Group Buying on the Web: A Comparison

of Price-Discovery Mechanisms
Krishnan S. Anand Ravi Aron
Operations and Information Management Department, The Wharton School, University of Pennsylvania,
Philadelphia, Pennsylvania 19104
anandk@wharton.upenn.edu raviaron@wharton.upenn.edu
W
eb-based group-buying mechanisms are being widely used for both business-to-
business (B2B) and business-to-consumer (B2C) transactions. We survey currently
operational online group-buying markets, and then study this phenomenon using analytical
models. We build on the literatures in information economics and operations management
in our analytical model of a monopolist offering Web-based group-buying under different
kinds of demand uncertainty. We derive the monopolists optimal group-buying schedule
under varying conditions of heterogeneity in the demand regimes, and compare its prots
with those that obtain under the more conventional posted-price mechanism. We further
study the impact of production postponement by endogenizing the timing of the pricing and
production decisions in a two-stage game between the monopolist and buyers. Our results
have implications for rms choice of price-discovery mechanisms in e-markets, and for
the scheduling of production and pricing decisions in the presence (and absence) of scale
economies of production.
(Group Buying; Pricing; Demand Uncertainty; Quantity Discounts; Price Discovery; Demand
Heterogeneity)
1. Introduction
Group-buying schemes have been in vogue for many
years, particularly in the context of selling on televi-
sion via the popular Home Shopping Network. Web-
based variants of group buying have recently received
a lot of attention as part of the wave of innovative
online market-based mechanisms such as auctions,
reverse auctions, and Pricelines name-your-own-
price scheme. While there is a rich history of analyt-
ical research in auctions, spanning at least 40 years,
we believe that this paper is the rst analytical model
of group buying.
To illustrate how group buying works, consider
the sale of a particular productsay the Compaq
iPaq 3650 (a personal digital assistant)by a Web-
based group-buying operator. Suppose the retail price
of the iPaq 3650 is $500. The seller (group-buying
operator) might announce a base unit price of $500.
To entice buyers, the seller offers to sell at the dis-
counted unit price of $480 if at least three units are
demanded, which drops to $450 each if at least ve
units are demanded, and $400 per unit if demand is
for eight units or more. In effect, quantity discounts
are offerednot on an individual customers order,
but on the total of all customer orders. The seller publicly
announces, and commits to, his pricing scheme and a
closing date for the sale. Further, the sales statusthe
current number of bidders and the imputed price
is updated dynamically and displayed on the Web.
Thus, buyers with different valuations for the prod-
uct can follow the sales trajectory and jump in with
a bid at an opportune moment. Bids are rm com-
mitments (reneging on bids is not allowed), and the
products current price acts as a price ceiling for the bid-
der. Eventually, all winning customers get the product
at the same clearing price, irrespective of what their
actual bids were. In contrast to auctions and most
other dynamic pricing mechanisms, there is no ex post
Management Science 2003 INFORMS
Vol. 49, No. 11, November 2003, pp. 15461562
0025-1909/03/4911/1546
1526-5501 electronic ISSN
ANAND AND ARON
Group Buying on the Web
valuation-based price discrimination. Because prices
fall with the number of bids, customers who bid early
induce other customers to bid, leading to a cascade
of more bids and lower prices. In turn, the seller can
stimulate demand and increase his revenues by the
judicious use of discounts. Thus, in principle, both cus-
tomers and the seller are better off, leading to Pareto-
improving welfare gains.
The remainder of this paper is organized as follows.
In the next section, we lay out our research objectives.
Section 3 provides a comprehensive review of the cur-
rent state of praxis by surveying a variety of group-
buying practices, worldwide, in the B2B, B2C, and
nonprot sectors. Section 4 reviews the relevant prior
research. In particular, two streams of researchthe
quantity-discounts literature in operations manage-
ment and the literature on price discovery under
demand uncertainty in information economicsserve
as the building blocks for our analysis. In 5 and 6,
we develop analytical models of group buying under
different kinds of demand uncertainty and compare
the performance of group buying versus posted pric-
ing in each case. We derive the conditions under
which each dominates. In 7, we discuss the manage-
rial implications of our research and highlight possible
extensions.
2. Research Objectives
Group-buy operators seek to aggregate disparate buy-
ers (who can operate remotely and asynchronously)
via the Web by providing them price-based incentives
for volume purchases. Kauffman and Wang (2001,
2002),
1
the only previous studies by academics to
focus on group buying, empirically analyzed pat-
terns in bidder behavior using customer data from
MobShop.com. Our section on Praxis (3) provides
a survey (up to October 2002) of worldwide group-
buying practices. Further, our analysis focuses on
answering two key issues of immediate concern to
current practice: (1) Given that a rm uses group
1
An extensive list of references from both the trade and academic
literatures is available in our online appendix at mansci.pubs.
informs.org/ecompanion.html.
buying to sell its products, what is its optimal group-
buying schedule?
2
(2) How does the performance of
this optimal group-buying schedule compare with
that of simple (and widely used) posted prices, and
what market conditions and product characteristics
would vindicate the use of group buying? We believe
that this paper is the rst to study group buying using
rigorous analytical models.
3. Group Buying: The Current
State of Praxis
We quote the mission statements of two of the market
operators that we surveyed to illustrate that demand
aggregation and volume discounting are at the core
of group buying.
Co-buying is co-operative shopping for the 21st
century. Its a really simple way of getting better
value by bringing people together via the Internet.
By bringing together as many members as possible,
LetsBuyIt.com can negotiate lower prices with mer-
chant partners (also referred to as suppliers) or
manufacturers. The more people, the lower the prices.
LetsBuyIt.com (2003)
The more we are, the bigger the negotiation power
towards trusted suppliers. The immediate result?
Prices that fall without having to negotiate on your
own. The more we are (many), the happier we are
(happy) because everyone pays less and everyone
benets from a better service.
HappyMany (2003)
In our study of currently operational,
3
online mar-
kets, we found that group buying is a widely
deployed price-discovery mechanism in a variety of
markets and contexts. We will see that the group-
buying mechanism is extensively used in the United
States, Europe (including Germany, France, and the
United Kingdom), and Asia (including in Egypt,
India, Singapore, and Thailand) for many different
product categories. This review discusses interest-
ing and representative examples of this deployment.
2
Under group buying, the seller offers a menu of price-quantity
tuples, one of which is eventually realized. For brevity, we refer
to the price-quantity schedule offered under group buying as the
group-buying schedule.
3
As of October 2002.
Management Science/Vol. 49, No. 11, November 2003 1547
ANAND AND ARON
Group Buying on the Web
Given the large number of active group-buying sites
(more than 50), an exhaustive review is beyond the
scope of this paper. However, a fairly comprehen-
sive list of 26 representative group-buying websites
from the B2B, B2C, and nonprot sectors is provided
in the online appendix (at mansci.pubs.informs.
org/ecompanion.html).
Group-buying schedules are employed for branded
consumer products and intangible services such as
bandwidth and network security, in both B2B and
B2C markets, and in public and private markets.
These markets are characterized by the belief that
both suppliers and buyers stand to benet through
group buying. Evidence of this belief is adduced by
the websites of both buyer (IRPG, APPA, TBG, GBP)
4
and supplier (LBI, THM) consortia in a variety of
markets that prominently tout the benets of group
buying to their members.
The successful U.S.-based group-buying site
e.conomy, operated by PricewaterhouseCoopers,
allows buyers to purchase indirect goods and ser-
vices such as ofce supplies, temps (temporary
stafng services), furniture, commercial print, com-
puter hardware and software, telecommunications
and connectivity, and company travel at substan-
tially reduced prices that result from higher purchase
volumes (ECON). Members are assured of an initial
maximum price (a price ceiling). As they place
orders, incrementing the total volume demanded, the
resulting price declines are broadcasted until the mar-
ket clears at a predetermined time. LetsBuyIt.com,
quoted in the beginning of this section, is a consumer
group-buying market currently functional in the
United Kingdom, Germany, and France, that works
similarly. A variation on the standard group-buying
mechanism that this company permits is that buyers
can choose not to declare a price ceilingbuyers
greater risk without a price ceiling may be offset by
guaranteed availability of the good at the common,
lowest price. In the words of LetsBuyIt.com, Before
you join a cobuy, you need to decide whether you
would like to buy at the current pricethe price
4
Acronyms refer to websites of companies listed in Group-Buying
Markets at the end of 3. For an extended list of academic and
trade references, refer to the online appendix.
reached when the cobuy closesor at the Best Price
only. If you choose to buy at the current price, you
will receive the product and you will pay the closing
price. If you choose to buy at the Best Price and the
