Efficiency Analysis for Display Ads and Contextual Search

Efficiency Analysis for
Display Ads and Contextual Search
Yung-Ming Li
Jhih-Hua Jhang-Li
Institute of Information Management
National Chiao Tung University, Taiwan
Institute of Information Management
National Chiao Tung University, Taiwan
886-3-5712121 Ext 57414
886-3-5712121 Ext 57340
[email protected]
[email protected]
ABSTRACT
Most News Web sites provide display ads and contextual search,
which are the main ad formats in Internet. Unlike display ads,
Contextual search is a performance based advertisement and
allows clients to bid for their exposure rate. Not only do Web sites
save operation costs but also clients plan their budget flexibly.
However, in addition to accuracy of search results, contextual
search like a common resource pool will spend much more time to
take the same number of clicks as display ads do. In this paper,
we develop a simple economic model to examine the profitability
and social efficiency of contextual search under monopolistic and
duopolistic market structures.
Keywords
Display ads, Contextual search, Game theory, Pricing strategies,
Competition
question stems from the theatrical difference between pushing and
pulling. Unlike traditional media pushing messages to individuals
alternately in everyday life, people are active in Internet and pull
information through this new medium, which means that audience
have ability to read preferred content and avert their eyes from ad
trap set by advertisers [19]. Consequently, click-thru rate (CTR),
derived from dividing the number of visitors who clicked on an ad
embedded in a web page by the number of times the web page
was delivered, becomes an index for measuring the performance
of advertisement. For example, Procter & Gamble, one of biggest
advertisers in American, was the first company to contract with
Yahoo to pay ad fee grounded on CTR in 1996. Since click
behavior is useful to demonstrate online advertising effectiveness,
the first attempt for modeling such effects over a relatively short
period of time had been proposed by Chatterjee et al., which
highlights the importance of explicit consumer identification
procedures to enhance the value of clickstream data [6].
1. INTRODUCTION
The role of advertising is to deliver the message of individual
products to consumers and higher exposure for a product means
that it impresses more of the potential market [21]. According to
“Internet Advertising Revenue Report” announced by IAB in
2006 April, Internet ad revenue in the United States amounted
over $12.5 billion for the full year 2005, up from $9.6 billon
reported in 2004. The increasing number suggests that Internet is
one of most fundamental brand-building components in marketing
platforms. The Internet ad formats can be divided into search,
display-related advertising, and classifieds based on revenues
ranked from high to low, accounting for more than 90 percent of
revenues during these two years [9]. Because playing the role of
content providers attracting different types of users, a News web
site may carry display ads to offer clients more ways to reach their
target audience and specific marketing objectives. By clicking a
concerned hyperlink the clients may drive their potential
customers to their web sites, enabling them to visit again after a
few days.
Traditional ad pricing strategy used in newspaper, radio, and
television is based on cost per thousand, abbreviated as CPM,
which means the cost for showing the ad to one thousand viewers.
Web sites may guarantee an advertiser a certain number of
impressions, an ambiguous term stamped on everyone’s mind;
however, nobody can show whether it is worth the price to pay for
the number of times an ad banner is seen by visitors. This
Therefore, a new ad pricing mechanism plus contextual search
has turned into the focus of attention because of a highly costeffective feature. The rule is simple: clients set their bid for each
placement on sections or specific pages and only pay for a
qualified click when a visitor clicks on their ads. In other words,
clients can pay for performance and find their most valuable
targets interested in the unique type of content a News web site
offer. For each web page such program often represents three to
five different ads in a frame, each including title, description, and
display URL, whose size often corresponds to a large rectangle
affording to place an independent display ad. The order of
placement is determined by specific rules based on a combination
of factors, such as relevancy, bid amount, click-thru rate, and so
on. Because clients share a common resource pool, they have no
ability to know when their ads appear on appointed web pages,
but know that the chance of exposure decreases with the number
of participants if more people enter the pool. Therefore, if time
pressure caused by sharing is low or moderate, this seems a
reasonable and economic way for advertisers to control their
budget and implement their marketing plan in Internet.
Because the expense to establish and maintain an online
auction-based and pay-per-click platform is respectable, many
News web sites cooperate with other firms which concentrate on
search techniques to reduce their cost and take their profits from
each page in pre-agreed proportions. The News web sites just
copy and paste a block of JavaScript codes to their web page and
then content-related ads start working without maintaining
advertiser relationships. In addition to deliver relevant ads
precisely targeted to the content, the mechanism may filter out
improper and competitive ads. Like advertorials consumers trust,
contextual search may have higher response rates than other web
advertisements because it can provide useful information related
to content for viewers. Thus, advertisers must be careful not to
create a negative brand impression by causing consumers to feel
deceived [17].
2. Literature Review
The effect advertising is mainly determined by ad expenditure
level, time in market, number of competitors, and order of entry
[22]. To expand the entire market and win market share, Bass et
al. define advertising as “generic advertising” and “brand
