of reviewers.
(February 14) Al.
Vt1mm(y)
v'' (2vp + vF)/4 + y2(2up - vF)/A Photo Marketing Association International. 2001. 2000 Digital Imag
v Jy) (2vp + z;F)/4 - yz;f/2 - y2(2?p - vF)/?_
ing Marketing Association Digital Imaging Consumer Survey.
Jackson, MI.

References Raju, J. S., V. Srinivasan, R. Lai. 1990. The effects of brand loyalty
on competitive price promotional strategies. Management Sei.
Admati, A., P. Pfleiderer. 2004. Broadcasting opinions with an over 36(3) 276-304.
confident sender. Internat. Econom. Rev. 45(2) 467-498. Raymond, E. 2006. Digital cameras among consumer electronics asso
Alba, J., W. Hutchinson. 1987. Dimensions of consumer expertise. ciation's five hottest markets. http://www.digitalcamerainfo.
/. Consumer Res. 13(March) 411-454. com / content / Digital-Cameras-Among-Consumer-Electronics
Association%E2%80%99s-Five-Hottest-Markets.htm.
Avery, C, P. Resnich, R. Zeckhauser. 1999. The market for evalua
tions. Amer. Econom. Rev. 89(3) 564-584. Shi, M. 2003. Social network-based discriminatory pricing strategy.
Bickart, B., R. M. Schindler. 2001. Internet forums as influential Marketing Eett. 14(4) 239-256.
sources of consumer information. /. Interactive Marketing 15(3) Shugan, S. 2002. Marketing science, models, monopoly models, and
31-40. why we need them. Marketing Sei. 21(3) 223-228.
Biyalogorsky, E., E. Gerstner, B. Libai. 2001. Customer referral man Simester, D. 1995. Signaling price image using advertised prices.
agement: Optimal reward programs. Marketing Sei. 20(1) 82-95. Marketing Sei. 14(2) 166-188.
Bone, P. F. 1992. Determinants of WOM communication during Wernerfeit, B. 1994a. On the function of sales assistance. Marketing
product consumption. J. F. Sherry, B. Sternthal, eds. Advances in Sei. 13(1) 68-82.
Consumer Research, Vol. 19. Association for Consumer Research, Wernerfeit, B. 1994b. Selling formats for search good. Marketing Sei.
Provo, UT, 579-583. 13(3) 298-309.
Brown, J. J., P. H. Reingen. 1987. Social ties and word-of-mouth Wilson, E. J., D. L. Sherrell. 1993. Sources effects in communication
referral behavior. /. Consumer Res. 14(December) 350-362. and persuasion research: A meta-analysis of effect size. /. Acad.
Chen, Y, J. Xie. 2004. Online consumer review: A new element of Marketing Sei. 21(Spring) 101-112.
marketing communication mix. Working paper, University of Xie, J., E. Gestner. 2007. Service escape: Profiting from customer
Arizona, Tucson. http://SSRN.com/abstract=618782. cancellations. Marketing Sei. 26(1) 18-30.

This content downloaded from

183.83.158.92 on Thu, 21 Aug 2025 13:39:48 UTC
All use subject to https://about.jstor.org/terms