Best Price is not reached (i.e., the required number
of participant cobuys is not reached), you will not
receive the goods. Your order will be cancelled and
you will not be charged. Product categories sold in
this market include exercise equipment, consumer
electronics, sport and leisure, food and wine, and
jewelry.
The Group-Buying Partnership in the United
Kingdom is a consortium of small- and medium-sized
businesses that negotiates volume discounts from
suppliers of items such as electricity and gas, com-
puter equipment, ofce supplies and vehicles, and
passes these discounts on to buyers (GBP). They claim
on their website that group buying allows small
to medium sized businesses to combine their indi-
vidual buying power for the overall benet of each
individual member (GBP). The mechanism works
through a process of demand aggregation wherein
buyers commit to purchase quantities subject to cer-
tain price ceilingsshould the consortium succeed in
procuring the quantities at or below these price ceil-
ings, buyers are then obliged to follow through on
their commitments. This is a direct correlate of our
model of purchasing commitment conditional on demand
realization (6 of this paper). A different avor of
this practice can be found in the India-based market
Chennai Online (COL) whose customers are redis-
tributors. This private electronic group-buying market
allows buyers to place orders by specifying price-
quantity schedules where purchase is required only
if the supplier is able to meet the price specied
by the buyers. What is interesting in this context
is the order of events that lead to price discov-
ery. In the case of products such as software suites
for ofce use, magazines and periodicals from the
popular press and computers (servers and PCs),
the supplier announces prices at different levels of
demand (price-quantity schedules) and the buyers
place bids serially until the market clears at a particu-
lar price. In a second category of goods that includes
kitchen appliances, computer peripherals, consumer
electronics, and electrical appliances, the sequence
1548 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
of demand solicitation and production (or procure-
ment) of the goods is reversed. In this case, suppliers
commit to a group-buying schedule, and demand is
realized when a stable price-quantity tuple is reached.
Then, suppliers either produce the goods (kitchen
appliances and computer peripherals) or buy them
from third parties (consumer electronics and electrical
appliances) in exact quantities. Thus, while production
and inventory stocking decisions precede pricing in the
rst category (modeled in Case 1, 6 of this paper),
the second category is an example of production post-
ponement (Case 2, 6, model).
Another version of the group-buying mechanism
called StockBuzz (based in Thailand) is operated
by Asian producers of high-quality yarn who sell
their products to manufacturers of high-end apparel
(STBZ). The apparel manufacturers forecast demand
based on a number of macroeconomic factorspast
sales data, seasonality, and most importantly, retailer
feedback. Because most of the manufacturers sell to
the same retailers and in the same urban markets,
their demand estimates tend to be highly correlated,
and reect consumer demand cycles. The yarn pro-
ducers, in turn, face uncertainty in demand, which
tends to uctuate between a robust (high demand)
regime and a weaker demand regime reective of an
adverse consumer demand cycle. Thus, the yarn pro-
ducers estimate the probabilities of these two market
states, based on prior seasonal data and other avail-
able information. Using the two estimated demand
schedules as guidelines, they commit to a schedule
of prices for different quantities with prices declin-
ing in quantities. Buyers (apparel manufacturers) are
allowed to place conditional orders, which they are
then obliged to honor should the price reach the levels
specied in the order contract. In this case, yarn sup-
pliers use group buying as a response to uncertainty
about which of the two states of demand will be real-
ized (modeled in Cases 1 and 2, 5 of this paper).
Group-buying mechanisms are also observed in
specialized markets. In the hotel and restaurant
industry, the Independent Restaurant Purchasing
Group (IRPG) brings together more than 3,000 restau-
rants across the United States and negotiates volume-
based discounts with suppliers on their behalf.
IRPG argues that the benets of its group-buying
scheme, called the IRPG Volume Rebate Program,
for its members are twofold. First, buying power is
enhanced for large chains and small restaurants: Our
national and regional contracts are based on the total
volume of our group versus that of an individual
restaurant. No matter how many restaurants you
have, the IRPG can add strength and stability to
your purchasing power. Second, many manufac-
turers insist on minimum volumes for doing busi-
ness with them, out of efciency considerations
arising from their production and transaction scale
economies. Smaller restaurants often nd it difcult
to meet these minimum criteria. As IRPG argues,
Most restaurants, by themselves, would probably not
meet the manufacturers minimum volumes. How-
ever, when combined with those of the entire mem-
bership, they exceed the minimum, and in many
instances, qualify for further multi-unit-allowances
normally only available to large chain accounts. In
the skin care product market, the market operator The
Buying Group (TBG) operates a relatively straightfor-
ward group-buying scheme where buyers commit to
quantities at a maximum price. The market operator
then negotiates a price from suppliers for the aggre-
gate quantity, and shares the benets of the lower
prices with the buyers (here again, we see procure-
ment postponement, studied in 6 of this paper). A
noteworthy feature seen in this market is the preorder
option, in which buyers express an interest in a prod-
uct that is not currently featured in the market, and
pay for their desired quantities up front, at prespec-
ied prices. If there is enough interest, the market
operator (in this case, The Buying Group) attempts to
procure the product at a price that will make trade
possible for at least some buyers. If the market opera-
tor succeeds, he charges all buyers (who are willing to
pay the clearing price or more) the same oor price.
If trade is not possible, the market operator refunds
the buyers paid-up amounts. The requirement that
buyers commit to a quantity and price by paying the
amounts up front ensures that buyers do not make
frivolous bids, and also provides an incentive for the
market operator to negotiate with suppliers.
Branded consumer goods are also sold via the
group-buying mechanism. Two companies, McNopoly
(MPLY) and Online Choice (OLC), operate markets
Management Science/Vol. 49, No. 11, November 2003 1549
ANAND AND ARON
Group Buying on the Web
that allow consumers to bid for specic branded items
from a menu of offered items. If their suppliers offer
discounts that support a market-clearing price, those
consumers whose bids for the products were higher
than the market-clearing price are required to buy the
products; the others exit the market.
In the nonprot sector, the Maryland Public Service
Commission (MPSC) supports individual buyers in
their efforts to form buying groups to negotiate
better rates for higher aggregate levels of power
consumption. Similarly, the American Public Power
Association (APPA) and the Environmental Action
Foundation have launched a number of group-buying
initiatives that bring together consumers to negoti-
ate lower rates with power utilities across the United
States.
To summarize, we nd evidence of widespread de-
ployment of group buying as a price-discovery mech-
anism in both B2B and B2C sectors. A key feature that
sets all group-buying schemes apart from other mar-
ket mechanisms is suppliers beliefs that getting buy-
ers to precommit to a price-quantity schedule where
the prices are monotonically declining in total pur-
chase quantities (and not just an individual buyers
purchase quantities) will maximize supplier revenues
by inducing greater buyer demand. Thus, group buy-
ing is often targeted toward buyers with low bar-
gaining powerindividual consumers or small to
medium-sized businesses. In fact, many companies
across all types of products, services, and coun-
tries, explicitly and prominently claim that enhanced
buyer bargaining power is the single biggest advan-
tage of group buying. The group-buying intermediary
Printing Industries of New England (PINE) asserts on
its website, PINE has made arrangements with sev-
eral companiescompanies you use in running your
businessto provide discounts on their products and
services. PINE combines the overall buying power
of its 513-member companies to negotiate these dis-
counts. These are discounts you may not be able to
receive if you were to directly negotiate with each
company (PINE). The construction industrys portal
site in India, BuildByte (BBYT) puts it even more suc-
cinctly, Bigger the volumeLesser the price.
We saw that group-buying schemes can differ in
the order of events leading to price discovery. In
some cases, suppliers rst commit to a price-quantity
schedule leading to demand realization, and then
produce (or procure) the products in volumes exactly
calibrated to the realized demand. In other cases, sup-
pliers rst produce (or procure) products and then
sell via a group-buying market through a volume
discounting mechanism. Ceteris paribus, the former
approach leads to higher supplier prots but is not
always feasible due to long production/procurement
lead times. However, the important practical ques-
tion of whether to adopt group buying when produc-
tion postponement is or is not feasible hinges on the
relative performance of group buying vis--vis posted
pricing in these two cases. Because the answer is
by no means obvious, our model and analysis in 6
attempts to shed light on this question. More gener-
ally, the widespread use of group buying in practice
is predicated on a number of unveried, seemingly