advertising” according to the effect of advertising, respectively.
Their results suggest that generic and brand advertising must be
properly coordinated for preventing suboptimal allocation of
advertising budget [2]. Because media purchasing is the largest
portion of advertising spending [14], a large number of
mathematical models (e.g., game theatrical model, regression
model and so on) had been proposed.
Regarding the advertising strategies, Nguyen and Shi address
the issue of optimal advertising strategies incorporating both
market-share dynamics and market-size dynamics [16]. Shaffer
and Zettelmeyer examine how the choice of advertising and the
extent of ad exposure drive firms’ optimal strategies based on the
Hotelling model of demand [20]. Saeed et al. highlight that
advertising spending alone has only a negligible impact on firm
performance; thus, companies should provide a better productownership experience to execute effective advertising policies
[18]. Jedidi et al. measure the effects of changes in advertising
and promotion policies on sales and profits, suggesting that
advertising has a positive effect on “brand equity” in the long
term [11]. Banerjee and Bandyopadhyay construct a multistage
game theoretic model of advertising and price competition in a
differentiated products duopoly, which exhibit that advertising
spending by the large firm provides a positive externality on the
small firm’s profits [1].
Because characterized by “two-way communication” and
“conveying detained information”, Web is perceived to be very far
from most of the other media [5]. Two-way communication, also
known as interactivity, defined by Bezjian-Avery et al. et al. [3],
turns Internet into a more efficient medium than traditional media
for high involvement products [23]. Statistical data also suggest
that Internet ads are more likely conducted by companies in the
automotive-and-durables and financial-services sectors [15].
Advertisements provide a majority of revenue for advertisementsupported Web sets; however, increasing advertising may repels
visitors [8]. Therefore, allocating ads efficiently becomes a
solution to enhance revenue without hurting visitors’ feelings. For
instance, a technique report shows that experts in Microsoft
Research had developed an approach delivering banner
advertisements efficiently on the msn.com Web site, running
roughly 500 advertisements at a time [7].
Prior research suggests developing one common Web
advertisement for all users may not be the most effective way to
drive intended visitors to a certain web site [13]. One of the
fundamental questions that marketers face in advertisement is how
to implement target advertising to specific consumers and how to
allocate their media budgets [10]. In the recent studies, Kim et al.
use tree induction techniques and data-mining tools to generate
marketing rules that match customer demographics, providing
personalized advertisement selection when a customer visits an
Internet store [12]. Bhatnagar and Papatla examine the issue of
how to identify ideal paid advertising based on consumer search
behavior. They assume that consumer interest in obtaining
information on different categories can be represented by the
locations of the categories on a search continuum in
multidimensional space [4]. Similarly, we consider that the
sections of News web sites also maintain the ability to classify
particular visitors with the same interest.
3. Model
We consider an ad pricing problem where clients want to
publish their advertising on News web sites through two different
platforms: display ads or contextual search. In our assumption, the
advertising market of News web sites is dominated by a
monopolistic firm or two firms competing against each other,
which have the same traffic volume. For acquiring maximum
profit, these firms determine which services they provide and
make a proper pricing decision. We assume that each client has an
adequate budget and carries individual ad on one News web site
at most. Thus, let d denote the total number of clients buying
display ads, and s the total number of clients buying contextual
search. For the sake of analytical convenience, the market size is
normalized to one, which means d  s  1 . Furthermore, in this
paper we consider that the most value of web advertising stems
from the number of times visitors click on an ad; therefore, we
omit “impression” utility in the following discussion.
3.1 Client utility functions
To begin with, we assume that in the market News web sites
have sufficient contents to serve all clients who want to publish
their ads on the News web sites; in addition, based on traffic
volume, these News websites claim that the expected number of
times visitors click display ads in a unit time interval is N . Then,
we denote  as the “click” utility each client derives from using
display ads in this unit time interval. Thus, a News web site may
charge pd for display ads (e.g. a banner ad) for the fixed period. If
a client pays pd to carry her ad on the News web site, everyone
will touch it every time once they enter a specific web page
containing it during the period. In contrast, News web sites
providing contextual search may charge p for one-click.
Therefore, in order to receive the “click” utility as the same as that
received by clients using display ads, ones using contextual search
must pay ps , where ps  N  p . In addition, the clients using
contextual search must spend much more time to derive the same
“click” utility than that using display ads because their ads appear
alternatively and search results may not agree with contents.
Therefore, a typical client i faces heterogeneous time cost wsi
caused by s clients sharing the common resource pool when she
carries advertisement through contextual search, where
variable i stands for individual sensitivity on the value of time,
and is uniformly distributed over an interval [0,1] . A client i with
higher value of i has more time pressure than others whose levels
of sensitivity are less than hers. The parameter w reflects the
degree of relevancy search engine can achieve and decreases with
the accuracy of relevant target content. The utility function of
each client is given by

