Market Segmentation and Information Development Costs in a Two-Tiered Information Web Site

Market Segmentation and Information Development
Costs in a Two-Tiered Information Web Site














Frederick J. Riggins
Carlson School of Management
University of Minnesota, Minneapolis, MN
friggins@csom.umn.edu
Phone: 612-624-5760
Fax: 612-626-1316


February 2002


An earlier version of this paper was nominated for the Best Paper award at the Thirty-
Fifth Annual Hawaii International Conference on Systems Sciences, January 2002.
1

---

Market Segmentation and Information Development
Costs in a Two-Tiered Information Web Site

Abstract
We develop an analytical model of a separating equilibrium for a
two-tier fee-based and sponsored-based information Web site. We
examine the monopolist’s choice of content quality and price for a fee-
based site targeted at high type consumers and the content quality level for
a sponsored site offered free to all consumers. We show how a reduction
in the potential for advertising revenues results in lower content quality on
the free site, but permits the seller to raise the fee charged to high type
consumers. We also show how differences in consumer tolerances to ads
impacts content quality, banner ad volume, and usage fees. In particular,
the seller can increase profits by making ads more attractive to either high
type consumers or low type consumers, but never both at the same time.
We show the conditions that determine which consumer segment the seller
should seek to improve ad relevancy.


2

---

Market Segmentation and Information Development
Costs in a Two-Tiered Information Web Site

1. Introduction
The Internet and the World Wide Web have altered the way many people gather
various types of information, ranging from daily news stories, to contact information for
old high school classmates, to highly technical consultancy reports. In their search for a
path to profitability, information providers have attempted various combinations of free
sponsored sites and fee-based sites. Daily news stories, sports scores, and stock price
information can be found on numerous sponsored sites such as CNN.com, Yahoo!, and
BusinessWeek online. Due to the collapse of the dot-com economy, these sites have
struggled as their potential for revenues from banner advertisers has plummeted. Other
sites such as Classmates.com, Britannica.com, and AOL.com provide some content for
free on a sponsored site, but seek to entice users to pay a subscription fee for unique
content not easily found elsewhere. For example, Classmates.com allows the non-paying
user to read a short bio of old classmates, but requires a $29.95 annual Gold Membership
fee to be able to actually contact old friends and engage in chat and discussion board
community features. Britannica.com offers some free content on a sponsored site, but
continuously prompts the user to pay a $50 annual membership fee in order to receive
full use of Britannica’s unique information archive. AOL.com functions as a competitive
Web portal, but offers various monthly pricing options whereby the user can upgrade to
the proprietary AOL interface with its many unique features. A variation of the two-tier
site model include companies such as Gartner and Forrester Research who provide some
free information on their advertisement-free company home pages, with the hopes of
enticing users to pay for other expensive services such as technical whitepapers and
industry-specific consultancy reports. In a sense, the home page functions as an
advertisement for the company itself, with the goal of building brand image and
promotion of unique information products.
For two-tier sites like Britannica.com and Classmates.com, the challenge is to
provide enough free content to keep users coming back in order to increase banner ad
revenues, but at the same time, limit the free content such that high end users will still be
3

---

willing to pay to access the premium services and information. Similarly, the consulting
firms desire to provide enough free content on their home pages such that visitors will
think highly of the firm whereby guests can be converted to clients. What these
information providers are essentially doing is degrading their information product to
create a free version of the good that satisfies low type consumers, but holding back
enough content so that high type consumers are not entirely satisfied and therefore are
willing to pay for the fee-based site [12, 13].
In this paper, we consider an information monopolist that markets its web site to
low type consumers who are not willing to pay for content on the Web, but may tolerate
some degree of banner advertising, and high type consumers who are willing to pay for
high quality content but have less tolerance for online advertising. We develop an
analytical model of a separating equilibrium where the monopolist provides a free
sponsored site to both high and low type consumers and a fee-based portion targeted at
high type consumers. The seller must select content quality and price levels for the fee-
based portion of the site, and the degraded quality level for the free sponsored site.
Firms that simultaneously cater to high-end and low-end clientele have been the
focus of much analytical modeling by researchers employing economic modeling
techniques. Based on work by Spence [14], Tirole [15], and Varian [16], information
systems researchers have analyzed the use of second degree price discrimination where
firms operate within the Internet economy to sell a high quality, expensive product to
high type consumers and a low quality, inexpensive product to low type consumers
simultaneously [2, 4, 5, 6, 10]. Because the seller is not able to know the type of a given
consumer, the seller is forced to create a separating equilibrium by setting price and
quality levels so that consumers self-select into the appropriate categories.
Much of this research examines the implications of the so-called cannibalization
problem where high type consumers may be tempted to purchase the product aimed at the
low type consumers. Moorthy and Png [8] show how this distortion in the marketplace
may result in low type consumers being under-served or in some cases not being served
at all. Riggins and Narasimhan [10] extend these results to model a firm that sells to high
and low type consumers in online channel and uses personalization technology,
community access limitations, and intertemporal price discrimination to segment the
4