commonsensical assumptions (reected above in
the quotes from various companies). This paper aims
to provide a more analytical response to this dis-
course, and veries the extent to which these underly-
ing beliefs are true under various market conditions,
through mathematical modeling.
Group-Buying Markets
APPA: American Public Power Association. www.
md-electric-info.com/index.html.
BBYT: BuildByte Construction Portal (Indian).
www. buildbyte.com/.
COL: Chennai Online Bazaar (India). www.
chennaionline.com/services.
ECON: Operated by PricewaterhouseCoopers.
www.pwcglobal.com.
GBP: The Group Buying Partnership. www.
groupbuying.co.uk/products.htm.
IRPG: Independent Restaurant Purchasing Group.
www.independentrestaurants.com/.
LBI: LetsBuyIt.com. www.letsbuyit.com/lbisite/
index.jsp.
MPLY: McNopoly.com. www.mcnopoly.com/.
MPSC: Maryland Public Service Commission.
www.md-electric-info.com/info-center /aggregators.
html.
OLC: Online Choice. www.onlinechoice.com/home
/default.asp.
1550 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
PINE: The Printing Industries of New England.
www.pine.org/MS/groupbuying.htm.
STBZ: The StockBuzz Market. www.stockbuz.com.
TBG: The Buying Group. www.the-buying-group.
com.
THM: HappyMany. www.happymany.com/index_
en.html.
4. The Group-Buying Mechanism:
Theoretical Underpinnings
It is useful to conceptualize the group-buying mech-
anism as composed of two interacting components:
(1) demand uncertainty faced by the seller, and (2) a
quantity discount scheme offered by the seller. The
relevant literatures on Quantity Discounts and Infor-
mation Economics are discussed below.
4.1. Quantity Discounts
There are two popular forms of quantity discounts:
(1) all units and (2) incremental (Nahmias 1997). Under
all units, the discount is applied to all the units in a
given order; under the incremental discount scheme,
the discount applies only to additional units beyond
prespecied breakpoints. All the companies operat-
ing group-buying schemes offer all-units discounts
rather than incremental quantity discounts (see the
survey in 3). This avoids penalizing early bidders
(presumably, those with higher expected utility from
the good) and encourages their early entry into the
market, promoting the bandwagon effect.
Quantity discounts have been studied as an instru-
ment to (1) facilitate transactions efciencies in trade,
in situations involving scale economies (cf. Kohli and
Heungsoo 1989, Lee and Rosenblatt 1986, Monahan
1984, Weng 1995), or (2) structure incentives in a ver-
tical market (cf. Anand et al. 2003, Cachon 2003,
Maskin and Riley 1984, Tang et al. 2001, Weng 1995).
Dolan (1987) provides an extensive review of the
early research on quantity discounts in the economics,
operations management, and marketing literatures.
(The reader wishing to delve deeper is referred to
the online appendix for an extended list of trade and
academic references.)
The study of quantity discounts as an enabler of
transactions efciencies in trade is closely related
to the concept of Economic Order Quantity (EOQ).
Monahan (1984) extends the idea of EOQ by deriving
the optimal price-quantity schedule that a supplier
should offer, given that the buyer subsequently opti-
mizes her prots by ordering appropriate quantities.
While Monahan (1984) assumes that the supplier pro-
duces in lot sizes that mimic the buyers orders, Lee
and Rosenblatt (1986) extend the analysis to the case
where the supplier can use a different lot size from
the buyer. Quantity discounts are also used by sellers
to encourage early season purchases, to reduce their
risk in environments characterized by demand uncer-
tainty (cf. Cachon 2003, Tang et al. 2001).
Weng (1995) studies the joint problem of garner-
ing transaction efciencies (i.e., minimizing operating
costs as above) and aligning incentives in the channel.
He shows that under price-sensitive demand and
scale economies, franchise fees are needed in addition
to quantity discounts to maximize prots and coordi-
nate the channel. Anand et al. (2003) show that when
the classical single-period vertical contract is extended
to a dynamic setting, two-part tariffs (a version of
quantity discounting) cannot coordinate the channel
when the buyer can carry inventories strategically.
This failure of coordination holds true even in a deter-
ministic, stationary environment and in the absence
of scale economies.
4.2. Information Economics
While transaction efciencies, and augmented buyer
bargaining power, may be part of the proffered
rationale for all Web-based demand aggregation mod-
els, the major use in practice of group-buying chan-
nels has been for price discrimination through demand
discovery.
5
Cremer and McLean (1985, 1988) ana-
lyze the case of selling to buyers with interdepen-
dent demands, with application to auctions. In their
model, when this interdependency is known to the
seller, there exists an ex post Nash equilibrium in
which the seller is able to extract as much surplus
as he would with full information, i.e., the informa-
tion rent paid by the seller to the buyers is zero.
5
For a discussion of the use of quantity discounts for second-degree
price discrimination, see Tirole (1988).
Management Science/Vol. 49, No. 11, November 2003 1551
ANAND AND ARON
Group Buying on the Web
Anand and Mendelson (1997) analyze the coordi-
nation problem faced by buyers with interdependent
costs, where the interdependency arises due to the
manufacturers scale diseconomies.
While group buying, unlike auctions, does not
lead to ex post price discrimination among different
buyers (who are all charged the same price), it does
aim to set the price dynamically based on market-
wide demand. Thus, a paper that is closely related
to the domain of our research is Maskin and Rileys
(1984) study of a monopolist facing an adverse selec-
tion problem. Maskin and Riley demonstrate that
when certain conditions (most notably, an increasing
hazard rate function for the distribution of buyers
types and linear production costs) are met, the sellers
prot-maximizing price-quantity schedule inducing
self-selection among different buyer types has the
structure of quantity discounts. This paper differs
from theirs both semantically and technically, and
ultimately in some of the results.
Semantically, group buying is a phenomenon dis-
tinct from the more traditional setting of quantity
discounts. Group buying is one form of a quantity
discount whereby the seller declares demand bands
within which price is stable while price declines
between bands (higher demand bands are character-
ized by lower prices). Where our modeling primitives
overlap with those of Maskin and Riley (1984), we
show that under a variety of settings discussed in
our analysis, group buying (which is a special case
of quantity discounts) is not optimal, and can per-
form no better than simple posted prices (trivially,
by mimicking the latter, it can do no worse). This
demonstrates that the form of discounts matters and
that group buying is not the optimal form of vol-
ume discount. Where the primitives of our model
diverge from Maskin and Rileys (1984)specically
in a case where the demand curves intersect (see 5.2
for details) or where the marginal costs of production
are nonlinearwe show that group buying dominates
posted prices depending on the nature and extent
of the heterogeneity in the demand regimes that the
supplier faces.
5. Models of Demand Uncertainty
In the survey of 3, we found that in diverse in-
dustries, the group-buying operator acted as a de
facto monopolist. In fact, the mechanism would
seem to promote a monopolistic structure among
group-buying operators. Given that the group-buying
operator aggregates potential buyers to increase their
collective bargaining power, and also in some cases, to
meet minimum quantity thresholds, the nature of the
business favors a single, large buyer group over mul-
tiple smaller buyer groups. Thus, we see consolida-
tion in different businessesfor example, StockBuzz
aggregating apparel manufacturers on behalf of yarn
sellers, IRPG aggregating 3,000 restaurants to nego-
tiate supplier discounts, or PINE wielding the bar-
gaining power of its 513-member companies in the
printing sector. In the B2C sector, the lack of com-
petition is exacerbated by the relative novelty of the
group-buying scheme to end consumers, and the spe-
cialized (niche) product lines. For these reasons, we
focus on a monopolistic seller in the analysis of this
paper; the extension of our framework to competing
group-buying operators is a topic for future research.
In the models that follow, we will analyze the
impact of different kinds of demand uncertainty on
the monopolist sellers optimal group-buying pricing
strategy, and compare its performance with that of
simple posted prices.
5.1. Model of Parallel Demand Regimes
We construct a theoretical model of a monopolist
seller facing demand uncertainty. The seller has an
estimate of the demand for his product (at every pos-
sible price) but does not know it exactly. The mono-
polist operates in one of two demand regimes (which
we term high/low), each of which is equally likely.
The monopolist cannot observe or infer the demand
regime prior to the pricing decision. The market con-
sists of a xed number of buyers transacting within a
single period. Each buyer demands exactly one unit
of an indivisible good. The monopolist seeks to sell a
quantity Q. The demand for the monopolists product
is determined by a number of factors, including prod-
uct attributes and consumer preferences, the avail-
ability of complementary or substitute products, and
a host of macroeconomic factors. He has to choose
1552 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
from one of two pricing mechanisms: group buying
or posted prices.
We preserve the general feature of all group-buying
schemesprice is decreasing in the total quantity
demanded. In this model, the price-quantity schedule
consists of two or more price points corresponding
to different quantities demanded. Further, the offered
price
6
should be decreasing in the quantity ordered,
or equivalently, we need that (q
i
, p
i
) and (q