   pd
Ui  

   ps  wsi
def
display ads
(3.1)
contextual search
survive (i.e., positive profit) even if their competitors operating
contextual search lower price to zero.
3.3 Monopolistic Market
In a monopolistic market a firm operating News web site may
choose to provide a single platform or hybrid platform. If only
running display ads, the firm will set the price at  , extracting all
surplus of the clients. Each client accepts the price ( dm  1 ) and
firm’s profit is  dm    cd . In contrast, if only offering
contextual search, the firm will set the price at   ws2 , and the
profit function becomes
3.2 Platform demand functions
In this section we consider three types of platform configuration:
display ads, contextual search, and hybrid ads. First, if only
display ads are provided, all clients pay ad fees as long
as pd   holds. Second, if only contextual search is provided,
because the maximum value of i is equal to the total number of
clients
purchasing
contextual
search,
clients
  

Max  sm     w sm
psm

2
m
s

won’t carry their ads through the platform. Thus, the demand
Thus, firm’s profit is given by
   ps 
d
m m
s , ps
w . Last,
if both display ads and contextual search are offered, the marketsegmentation condition is given by   p    p  w ˆ ,
s
s
where the utility with the index of time pressure ˆ a client derives
is indifferent between these two services. Consequently, client i
with   ˆ will choose contextual search; otherwise, she select the

d
s
where ˆ is given by
, pd  ps  w
1,

 p  ps
ˆ   d
, 0  pd  ps  w
w

0
, pd  ps  0


s 
pd  ps
.
w
(3.5)
3
 sm   2 4 27w .
(3.6)
Then we consider a hybrid platform, where price decisions of the
two platforms are taken simultaneously by a monopolistic firm.
Thus, the profit-maximization problem of the monopolistic
provider is written as
   pdm  psm  w 

Max
 m  pdm  cd 1 
m m d s
pd , ps
 psm

pdm

psm
w
.
(3.7)
Since both demand functions of display ads and contextual search
are associated with the difference of these two platform prices
dividing by the accuracy of contextual search, denoted
(3.2)

as p  pdm  psm

w , instead of solving the first order condition
with respect to pdm and psm , we rewrite the objective function as
Thus, the demand function of these two platforms are given by
pd  p s
and d  1 
w
(3.4)
 3w ,2 3 .
i
other platform. Let d and s denote the total number of clients
who prefer to display ads and contextual search, respectively. The
market share for ads can be derived from   1  ˆ and   ˆ ,
3
Solving  sm sm  0 , the optimal price and market share for
contextual search is given by
with i2     ps  w utilize contextual search and the others
function of contextual search is given by  s 
 .
 sm  w sm

(3.3)
Subsequently, we assume average marginal operations cost per
client is cd for serving a client using display ads. For example,
clients and editors of News web sites may discuss how to prepare
adaptable content to fit for their ads. In contrast, because the
workflow of contextual search is nearly automatic, the impact of
the number of clients on variable cost is sufficient small ( cs  0 ),
compared to display ads. Therefore, we assume that the variable
cost of contextual search is zero. To ensure that the participation
and incentive-compatible conditions hold, we assume
that   pd  cd , pd  ps , and   cd . We also
assume w  cd such that firms maintaining display ads may