---

market. Both of these studies examine the cannibalization problem for the sale of
physical goods. While several studies have examined the online pricing policies of
information goods, the implications of cannibalizing high-end information goods have
not been thoroughly examined.
The current analysis examines the cannibalization problem as it applies to
information goods within the online channel. If a firm operates a two-tier Web site where
a free, sponsored portion is offered to all consumers and a fee-based portion is targeted at
high types, the seller may face the cannibalization problem if high type consumers
receive sufficient value from the free content. We examine this problem in light of two
trends regarding online advertising. First, due to the collapse of the dot-com economy
the amount of money sites can charge for ad placement has declined in recent years,
thereby putting pressure on the revenue generating potential of banner ads. Indeed, this
trend casts some doubt on the very efficacy of the sponsored site business model. We
show how a reduction in revenue potential from ads on the sponsored site results in lower
quality of information on the free site, an increase in price or fees on the fee-based
portion of the site, and lower total profits. Second, personalization tools allow Web sites
to manage the placement of banner ads by providing impressions of ads that are more
relevant and interesting to the specific user. Therefore, it could be that users’
“discomfort” from having banner ads displayed might be lessened as personalization
increases. We show how information quality, banner ad volume, and fees change as high
and low type consumers’ tolerance for online banner ads change. In particular, we show
the conditions under which the seller will seek to make ads more relevant to either low
type consumers or high type consumers.
In §2, we discuss the relevant literature concerning the pricing of goods when the
cannibalization problem potentially exists, and recent work on the pricing of information
goods. In §3, we present a stylized example of the cannibalization problem to illustrate
its implications on the seller and both types of consumers. This is followed in §4 with the
development of our model of a separating equilibrium for a two-tier Web site where a
free, sponsored site is offered to all consumers, and a fee-based portion of the site is
aimed at high type consumers. We develop several propositions related to information
5

---

quality and pricing levels as the potential for banner ad revenue declines and as
consumers’ tolerances for banner ads change. We conclude the discussion in §5.

2. Background Literature
A seller may practice first-degree price discrimination when it faces
heterogeneous consumers and can distinguish the particular type of a given consumer.
When this occurs the seller may potentially extract all consumer surplus by charging at
each consumer’s willingness-to-pay. However, in most cases the seller is not able to
know the type of a given consumer. In addition, these consumers may not be inclined to
truthfully reveal their type. In this situation, the seller may resort to second-degree price
discrimination by offering a range of goods where consumers may be induced to reveal
their preferences by self-selecting into the appropriate category. To investigate this
problem, researchers commonly incorporate analytical models that investigate the pricing
strategy of sellers that practice intertemporal price discrimination or that sell multiple
products to multiple types of customers within a given channel.
For example, Conner [3] shows how a firm may spend aggressively on research
and development to create a new product, only to set this new product aside until the old
one is challenged by a competitor. Bensanko and Winston [1] and Levinthal and Purohit
[7] both examine intertemporal price discrimination where prices of products today are
influenced by the expectations of a new product tomorrow. The expectation of an
improved product in the future can cannibalization sales in the current market. Purohit
[9] further develops a two-period model of a manufacturer that offers a product in a rental
market and a sales market.
Dewan, et al. [4] develop a model that shows how an early entrant into the online
sales channel can use the personalization and customization features of the Internet to
create a barrier to entry discouraging other online sellers from entering the market.
Indeed, the early entrant may over-customize in an effort to block entry. In a follow-up
analysis, Dewan, et al [5] extent Salop’s circular city model [11] to develop a model of a
firm that sells a range of customized products in addition to a standard product. The
authors develop pricing and product-positioning strategies for a monopolist and a
duopoly in light of changing customization and personalization capabilities.
6

---

Moorthy and Png [8] develop an analytical model of a seller that sells a product
that can be differentiated along some measure of “quality” to both high and low types
consumers within a single channel. High type consumers are distinguished from low
types in that high type consumers are willing to pay more for a given level of quality than
are low type consumers. Because the monopolist seller is not able to know a given
consumer’s type, it must create a separating equilibrium by setting prices and quality
levels so that consumers self-select into the appropriate category. The problem is that in
order to sell to low type consumers at a price level they are willing to pay, high type
consumers will view the low product as an attractive alternative to the high quality good.
This potential for cannibalization is a distortion that ultimately harms the low type
consumers as the monopolist is forced to lower the quality of the low type product to
make it less attractive to the high type consumers. The authors further develop the model
by considering intertemporal price discrimination where the seller can segment the
market using delay in addition to quality differences.
Riggins and Narasimhan [10] extend this model to examine the case where a
monopolist markets a physical product to two types of consumers in the online channel.
They show how the seller may use personalization technology to limit or perhaps
eliminate the distortion caused by the cannibalization problem. Therefore, the seller may
provide higher product quality levels and earlier product introductions for goods aimed at
low type consumers. They show how personalization technology also lessens the adverse
effects caused when the seller cannot credibly commit to future actions. In addition,
because the seller overcomes the cannibalization problem by making sure the high type
consumer buys the “correct” product the seller should focus all of its personalization
efforts on high type consumers, and must concern itself with the level of patience of high
type consumers and not low types. They also examine three different seller strategies for
bundling the sale of the physical good with customer access to a seller-sponsored online
community. When personalization capabilities are relatively low, the seller should
segment consumers into two communities by allowing consumers to have access to the
community associated with the type of physical good that was actually purchased. When
personalization capabilities are sufficiently high, the seller should switch from the
7