, p

): q

>
q
i
p

p
i
. Note that implicitly, the pricing struc-
ture is akin to all units rather than incremental quan-
tity discounts. With the above formulation, we turn
to the monopolists revenue maximization problem
under group buying. We assume that demand
7
for the
good is linear, and given by
q =u
l
p and q =u
l
p
in the high and low demand regimes, respectively.
This formulation results in parallel demand curves
for the two equiprobable
8
regimes as discussed above.
The parameters of each demand regime are com-
mon knowledge to the potential buyers and the seller.
However, only the buyers can infer which regime is
perfectly operative, based on their individual pref-
erences, i.e., their positions on the demand curves
(because the curves do not intersect). We will con-
sider the case when the two regimes have nonpar-
allel (intersecting) demand curves in 5.2. The two
demand regimes are shown in Figure 1(a). The high
demand regime is the outer demand curve.
Thus, the monopolists problem is to pick a price-
quantity schedule that will maximize his expected
total revenue
9
without knowing which of the two
demand regimes is realized. Because there are only
6
We distinguish between offered and realized prices here because
a single price may be realized which is one of the many offered
prices.
7
See the online appendix of mathematical proofs and derivations
for how the generic demand curves are normalized to arrive at this
formulation.
8
Our results are qualitatively unchanged by generalizing these
probabilities to p and (1p).
9
Because production precedes pricing in our model, we are justied
in treating the marginal cost as sunk and, therefore, maximizing
revenues (as opposed to revenues net of production costs). It will be
clear from inspection that the solution and the ordering of market
Figure 1 (a) The Monopolist Attempts to Set the Higher of the Two
Price Points from the High Demand Regime. (b) The Higher
Price Point is Captured from the Low Demand Regime While
the Lower Price Point is Captured from the High Demand
Regime.
a
H
a
L
P
1
P
2
q'
2
q
1
q
2
q'
1
a
H
a
L
a
H
a
L
P
1
P
2
q'
2
q
1
q
2
a
H
a
L
Quantity
Quantity
P
r
i
c
e
P
r
i
c
e
(A)
(B)
two demand regimes, the monopolist will pick a
pair of price-quantity tuples |(q
1
, p
1
), (q
2
, p
2
)], with
one tuple corresponding to each regime such that
the expected revenue is maximized, subject to the
mechanisms by revenues will not change if a positive marginal
cost were to be factored into the analysis. When pricing precedes
production, the marginal cost might matter: in this case (studied
later in 6), our analysis assumes a nonzero marginal cost.
Management Science/Vol. 49, No. 11, November 2003 1553
ANAND AND ARON
Group Buying on the Web
condition that q
2
> q
1
and p
2
p
1
. The notation
|(q
1
, p
1
), (q
2
, p
2
)] means that the seller charges the
price p
1
when the total sales quantity is q
1
, and
p
2
when the total sales quantity is > q
1
but q
2
. In
this shorthand for the schedule under group buying,
the price p
1
is to be interpreted as the base price,
and the quantity q
1
is to be interpreted as a quan-
tity ceiling corresponding to the base price, i.e., the
quantity up to which the price p
1
will be operative.
Thus, if the total sales quantity is any number less
than or equal to q
1
, the sales price will be p
1
. If the
total quantity demanded in the market exceeds q
1
, the
lower price p
2
will obtain.
10
In the absence of capacity
(or procurement) constraints, the monopolist supplier
would simply set q
2
=+, or a large number. Further,
incentive compatibility constraints have to be satised, to
ensure that under each regime, the price-quantity pair
that is operationalized is the one that the monopolist
intended for that regime. Thus, under each regime,
the equilibrium outcome should be that the cus-
tomers, while maximizing their own value, prefer the
price-quantity pair intended for that demand regime,
rather than the alternative. The implementation of
this idea, in connection with the well-known revela-
tion principle of economics,
11
will become clearer in the
exposition below.
The seller has two alternatives: (1) set the higher
price point from the high demand regime and, there-
fore, the lower price point from the low demand
regime (this would imply that the seller expects
to realize a higher quantity of sales from the low
demand regime), or (2) set the lower price point
(and, therefore, higher sales quantity) from the high
demand regime and the higher price point (and,
therefore, lower sales quantity) from the low demand
regime. Figures 1(a) and (b) illustrate alternatives (1)
and (2), respectively, which we analyze below.
10
We use this notation for simplicity. It is worth noting that the
interpretation of a price-quantity tuple such as (q
1
, p
1
) under group
buying is different from the take-it-or-leave-it price-quantity tuples
more commonly observed in the literature (in which the buyer(s)
only have the option of buying exactly q
1
at the price p
1
). In effect,
pricing under group buying is a monotonically decreasing step
function of total quantities demanded.
11
For an excellent exposition of the revelation principle, see
Mas-Colell et al. (1995) or Myerson (1991).
Case 1. Higher price point is set from the high demand
regime (see Figure 1(a)). In Figure 1(a), q
1
=u
l
p
1
and
q
2
= u
l
p
2
. The quantity q
1
acts as a limit for the
higher price p
1
. The seller will offer the following
pricing scheme:
12
Price P =