 
Max  dm s  pdm  cd 1  p  pdm  wp
pdm , p

p .
(3.8)
It can be easily observed that the joint profit is linearly increasing
with pdm , and  dm s is concave on  p . Solving the first order
condition  dm s p  0 , the optimal price offset is given
by pdm  psm  cd 3 , where p  cd 3w . Then, the price of
display ads
pdm
should be as high as possible such that the
maximal profit is reached. Hence, the pricing schemes and
corresponding market shares in the hybrid platform are given by


m m
s , ps
m m
d , pd
 c
  1 
d

3w ,  3  cd  3 and

cd 3w ,  ,
(3.9)
yielding a total profit of
 dm s
1
5
cd
w
dc   4  1  5

 2 cd 
.
   cd 1 
 3 3w 


(3.10)

 ,

(3.16)
3
 sc 
c
2w 
1  1  5 d
125 
w

 , and

 dc 
more profit from a hybrid platform than a single one. (c) In a
monopolistic market, the lowest allowable bid in a hybrid
platform is larger than that in a single one.
c
3w 
1  1  5 d
125 
w



c
 d
5

c
 4  1  5 d
w

3.4 Duopolistic Market
If the variable cost of display ads is very small  cd  0  , profit
In this section we examine the pricing scheme of ad platforms
instituted by independent profit-seeking firms, which own and
operate different ad platforms. The business environment for
display ads and contextual search is largely determined by relative
market power between these two firms. We consider three cases of
market power distribution in non-cooperative pricing dynamics.
In the first case both firms offering different platforms participate
in a simultaneous Bertrand pricing competition game, each of
which with symmetric market power. In the subsequent cases we
consider a market structure that one of the firms gets the
leadership of pricing decision; in other words, these two firms
participate in a Stackelberg (leader-follower) pricing competition
game.
levels of both platforms can be simplified to
Proposition 1. (Monopolistic market) (a) Contextual search will
be more profitable than display ads for a monopolistic firm if the
accuracy of contextual search is sufficiently high, where
w  4
3
27    cd  . (b) The monopolistic firm always makes
2
 sc 
2

c
 4  1  5 d
w




(3.17)


.
(3.18)
16
36
w and  dc 
w.
125
125
(3.19)
Proposition 2. (Simultaneous price competition) (a) The market
share of contextual search (display ads) increases (decrease) with
the variable cost cd of display ads and the accuracy of search
results. (b) If the variable cost cd of display ads is sufficient small,
the profit level of display ads is larger than that of contextual
search.
3.4.2 Sequential pricing competition
3.4.1 Simultaneous pricing competition
In a simultaneous price competition environment, both firms
maintaining different platforms attempt to create their maximal
profit. Let  d denote the profit function of display ads, and  s the
c
c
profit function of contextual search, respectively. Thus, the profit
functions of these two firms are given by
 sc  pscsc  psc

p
 psc
c
d

w and
(3.11)

 dc  pdc  cd dc



 pdc  cd 1 

p
c
d
 psc


w

.
(3.12)


.


cd
cd
w


4  10  4 1  5
25 
w
w



.


(3.13)
cd
w

 ,


Solving first order condition of display ads directly yields the
subgame perfect Nash equilibrium results as follows:
c
(3.14)
(3.15)
(3.20)
where dcds  1  pdcds 3w .
pdds 
Hence, market share and profit levels can be derived from using
(3.3), which are given by
1
5

 dcds  pdcds  cd dcds ,
cd
cd
w

pdc 
6  15  6 1  5
25 
w
w

sc  1  1  5

 
response function of contextual search, pscds , and rewritten as
Solving  sc psc  0 and  dc pdc  0 simultaneously yields
equilibrium prices given by
psc
Instead of pricing simultaneously, we assume that which firm
makes pricing decision first is dependent on its own market
power. Therefore, we assume that the firm operating display ads
offers a price first. Then, the firm operating contextual search
makes its pricing decision after observing the offer. Thus, the best
response function of contextual search to display ads is the same
as
that
results
from
simultaneous
decision,
i.e.,
2 cds
cds
cds
ps  R pd  pd . Utilizing backward induction approach,
3
the profit function of display ads is given by plugging in the best
2
c
w
1  1  d
3
w