---

segmented communities strategy and merge all consumers into one community to take
advantage of externality effects.
The models considered above examine the pricing strategies of traditional
physical goods. In these models, the marginal cost of producing a product of higher
quality in convex as shown in Figure 1. Because high type consumers have a higher
reservation value for a given level of quality, the seller will create two versions of the
product and sell the higher quality version to the high types and the lower quality version
to low type consumers.

Segment h
Segment l
$
Costs
q *
q *
l
h
quality
Figure 1. Consumer Reservation Values and
Marginal Costs [from (8)]


Recent analytical modeling has examined the unique characteristics of
information goods. Information goods are goods where the main value is derived from
digitized content [2]. The major portion of the costs associated with information goods is
in the development phase, rather than the production phase. Works of literature, music
recordings, software programs, and full-production movies all require a significant up
front-cost to produce the first copy of the good, but can then be duplicated for a minimal
8

---

cost [13]. Indeed, the presence of the Internet allows near costless download of countless
copies of an information good that has already been produced. In this sense the marginal
cost of producing another copy for sale is essentially zero. In contrast to the function in
Figure 1, much of this literature assumes a concave marginal cost function or simply
assumes a marginal cost of zero. In this context, Bhargava and Choudhary [2] show that
the firm’s optimal product line is dependent upon the benefit-to-cost ration of qualities in
the choice menu. Therefore, product differentiation is not an optimal strategy and the
seller should produce and sell only the highest quality product feasible. Replacing the
convex function in Figure 1 with a concave function will clearly result in the seller
maximizing profits by increasing quality as much as possible. In a similar approach
Jones and Mendelson [6] show that the seller should offer only one product quality level.
Clearly, for information goods, the marginal cost is zero for producing additional
quantity. But what about creating a higher level of quality in the second unit? Is that
costless? Consider a vendor of a software product of a certain quality. If a different
customer desires a higher level of quality, producing a perfected version of the program
can become very costly. Indeed, the cost function associated with improving the quality
of software would likely be convex as it becomes harder and harder to perfect the
program. This is because creating a higher quality version of an information good
requires the seller to return to the development phase rather than the production phase.
Indeed, the very concept of a “production phase” is misleading when considering
information goods. Any deviation from the continuous duplication of the original copy
of an information good could be considered a return to development. On the other hand,
a manufacturer of a physical good need not return to the product development phase
when a customer desires a higher quality product. Rather, the production line continues
as before, albeit with higher quality raw materials. As another example, consider the
writing of a mystery novel. One could create a version with shallow characters and a
straightforward plot (Copy 1), but the costs to create a higher quality version with well-
developed characters and an intricate plot line (Copy 2) can become a very daunting task.
The same can be said of writing academic conference and journal papers where the cost
of improving the quality of the manuscript is certainly convex.
9

---

The Bhargava and Choudhary [2] analysis lets the seller choose quality levels and
shows how the seller would choose the one of highest quality. Suppose a software maker
has already developed two versions of a software product (one of higher quality than the
other) and the software can be costlessly downloaded from the seller’s Web site. If a
consumer is willing to pay for higher quality, and one version is no more costly to
provide than the other, then only the higher quality version should be sold. Therefore, if
we disregard the initial development costs the lower quality version will not be developed
in the first place. Only the highest quality feasible will be offered, where the authors
assume the quality limit is exogenously given. By disregarding convex development
costs and considering zero or concave marginal costs, they avoid the issue of choosing
the appropriate level of quality to develop in the first place.
In reality, sellers serve a variety of consumers who have differing valuations of
quality and therefore willing to pay different amounts for a given product. In order to
meet this demand, information providers typically provide versions of their information
products where a free or low priced version may be a time-limited trial version, have
limited functionality, or may contain some time delay to make it less valuable [12, 13].
What information providers typically do is develop the high-end good and then purposely
degrade the quality of the low-end good. In our model in Section 4, we examine this
particular problem where an information monopolist develops a high end information
good based on a convex development cost function, then incurs some degradation cost to
create the low end good. The seller’s task is to select the two quality levels and the high
quality price level that maximizes the firm’s profit.

3. An Example of the Cannibalization Problem
The current analysis examines the implications of the cannibalization problem as
outlined by Moorthy and Png [8] and further expanded upon by Riggins and Narasimhan
[10]. In our model, we consider the implications of a monopolist offering a Web-based
information good. However, before considering the information good case, it would be
instructive to illustrate the classic cannibalization problem by considering an example of
a simple physical good.
10

---