p
1
if quantity q q
1
=u
l
p
1
,
p
2
otherwise,
subject to the constraints
13
p
1
p
2
and q
1
q
2
. Under
this scheme, the price (outcome) will always be p
2
irrespective of the demand regime that is realized.
Consider the following two scenarios: If the lower
demand regime is realized, the price that will pre-
vail is p
2
and the quantity sold will be given by
q
2
=u
l
p
2
. If the higher demand regime is realized,
then the seller would want the price to remain at p
1
,
however, buyers in this demand curve would demand
a quantity given by q
2
=u
l
p
2
, which is greater than
u
l
p
1
. As a result, the seller would have to offer
the lower price corresponding to the higher quantity
level. This leads us to our rst result about the sellers
optimal revenue.
Lemma 1. The sellers total revenue in expectation
when he chooses the higher price point from the higher
demand regime is given by
r

|(q
1
, p
1
), (q
2
, p
2
)] =
(u
l
+u
l
)
2
16
, (5.1)
and the optimal price P

is given by P

=(u
l
+u
l
),4.
Proof. The proof of Lemma 1 and all propositions
in this paper are provided in the online appendix.
The above scheme is in essence a single-price
scheme (xed price) because only a single price
will prevail irrespective of which demand regime is
realized.
Case 2. The seller picks the lower price point from
the high demand regime and the higher price point from
the low demand regime (see Figure 1(b)). In this case, the
seller will offer the following pricing scheme:
Price P =

p
1
if quantity q q
1
=u
l
p
1
,
p
2
otherwise,
12
It will be clear by inspection that this pricing scheme maximizes
the sellers revenues under Case 1.
13
These constraints result from the denition of group-buying
mechanisms: higher quantities lead to lower prices.
1554 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
subject to the constraints p
1
p
2
and q
1
q
2
.
Now, the separator between the two pricing points
is the quantity q

2
, which ensures that, in the event
the low demand regime is realized, the price does not
slide down to p
2
. In this case, separation of prices
across the two demand regimes seems possible.
14
Of course, if the seller could individually maximize
revenues in the two markets, the revenue-maximizing
prices in the high and low demand markets would
be, respectively, p

1
=u
l
,2 and p

2
=u
l
,2, with the cor-
responding sales quantities q

1
= u
l
,2 and q

2
= u
l
,2.
However, this solution is infeasible under the group-
buying mechanism, because the constraint that the
price should be decreasing in quantity demanded is
violated (because p

1
> p

2
and q

1
> q

2
). Because sepa-
rate maximization of revenues across the two demand
regimes leads to an infeasible price-quantity schedule,
we need to maximize the sellers revenue within the
feasible set of prices. Lemma 2 provides the optimal
solution.
Lemma 2. The optimal (revenue-maximizing) prices
and seller prots under group buying are given by

1
=p

2
=
u
l
+u
l
4
and r
G
|(p
1
, q
1
), (p
2
, q
2
)]
=
(u
l
+u
l
)
2
16
if u
l
u
l
(

21),
p

1
=p

2
=
u
l
2
and r
G
|(p
1
, q
1
), (p
2
, q
2
)]
=
u
2
l
8
otherwise.
In particular, any feasible group-buying schedule that sets
the prices such that p

1
= p

2
yields revenues less than the
above solution.
Remarks. Because p

1
= p

2
under the optimal
group-buying scheme (i.e., the price is independent
of the demand regime), the monopolist can clearly
do as well using posted prices, by setting his price
p

=p

1
(=p

2
). The sellers revenues under group buy-
ing and posted prices are identical.
To summarize, the price p
1
always slides down
to p
2
in Case 1, thereby establishing a single-price equi-
librium across the two demand regimes. In Case 2,
14
Recall that the sellers objective is to implement a different price
in each demand regime to optimize his total expected revenue.
the monopolist is able to implement a separation of
prices based on quantity demanded, but the revenues
from a single posted price dominate the resulting
revenues from group buying. Proposition 1 derives
from Lemmas 1 and 2.
Proposition 1. In the model of demand uncertainty
with parallel demand regimes, the sellers revenues from
group buying can never exceed his revenues from simple
posted prices. In fact, the optimal (revenue-maximizing)
group-buying scheme simply mimics the optimal posted
price. Any deviation from posted pricing (such as dif-
ferential pricing to discriminate between realized demand
regimes) leads to suboptimal revenues for the seller.
5.2. Model of Intersecting Demand Regimes
In the previous section, we considered the case
of demand uncertainty in which the two demand
regimes were parallel. This implied that at any price,
the quantity demanded under one demand regime
dominated that demanded under the other. On the
other hand, intersecting demand regimes represent
uncertainty about market size (density of buyer distri-
bution) and valuations. Most models deal with only
one of the two kinds of uncertainty. We now consider
demand uncertainty where one demand regime does
not universally dominate the other. Specically, we
study the case when the two demand regimes cross:
one regime may result in a higher quantity demanded
for one range of prices, while the other regime may
dominate over another range of prices (see Figure 2).
We derive the optimal revenues from group buying
and posted prices and compare the two.
Figure 2 illustrates the case of intersecting demand
curves. The generic linear demand curve is given by
Q = mp. Let the two curves be given by q = u

m
1
p and q = |

m
2
p, where u

> |

and m
1
m
2
. Let
m = m
1
,m
2
. We normalize m
2
to 1 so that the two
demand curves are given by q =ump and q =| p,
where u > | and m 1. Further, as seen in Figure 2,
the two curves intersect only when | > u,m (thereby
making it a binding constraint for the existence of
intersecting demand regimes).
Each demand regime can occur with equal like-
lihood, and the seller has to set prices before the
Management Science/Vol. 49, No. 11, November 2003 1555
ANAND AND ARON
Group Buying on the Web
Figure 2 Intersecting Demand Curves: Neither Demand Regime
Dominates the Other Universally
(0,b)
(0,a/m)
P
1
P
2
q'
2
q
1
q
2
q = b - p
q = a - mp
(b,0) (a,0)
Quantity
P
r
i
c
e
demand regime is realized.
15
Under group buying, the
seller picks a price-quantity schedule as before. He
offers prices of p
1
and p
2
for quantities demanded up
to the thresholds q
1
and q
2
, respectively, where q
1
q
2
and p
1
p
2
. We now determine the optimal values of
p
1
and p
2
, and the corresponding expected revenues,
under group buying.
Lemma 3. For the model of intersecting demand
regimes described above, the prot-maximizing, group-
buying, price-quantity schedule is given by
Price P

1
p

|
2
if quantity demanded
q |
u
2m
,
u
2m
otherwise.
(5.2)
The sellers expected revenue is
r(p

1
, p

2
) =
u
2
+m|
2
8m
. (5.3)
15
Because the demand curves intersect at just one point, all buy-
ers other than the one atomistic buyer at the point of intersection
can perfectly infer which regime is operative from their individual
preferences, i.e., their positions on the demand curves.
We now derive the sellers optimal price and
revenues under posted pricing, and compare these
with the optimal prices and revenues under the
group-buying mechanism.
Proposition 2. In the case of intersecting demand
regimes,
(i) The prot-maximizing price and revenues under
posted pricing are given by
p