c 
2w 
c
 , ps ds 
1  1  d 
9 
w 

dcds   2  1 

1
3
cd
w

c
1
c
 , s ds  1  1  d
3
w

1
9
cd
w
2


c 
 

  w 1  1  d   3cd  ,
w 





 dcds   2  1 


 ,

2
(3.21)
(3.22)
(3.23)
dependent on the accuracy of contextual search (the variable
3
c 
2w 
1  1  d  .
27 
w 
and  scds 
(3.24)
If the variable cost of display ads is very small  cd  0  , then
firms’ profit levels become
 scds 
16
4
w and  dcds  w .
27
9
(3.25)
Proposition 3. (Sequential pricing competition, display ads
first) (a) Both the prices of these two platforms will increase with
the variable cost cd of display ads and decrease with the accuracy
of contextual search. (b) The market share of contextual search is
larger than that of display ads. (c) The profit of contextual search
is large than that of display ads if the variable cost cd of display
ads is sufficient small.
Next, we assume that the firm providing contextual search has
market power, drawing up its pricing strategy first. Similarly,
solving first order condition  dcsd pdcsd  0 yields
c
pdsd 

 c  pcsd
w
s
1  1  3 d

9 
w


Max  scsd  pscsd  scsd
pscsd

   pscsd .
 


csd

1  1  3  cd  ps


3 
w


(e.g.,
p
p
csd
s
(3.26)

.
 

(3.27)
), and grabbing its clients. In other words,
the firm maintaining contextual search may charge cd N for the
lowest allowable bid per click to make maximal profit.
Consequently, equilibrium price, market share, and the levels of
profit are given by
c
pdsd  cd 
4
w , pscsd  cd ,
9
dcsd  1 3 , scsd  2 3 ,
 dcsd 
4
2
w , and  scsd  cd .
27
3
W  d  s 
U .
(4.1)
i
4.1 Monopolistic market
First, consider a monopolistic firm operating and owning a
single platform. Suppose that the monopolistic firm decides on
either maintaining display ads or contextual search. Then, social
welfares are given by
Wdm    cd , and
m m
2
the firm providing contextual search should keep pscsd  cd to
preventing firm providing display ads from undercutting
csd
d
In this section, we analyze the economic efficiency (social
welfare) of different strategies adopted by firms in monopolistic
and duopolistic markets, comparing them with socially optimal
results. Social welfare of market is defined as summation of the
firms’ profit and the clients’ surplus, which are given by
m
We find that  scsd pscsd  0 always holds, which means that the
profit level of contextual search increases with its price. However,
c
pdsd
4. Analysis of Economic Efficiency
Ws  ps  s 
Hence, the profit maximization program of contextual search is
given by
pscsd
cost cd of display ads).

s
(4.2)
m
0
m
m
  ps  w s  d 
5w 
2



3w



3
(4.3)
Now suppose that a hybrid platform is offered by the
monopolistic firm, the social welfare of which is given by

m

m
m
m m
Wd  s  pd  cd  d  p s  s 
   cd 
5w 


2
cd
3w




s
0
m
m
m
  p s  w s  d
(4.4)
3
4.2 Duopolistic market
In a duopolistic market, the function of social welfare is given by
Wd  s   pd  cd d     pd d  ps s


s
0
  ps  w s d
    cd  
w  s 
2
(4.5)
3
 cd s .
This equation implies that the social welfare of either
simultaneous pricing competition or sequential pricing
competition can be evaluated directly by market share of
(3.28)
contextual search and variable cost cd of display ads. Thus, the
(3.29)
social welfare gained with respect to different pricing strategies is
derived as follows:
(1) Simultaneous pricing competition
(3.30)
Proposition 4. (Sequential pricing competition, contextual
search first) (a) The market share of contextual search is twice as
much as that of display ads. (b) The price of display ads decreases
with the accuracy of search engine while the price of contextual
search keeps the same level as the variable cost cd of display ads.
(c) The profit level of display ads (contextual search) is only
Wdc s     cd  
where
 sc
 
w  sc
2
3
 cd sc ,
c
1
 1  1  5 d
5 
w

 .