=
u+|
2(m+1)
and r(p

) =
(u+|)
2
8(m+1)
.
(ii) The difference in revenues between the optimal
schedules for group buying and posted pricing is given by
r
g
r
p
=(u|m)
2
,(8m(m+1)) > 0.
(iii) Thus, the revenue to the seller from group buying
strictly dominates the revenue from posted pricing.
Thus, unlike the earlier case of parallel demand
curves (in which the two demand regimes had iden-
tical slopes), group buying outperforms posted prices
for intersecting demand curves, in which the two
demand regimes have different slopes. Group buy-
ing does better under greater demand heterogeneity
(reected in the parameter m), because it enables the
seller to set (nonlinear) price-quantity schedules that
optimize revenues under each demand regime, thus,
maximizing total expected revenues. With greater het-
erogeneity in demand regimes, it is easier to induce
a self-selective second-degree
16
price discrimination
among customers, via quantity-discount schemes.
However, when the seller relies on posted prices, the
single quoted price forces the seller to make a trade-
off between revenues in one demand regime and
revenues in the other, which leads to lower overall
expected revenues. In contrast, demand homogene-
ity would neutralize the advantage of pricing exibil-
ity (enabling second-degree price discrimination) that
group-buying pricing mechanisms offer. Proposition 3
establishes that the above intuition holds, with m =
m
1
,m
2
being a measure of the difference in the
slopes of the two demand curves (which induces the
heterogeneity).
Proposition 3. The gains from using the group-
buying mechanism increase as the demand heterogeneity
(m) increases.
16
See Tirole (1988).
1556 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
6. Pricing and Production: Timing
and Scale Economies
Our previous models implicitly assumed a certain
sequence of events. For example, the assumption of
zero marginal costs is reasonable when production
costs are sunk, i.e., the commitment to production
quantities precedes the pricing decision. When pric-
ing follows production, there are fewer degrees of
freedom in the pricing decision, under both group
buying and posted prices. Put differently, under a dif-
ferent sequence of events, with greater freedom in the
timing of the pricing decision, group buying might out-
perform simple posted pricing even more. A second
issue is the absence of scale economies in our previous
models. After all, because group buying is an exten-
sion of traditional quantity discounting, its greatest
benet might be precisely to maximally exploit scale
economies. This section casts the spotlight on these
two issues. In the relative performance of group buy-
ing versus posted prices, (1) what is the impact of
economies of scale, and (2) does the order of pro-
duction versus pricing matter? We analyze two dif-
ferent scenarios and the effect of scale economies
in each: (1) production precommitment, and (2) pro-
duction postponement. Under production precommit-
ment, the rms pricing decisions are made after and,
hence, are constrained by commitment to produc-
tion/procurement quantities. Production precommit-
ment is almost inevitable under long production or
supply lead times, when the rm has to commit ex
ante to purchase or production quantities. Production
postponement is feasible when production or pro-
curement lead times are sufciently small. Here, the
rm determines its production/procurement quantity
after observing the demand as a function of its price
or price-quantity schedule.
17
Under both posted prices and group buying, the
rm is better off under postponement (when this is
feasible), because the additional market information
is useful in tailoring production. Our interest, how-
ever, is in comparing the relative performance of the two
market mechanisms with and without postponement,
17
For an analytical treatment of postponement under demand
uncertainty, and its relationship to information, see Anand and
Mendelson (1998).
to study the effect of the production-pricing sequence
on the choice of market mechanism.
In all the models in this section, we assume that
there is a risk-neutral, prot-maximizing monopolist
selling to two buyers (customers) with heterogeneous
types (values). Each customer buys at most, one
unit of the good, and seeks to maximize her expected
value. Customers valuations are drawn i.i.d. from
the standard uniform distribution (i.e., uniformly and
independently from the unit interval |0, 1]). While the
distribution is common knowledge, only the corre-
sponding customer knows her own exact valuation.
Case 1. Production precommitment. In this case, all
production is ex ante, i.e., the monopolist supplier
commits to the total production quantity before the
revelation of demand through the price (or price-
quantity schedule). We model the discovery of the
optimal group-buying schedule as a game between
the monopolist and the buyers. The sequence of
events is as follows. The monopolist rst deter-
mines his production quantity and then picks a price-
quantity schedule, and buyers choose to buy or
abstain. The monopolist folds the expected result-
ing demand from the stage game to arrive at the
optimal price-quantity schedule. Recall that, under
group buying, each customers bid creates positive
externalities for all other customers by lowering the
expected price.
18
Hence, each customer needs to fac-
tor in the probability of bidding by other customers
(which is a function of their values) in determining
her own bid. A customer will buy (bid) if her value
is greater than the expected price. Thus, a customer
may bid even when the current price under group
buying is greater than her value for the good. This
may induce other customers to bid, further lowering
prices. A direct approach to comparing the relative
performance of the two mechanisms (posted prices
and group buying) is to derive the optimal prices
and prots under eachto solve the group-buying
problem, we may employ the revelation principle (cf.
Myerson 1991). However, in this case, we may build
on Myerson (1981) to compare the two mechanisms
without deriving the optimal group-buying schedule.
18
Kauffman and Wang (2001) nd evidence of the positive
externality effect on customer bids in their empirical study.
Management Science/Vol. 49, No. 11, November 2003 1557
ANAND AND ARON
Group Buying on the Web
Myerson (1981, p. 60) focuses on the auction of a sin-
gle, indivisible object. His result on the optimal auc-
tion (extended for the multiunit case) is stated below
without proof:
Suppose the seller has | indivisible units, and there
are n buyers whose valuations are independent
uniform draws from the unit interval |0, 1]. The
optimal auction is to set a reservation price equal to
p

=max
p|0, 1]
p (1p) =1,2, and allocate the | units
to the | highest bidders at a price equal to the larger
of the reservation price and the highest losing bid.
The preceding implies that when | =n, the realized
auction price is exactly the reservation price, and the
optimal auction is equivalent to a posted-price mechanism.
The critical observation is that Myersons (1981) math-
ematical specication of the auction is quite general
in terms of probabilities of different bidders winning
the good, and expected money transfers between the
seller and buyers. Both group buying and posted
prices t within these specications. Proposition 4
builds on the preceding observations to compare the
two mechanisms under production precommitment.
Proposition 4. Under production precommitment, the
equilibrium solution and the suppliers prots are identical
under group buying and posted pricing.
Proposition 4 applies under production precom-
mitment for both linear production costs and scale
economies, because the production costs are sunk
before pricing and demand realization. Thus, with
production precommitment, the best that the sup-
plier can do under group buying is to mimic the
optimal posted pricing schemeoffering quantity dis-
counts does not improve the suppliers prots. This
result on the inefcacy of quantity discounts (as medi-
ated by the group-buying mechanism) even under
scale economies is contrary to previous research that
argued that quantity discounts contributed to transac-
tion efciencies, particularly under production scale
economies (cf. Kohli and Heungsoo 1989, Lee and
Rosenblatt 1986, Monahan 1984, Weng 1995). The
crucial difference is that the greater coordination costs
under group buying offset the improved transaction
efciency from quantity discounting.
Case 2. Production Postponement. This model differs
from that of Case 1 in the sequence of produc-
tion and pricing. The altered sequence of events is
as follows: the supplier rst quotes a price-quantity
schedule, and buyers then respond with their pur-
chase decisions (demand is realized). Finally, the sup-
plier tailors his production to exactly satisfy his com-
mitments (orders). In this case too, the monopolist
determines the optimal price-quantity schedule (one
that maximizes revenues) by backward induction. We
assume that the cost of producing i units is c
i
, for i =
1, 2; without loss of generality, we set c
0
=0, i.e., there
are no production-independent xed costs. Our anal-
ysis focuses on the effect of economies of scale in
production, thus, we assume that marginal costs are
weakly decreasing, i.e., c
2
c
1
c
1
. The limiting case
of c
2
c
1
=c
1
corresponds to linear production costs.
In this case, we need to explicitly derive the opti-
mal group-buying and posted-pricing solutions to compare
their performance; indirect approaches such as invok-
ing the optimal auction solution of Myerson (1981)
are not helpful. Under group buying, we denote the
price vector by |p
1
, p
2
], where p
1
is the unit price for
one unit of sales, and p
2
the unit price otherwisethe
quantity discount per unit is thus (p
1
p
2
). We solve
the group-buying problem using the revelation prin-
ciple (cf. Myerson 1991). Proposition 5 derives the
unique, subgame-perfect equilibrium for this game,
under both posted prices and group buying.
Proposition 5. Under production postponement, the
unique subgame-perfect equilibrium and supplier prots for
posted pricing and group buying are as given below.
(i) Posted pricing:
p

=
1c
1
+c
2
22c
1
+c
2
and r

p
=
(1c
1
)
2
22c
1
+c
2
.
(ii) Group buying:
p

1
=
1+c
1
2
,
p

2
=
1
6

4+c
2

6(1c
1
+c
2
)
22c
1
+c
2

1+3(2c
1
)c
1
+(c
2
4)c
2

,
and
r
G8
=
1
108
(5272c
1
+36c
2
1
15c
2
18c
1
c
2
+9c
2
1
c
2
+12c
2
2
2c
3
2
+2(1+3(2c
1
)c
1
+(c
2
4)c
2
)
3,2

.
We nd that in this (nal) case, the prots for the
monopolist from group buying dominate those under
posted pricing. As Figure 3 illustrates, group buying
1558 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
Figure 3 Prot Comparisons when Pricing Precedes Production Under
Scale Economies: Group Buying Does Better than Posted
Pricing for the Seller
0.4
0.6
0.8 0.4
0.5
0.6
0.7
0.8
0.5
0.7
C
1
C
2
0.025
0.05
0.075
0.1
does provide an exploitable informational advantage over
simple posted pricing.
The difference in seller prots under group buy-
ing and posted pricing when production costs change
is a function of two factors: (1) absolute protability
when costs (c
1
or c
2
or both) increase, prots fall
under both group buying and posted pricing. A drop
in absolute protability tends to drive the prot difference
down as well; (2) production scale economies. Observe
that under linear production costs,
19
the prot expres-
sions of Proposition 5 reduce to r

p
=r

G8
=(1c
1
)
2
,2.
In fact, the two market mechanisms yield identical prof-
its to the seller in this limiting case alone. Further, the
degree of scale economies (measured by the concavity
of production costs) is increasing in c
1
and decreasing
20
in c
2
. Thus, ceteris paribus, as c
1
increases or c
2
decreases, the relative production cost of the second
unit vis--vis the rst decreases. From the preceding
discussion, it becomes clear that when c
2
increases,
both of the above drivers, absolute protability and
scale economies, work in the same direction and
drive down the prot difference as seen in Figure 3.
A more interesting case is when c
1
increases: also
from the preceding discussion, the absolute protabil-
19
Linear production costs imply (c
2
c
1
) =c
1
0, i.e., c
2
=2 c
1
.
20
Because marginal production costs are c
1
and c
2
c
1
for the rst
and second units, respectively.
ity driver will cause the prot difference r
g
r
p
to
fall, while the scale economies driver will cause the
prot difference r
g
r
p
to increase in c
1
. Overall,
the direction of change of the prot difference when
c
1
increases, depends on which of the two drivers
dominates. Figure 3 shows that, in fact, the differ-
ence in monopoly prots between group buying and
posted pricing is an increasing function of c
1
for the
entire feasible range of c
1
(and c
2
). Thus, surprisingly,
the impact of the scale economies driver dominates that of
the absolute protability driver for the entire feasible range
of c
1
, illustrating the signicance of production scale
economies under production postponement.
Figures 4 and 5 shed further light on the behav-
ior of prices and prots under scale economies under
the two mechanisms. The cost parameter for the rst
unit, c
1
, is set to one of three values: Low (c
1
=0.1),
Moderate (c
1
= 0.25), and High (c
1
= 0.5). For each
xed c
1
, the cumulative production cost for two
units, c
2
, is varied along its entire feasible range.
Figure 4 plots the optimal prices under group buy-
ing and posted prices. It turns out that P
1
P

P
2
in all cases, with the equality holding exactly when
c
2
=2 c
1
. In fact, P
1
is independent of c
2
, but an
increasing function of c
1
. Intuitively, the seller cares
about the value of P
1
only when exactly one unit is
sold, because this is the only scenario under which P
1
affects revenues. But then, the only cost that matters
is c
1
. Both P

and P
2
are increasing in c
2
, and even-
tually converge to P

1
when c
2
= 2 c
1
. Under posted
pricing, the seller has to balance out the benet of sales
at a higher price (in the event of high customer valua-
tions) versus the risk of losing sales at the higher price (if
low customer valuations).
This trade-off is reected in the optimal posted
price P

. Group buying affords the monopolist greater

exibility. Because the probability of one customer
having a high valuation and the other having a low
valuation is high, the seller sets P

1
to be high to trap
the high value customers demand when one cus-
tomer has a high valuation and the other has a low
valuation. This risk is offset by the safety net of a
lower P

2
, to make the sales when both customers have
moderate valuations. Of course, group buying has its
drawbackswhen both customers have high valua-
tions, the price settles down to P

2
, leaving money
on the table; and when both customers have low
Management Science/Vol. 49, No. 11, November 2003 1559
ANAND AND ARON
Group Buying on the Web
Figure 4 Price and Prot Comparisons Under Varying Costs
(a) Low Costs (C
1
= 0.1)
0.12 0.16 0.2
0.53
0.55
0.51
C
2
P
r
i
c
e
s
P
1
P
2
0.5 0.6 0.7 0.8
0.55
0.63
0.7
0.50
C2
P
r
i
c
e
s
P
1
P
2
(c) High Costs (C
1
= 0.5)
0.3 0.4 0.5
0.52
0.57
0.62
C2
P
r
i
c
e
s
P
1
P
2
(b) Moderate Costs (C
1
= 0.25)
P* P*
P*
Notes. c
1
varies from 0.1 (low) to 0.5 (high), while c
2
varies between c
1
and 2c
1
. This is the range of values for c
2
that is feasible under production scale
economies. P
1
and P
2
are the prices under the optimal group-buying schedule, while P

is the optimal posted price.