(2) Sequential pricing competition, display ads first
(4.6)
Wdcdss
    cd  
where  scds
 
w  scds
3
 cd scds ,
2
c 
1
 1  1  d  .
3 
w 
.
(4.7)
w
(3) Sequential pricing competition, contextual search first
c
Wd sds
    cd  
where scsd 
 
3
w scsd
 cdscsd ,
2
.
(4.8)
2
.
3
4.3 Socially planned market
In a socially planned market, social welfare of display ads alone
and its socially optimal market share are given by
Wdw    cd and dw  1 , where cd  pdw   .
(4.9)
Similarly, social welfare of contextual search alone is given by
Wsw  pswsw 

sw

sw
0
  psw  wsw d
(4.10)
 
3
1
 w sw
2
Solving first order condition Wsw sw  0 yields socially
optimal market share, leading to pricing strategy and social
welfare in a planned market, which are given by
 sw 
 2
2
, psw   3 , and Wsw  w 
 3w
3w

3

 , respectively.

(4.11)
Last, we consider a socially planned market providing both
display ads and contextual search, the function of social welfare
of which is given by




Wdw s  pdw  cd dw    pdw dw  pswsw


 sw
0
  psw  w sw d


   cd 1   sw 
(4.12)
 
1
w  sw
2
3
Socially optimal market share, the prices, and welfare of the
market
are
derived
from
solving
first
order
condition Wdw s sw  0 and using the demand functions (3.3),
which are given by
 sw 
2
2cd
, pdw  cd  psw , and
3
3w
3
Wdw s
 2c  2
   cd  w  d  , respectively.
 3w 
market operating contextual search alone grabs more clients and
charges less than a monopolistic market. (c) The social welfare of
planned market (monopolistic market) providing contextual
search alone is larger than that providing a hybrid platform when
the accuracy of search results is sufficiently high, i.e.,
(4.13)
Proposition 5. (Efficient platform allocation) (a) The excellent
search technique is beneficial to social welfare in all markets if
there exits a contextual search platform. (b) A centrally planned
8  1.5  c1.5
d