valuations, no sales are made. But these events have
a relatively low probability. Furthermore, the cost sav-
ings from production postponement mean that these losses
are less than they would have been if production com-
mitments had to be made before orders are revealed. Fig-
ure 5 shows that for any xed level of c
1
, the prot
difference is greatest for c
2
c
1
(maximum scale
economies); it falls as c
2
increases, and eventually
becomes 0 when c
2
=2 c
1
.
To summarize, when production can be tailored
to meet revealed demand, scale economies allow the
exploitation of nonlinear pricing under group buying,
which outperforms posted pricing. Clearly, revelation
of demand information (through a nonlinear price-
quantity schedule) and the optimal exploitation of that
information (via sequencing production after pricing)
together play a role in driving our results. In other
words, neither postponement nor scale economies alone
justies the use of group buying. However, when both
coexist, group buying outperforms simple prices for
the seller.
7. Conclusion and Extensions
Because group buying does not permit ex post
price discrimination among customers with differ-
ent valuations, it cannot outperform posted prices in
the absence of demand uncertainty.
21
Thus, demand
uncertainty is a sine qua non for the viability of the
group-buying mechanism. Our analysis showed that
group buying outperformed posted pricing under
(1) demand heterogeneity (5), and (2) production
postponement in combination with scale economies
(6). We briey summarize our ndings in each con-
text below.
The analysis in 5 showed that under heteroge-
neous demand regimes, the seller is able to use group
buying to price discriminate and capture some of the
revenues lost by setting a single price across both
demand regimes. As the demand regimes become
more similar, the situation begins to resemble the
cases analyzed in Proposition 1: the advantage of
group buying over posted pricing shrinks and nally
vanishes.
22
21
Observe that if the nal price under group buying can be per-
fectly predicted, setting a list price equal to this realized price will
do equally well for the seller.
22
For the advantage of group buying to vanish, the demand
regimes do not need to be identical. It is enough if they are alike,
as for example, if their slopes are the same, even though they may
have different offsets.
1560 Management Science/Vol. 49, No. 11, November 2003
ANAND AND ARON
Group Buying on the Web
Figure 5 Prot Comparisons Under Varying Costs
(a) Low Costs (C
1
= 0.1)
0.12 0.14 0.16 0.18 0.2
0.405
0.415
0.425
c
2
0.25 0.35 0.4 0.45 0.5
0.29
0.31
0.32
0.30
0.30
c
2
P
r
o
f
i
t
s
(c) High Costs (C = 0.5)
1
0.5 0.6 0.7 0.8
0.18
0.19
0.21
0.22
0.23
0.24
0.20
Posted Pricing
Group-Buying
P
r
o
f
i
t
s
P
r
o
f
i
t
s
(b) Moderate Costs (C
1
= 0.25)
c
2
Notes. c
1
varies from 0.1 (low) to 0.5 (high). Low c
1
=01, moderate c
1
=025, and high c
1
=05, for feasible values of c
2
. This is the range of values
for c
2
that is feasible under production scale economies.
We also found that the value of group buying
depends on the nature of the uncertainty about buyer
valuations in the market. When the distribution that
characterizes buyer valuations is known beforehand,
sellers are almost always better off by running a
posted-price market, in fact, the only exception was
the combination of production postponement and scale
economies (6). Neither production postponement nor
scale economies was individually enough for group
buying to outperform list prices.
Other disadvantages of the group-buying mecha-
nism vis--vis posted pricing were not modeled in
this paper. Specically, group buying involves delays
and uncertainty before the nal price is realized, for
both buyers and sellers. The delay before consum-
mation of trade leads to utility decay for all par-
ties, and uncertainty coupled with risk aversion (on
the part of either party to the trade) leads to further
devaluation of the eventual trade. Because buyers
anticipate the delay, their willingness to pay during
the bidding phase is lowered, leading to a down-
ward shift in the demand curve under the group-buy
mechanism. This may well result in some adverse
selection wherein the buyers who value the product
least (and therefore, suffer the least utility decay) may
be attracted to the group-buying market. These issues
should be addressed in future research.
Another important area for future research, not
studied in the models of this paper, is the im-
pact of competition. Under competition (or under
oligopoly), group-buying markets may be run for
three reasons: (1) the market can be segmented
between price-sensitive buyers and buyers whose
Management Science/Vol. 49, No. 11, November 2003 1561
ANAND AND ARON
Group Buying on the Web
demand for the product is relatively inelastic,
23
which
allows price-sensitive buyers to choose the group-
buying option, (2) a consortium of buyers may
emerge to negotiate discounts on behalf of buyers
for products that are characterized by production or
logistical scale economies (such as ofce supplies,
pulp and paper, and utilities), and (3) a consortium
of suppliers may aggregate small to marginal retail-
ers for whom the costs of running individual discount
programs may be substantial.
Bounded rationality may limit the effectiveness of
these markets. Group buying is a relatively complex
mechanism demanding sophisticated calculations by
both sellers and buyers. Individual consumers, in par-
ticular, may be unable (due to bounded rationality)
to make complex calculations, or at least experience a
disutility from this process.
Another possible extension of this research is to
group-buying mechanisms under uncertainty about
product quality. When buyers can get signals about
product quality from the bids of other buyers (simi-
lar to common values auctions), group-buying mech-
anisms can be a powerful way for the seller to induce
buyers to signal product quality information to each
other. Future research could extend our model to
study the impact of these different factorsdelays,
uncertainty, risk aversion, bounded rationality, and
adverse selection. The impact of competition and mul-
tiple markets concurrently functioning could be mod-
eled through a dynamic simulation.
Acknowledgments
The authors thank Gerard Cachon, Art Geoffrion, Ramayya
Krishnan, and Howard Kunreuther for helpful comments and
suggestions. We also greatly beneted from the feedback provided
by the associate editor and three anonymous referees.
An electronic companion to this paper is available at man-
sci.pubs.informs.org/ ecompanion.html.
23
This holds especially when there is a strong inverse correlation
between delay sensitivity and price sensitivity and, hence, a direct
correlation between delay sensitivity and willingness to pay. In fact,
some group-buying sites offer products (e.g., electronic items) that
are less than state of the art at steep discounts, while the latest
product is sold via Web-based posted prices or traditional brick-
and-mortar channels. This quality-based, second-degree price dis-
crimination reduces the cannibalization of the sales of high-margin,
high-quality items by low-margin items.
References
Anand, K. S., H. Mendelson. 1997. Information and organization for
horizontal multimarket coordination. Management Sci. 43(12)
16091627.
, . 1998. Postponement and information in a supply chain.
Working paper, The Kellogg School, Northwestern University,
Evanston, IL.
, R. Anupindi, Y. Bassok. 2003. Strategic inventories in ver-
tical contracts. Working paper 02-05-06, The Wharton School,
University of Pennsylvania, Philadelphia, PA.
Cachon, G. P. 2003. The allocation of inventory risk and advanced
purchase discounts in a supply chain. Working paper, The
Wharton School, University of Pennsylvania, Philadelphia, PA.
Cremer, J., R. P. McLean. 1985. Optimal selling strategies under
uncertainty for a discriminating monopolist when demands are
interdependent. Econometrica 53(2) 345362.
, . 1988. Full extraction of the surplus in Bayesian and
dominant strategy auctions. Econometrica 56(6) 12471257.
Dolan, R. J. 1987. Quantity discounts: Managerial issues and
research opportunities. Marketing Sci. 6(1) 122.
Kauffman, R. J., B. Wang. 2001. New buyers arrival under dynamic
pricing market microstructure: The case of group buying dis-
counts on the Internet. J. Management Inform. Systems 18(2)
157188.
, . 2002. Bid together, buy together: On the efcacy
of group buying business models in Internet-based selling.
P. B. Lowry, J. O. Cherrington, R. R. Watson, eds. The E-Business
Handbook. CRC Press, Boca Raton, FL, 99137.
Kohli, R., P. Heungsoo. 1989. A cooperative game theory model of
quantity discounts. Management Sci. 35(6) 693707.
Lee, H. L., M. Rosenblatt. 1986. A generalized quantity discount
pricing model to increase suppliers prots. Management Sci.
32(9) 11771185.
Maskin, E., J. Riley. 1984. Monopoly with incomplete information.
RAND J. Econom. 15(2) 171196.
Mas-Colell, A., M. D. Whinston, J. R. Green. 1995. Microeconomic
Theory. Oxford University Press.
Monahan, J. P. 1984. A quantity discount pricing model to increase
vendor prots. Management Sci. 30(6) 720726.
Myerson, R. B. 1981. Optimal auction design. Math. Oper. Res. 6(1)
5873.
. 1991. Game Theory: Analysis of Conict. Harvard University
Press, Cambridge, MA.
Nahmias, S. 1997. Production and Operations Analysis, 3rd ed.
McGraw-Hill, Irwin.
Tang, C. S., K. Rajaram, A. Alptekinoglu, J. Ou. 2001. The benets
of advance booking discount programs: Model and analysis.
Management Sci. Forthcoming.
Tirole, J. 1988. The Theory of Industrial Organization. MIT Press,
Cambridge, MA.
Weng, Z. K. 1995. Channel coordination and quantity discounts.
Management Sci. 41(9) 15091522.
Accepted by Arthur Geoffrion and Ramayya Krishnan; received January 2002. This paper was with the authors 12 months for 3 revisions.
1562 Management Science/Vol. 49, No. 11, November 2003