27    cd 
2
2
(w

25  1.5  c1.5
d
108    cd 

2
2
).
5. Conclusions
In this paper, we have developed an economic model to study
the competence between display ads and contextual search in
either monopolistic or duopolistic market. In a monopolistic
market contextual search turns into a powerful weapon to make
respectable revenue for firms concentrating on search technique,
threatening the position of traditional display ads. However, in an
integrated market contextual search can be used as a vertical
product differentiation strategy to improve profit when the
variable cost of display ads exists and the results of contextual
search is not accurate enough. Comparing the profitability and
social welfare with respect to pure strategies (display ads or
contextual search) and hybrid strategy under monopolistic and
social planned market, we suggest that hybrid strategy is better
than pure strategy; however, maintaining contextual search alone
is the best strategy if search technique is prominent.
In duopolistic market, market share of contextual search ranges
from 40% to 80% dependent on which pricing scheme is
launched. It is noteworthy that the profit level of contextual search
won’t increase with the quality of search engine when the variable
cost of display ads is sufficiently small. This effect results from
the fact that the firm maintaining display ads will lower price such
that both firms’ profits are reduced. Comparing social welfare
with respect to different pricing schemes, we find that when the
ratio of the variable cost of display ads to the quality of contextual
search is close to one, sequential pricing with contextual search
first makes maximal social welfare in the market; however, if this
ratio is close to zero, simultaneous pricing does.
In our model we shed the “impression” utility and assume that
the parameter representing the quality of contextual search is
larger than the variable cost of display ads. However, if the case
reverses and firms proving display ads alone survive in a
competitive market, the “impression” utility should remain
valuable for some particular clients such that these clients still
purchase it. On the other hand, News web sites providing
contextual search, in fact, manipulate a starting price that serves
as both the lowest allowable bid and the minimum payment for
one-click. For simplicity in our model we assume that each client
exploiting contextual search pay the same price, equal to the
lowest allowable price announced by a News web site. Therefore,
if each client bids based on her budget and individual valuation,
this question will relate to Vickrey auction; thus, it is an
interesting topic for future research to study market share in the
environment where the exposure rate of ads is determined by bid
value.
6. REFERENCES
[1] Banerjee, B. and Bandyopadhyay, S. Advertising
Competition Under Consumer Inertia. Marketing Science,
22, 1 (2003), 131-144.
[2] Bass, F. M., Krishnamoorthy, A., Prasad, A., and Sethi, S. P.
Generic and Brand Advertising Strategies in a Dynamic
duopoly. Marketing Science, 24, 4 (2005), 556-568.
[3] Bezjian-Avery, A., Calder, B., and Iacobucci, D. New media
interactive advertising vs. traditional advertising. Journal of
Advertising Research, 1998, 23-33.
[4] Bhatnagar, A. and Papatla, P. Identifying Locations for
Targeted Advertising on the Internet. International Journal of
Electronic Commerce, 5, 3 (2001), 23-44.
[5] Calisir, F. Web advertising vs other media: young
consumers’ view. Internet Research: Electronic Networking
Applications and Policy, 13, 5 (2003), 356-363.
[6]
Chatterjee, P., Hoffman, D. L., and Novak, T. P. Modeling
the Clickstream: Implications for Web-Based Advertising
Efforts. Marketing Science, 22, 4 (2003), 520-541.
[7] Chickering, D. M. and Heckerman, D. Targeted Advertising
on the Web with Inventory Management. Interfaces, 33, 5
(2003), 71-77.
Storefronts. International Journal of Electronic Commerce, 5,
3 (2001), 45-62.
[13] [13] Korgaonkar, P. and Wolin, L. D. Web usage,
advertising, and shopping: relationship patterns. Internet
Research: Electronic Networking Applications and Policy,
12, 2 (2002), 191-204.
[14] Krishnan, T. V. and Jain, D. C. Optimal dynamic Advertising
Policy for New Products. Management Science, 52, 12
(2006), 1957-1969.
[15] [15] Lace, F. M. At the crossroads of marketing
communications and the Internet: experiences of UK
advertisers. Internet Research, 14, 3 (2004, 236-244.
[16] Nguyen, D. and Shi, L. Competitive Advertising Strategies
and Market-Size Dynamics: A Research Note on Theory and
Evidence. Management Science, 52, 6 (2006), 965-973.
[17] Riquelme, H. and Kegeng, W. The unintended effects of
hidden assumption: biases on the internet. Online
Information Review, 28, 6 (2004), 444-453.
[18] Saeed, K. A., Hwang, Y., and Grover, V. Investigating the
Impact of Web Site Value and Advertising on Firm
Performance in Electronic Commerce. International Journal
of Electronic Commerce, 7, 2 (2002), 119-141.
[8] Dewan, R. M., Freimer, M. L., and Zhang, J. Management
and valuation of Advertisement-Supported Web Sites.
Journal of Management Information Systems, 19, 3 (1992),
87-98.
[19] Schwartz, E. I. Webonomics : Nine Essential Principles for
Growing Your Business on the World Wide Web. 1997.
[9] Internet Advertising Revenue Report.
http://www.iab.net/resources/ad_revenue. asp.
[21] Soberman, D. A. Research Note: Additional Learning and
Implications on the Role of Informative Advertising.
Management Science, 50, 12 (2004), 1744-1750.
[10] Iyer, G., Soberman, D., and Villas-Boas, J. M. The Targeting
of Advertising. Marketing Science, 24, (3) 2005, 461-476.
[11] Jedidi, K. Mela, C. F., and Gupta, S. Managing Advertising
and Promotion for Long-Run Profitability. Marketing
Science, 18, 1 (1999), 1-22.
[12] Kim, J. W., Lee, B. H., Shaw, M. J., Chang, S. H., and
Nelson, M. Application of Decision-Tree Induction
Techniques to Personalized Advertisements on Internet
[20] Shaffer, G. and Zettelmeyer, F. Advertising in a Distribution
Platform. Marketing Science, 23, 4 (2004), 619-628.
[22] Vakratsas, D., Feinberg, F. M., and Bass, F. M. The Shape of
Advertising Response Functions Revisited: A Model of
Dynamic Probabilistic Thresholds. Marketing Science, 23, 1
(2004), 109-119.
[23] Yoon, S. J. What makes the Internet a Choice of Advertising
Medium? Electronic Markets, 11, 3 (2001), 155-162.