Merchant or Two-Sided Platform?

Andrei Hagiu

December 6, 2006

Abstract

This paper provides a first pass at clarifying the economic tradeos between two

polar strategies for market intermediation: the "merchant" mode, in which the inter-

mediary buys from sellers and resells to buyers; and the "two-sided platform" mode,

under which the intermediary enables aliated sellers to sell directly to aliated buy-

ers. The merchant mode yields higher profits than the two-sided platform mode when

the chicken-and-egg problem due to indirect network eects for the two-sided platform

mode is more severe and when the degree of complementarity/substitutability among

sellers products is higher. Conversely, the platform mode is preferred when seller in-

vestment incentives are important or when there is asymmetric information regarding

seller product quality. We discuss these tradeos in the context of several prominent

digital intermediaries.

1. Introduction

With ever more sophisticated logistics and the rise of information technologies, intermedi-

aries and market platforms have become increasingly ubiquitous and important agents in

Harvard Business School; ahagiu@hbs.edu

the digital economy. Sites such as Alibaba, Amazon, eBay, Google Video, i-mode, iTunes,

etc. help connect millions of sellers to tens of millions of buyers. At a broad level, one can

distinguish between two polar types of intermediaries: "merchants", who buy goods from

sellers and resell them to buyers, and "two-sided platforms", who allow "aliated" sellers

to sell directly to "aliated" buyers.

This paper is a first pass at clarifying the dierences between these two forms of market

organization by intermediaries and the economic tradeos involved. Indeed, despite the

recent surge of interest in the economics literature in two-sided markets (Armstrong (2006),

Caillaud and Jullien (2003), Evans (2003), Hagiu (2006a) and (2006b), Rochet and Tirole

(2003) and (2006), Schmalensee (2002)), all of the work in this area up to now takes

the form of market intermediation as given - namely, the two-sided platform mode - and

focuses on pricing and eciency issues. This has left a gap between this new literature

and the older economics literature on market intermediaries (Biglaiser (1993), Rubinstein

and Wollinsky (1987), Spulber (1996), Stahl (1988)). Our paper is the first to attempt to

bridge this gap.

The main dierence between the classic form of market intermediaries - which we will

call merchants from now on - and two-sided platforms is that pure merchants, by taking

possession of sellers goods, take full control over their sale to consumers. By contrast, pure

two-sided platforms leave that control entirely to sellers and simply determine buyer and

seller access to (or aliation with) a common marketplace. The following figure depicts

this dierence.

2

Consumers

Pure platform

Sellers

affiliation

sale

affiliation

Consumers

Pure merchant

Sellers

sale

sale

To give a few examples, retailers like Walmart and Amazon are (mostly) merchants.

At the opposite end of the spectrum, eBay is a pure two-sided platform. More interesting

are intermediaries, such as Apples iTunes digital music store, which exhibit both platform

and merchant features and therefore lie in-between these two extremes. Although Apple

does not take physical or full legal "possession" of the songs it distributes (the rights

remain with music publishers), it does obtain the right to repackage and price them as

it sees fit on iTunes. The 99 cents per song policy is entirely Apples unilateral decision

(and is increasingly contested by music studios), designed to provide simplicty of usage

to consumers and promote sales of Apples associated digital music player, the iPod1.

Thus, even though the combination iPod/iTunes exhibits indirect network eects (music

publishers obtain higher profits by signing a distribution deal with Apples iTunes when

more consumers buy iPods), the extent of control over pricing and distribution that Apple

maintains make iTunes more similar to a merchant such as Walmart, rather than a pure

two-sided platform, such as eBay.

Using a simple framework, we formalize the economic tradeos between the (one-sided)

1 Indeed, Apples profits from the iPod/iTunes combination come largely from sales of iPod, where thecompany enjoys margins higher than 20 percent. On the music side, it is estimated that Apple makes lessthan 10 cents for every song.

3

merchant "mode" and the two-sided platform "mode". First, we show that, unlike the

merchant mode, the platform mode exhibits indirect network externalities between sellers

and consumers (buyers). As a consequence, sellers might be unwilling to aliate with

the platform because they anticipate other sellers will not do so, leading to low consumer

demand for the platform, which ex-post justifies a non-aliation decision. In such cases,

the merchant model helps "break" these unfavorable seller expectations by eliminating

indirect network eects and achieves higher total profits. On the other hand however, the

merchant model usually entails higher costs per seller - inventory, risk, management, etc.

-, so that a platform model is more desirable when one wants to achieve higher product

variety (and seller coordination problems are not too severe).

Second, the merchant mode also dominates the platform mode whenever there are

significant complemetarities and/or substitutability between sellers products, that sellers

pricing independently (on a platform) do not internalize. By taking control over pricing

(as well as advertising, distribution, bundling, etc.) decisions, the merchant can create

more value and extract more profits from consumers. In addition we also show that a

platform faces an inherent hold-up problem when contracting with sellers before selling to

consumers, since it does not take into account seller profits when it sets its access price

to consumers and therefore tends to excessively limit consumer adoption. By definition, a

merchant fully internalizes seller profits2.

Third, whenever sellers can make quality enhancements to their products after con-

tracting with the intermediary and before selling to consumers, the platform mode is

more desirable since it preserves seller investment incentives by making them the resid-

2This is true in the simple framework we present in this paper because we assume away the existence ofseller "contact" with consumers outside of the intermediary. If such contact existed (for example, throughalternative sales channels or simply brand name), then even the merchant mode would leave certain exter-nalities on seller profits uninternalized.

4

ual claimants of those investments. Similarly, when quality is uncertain, the intermediary

is better o devolving the corresponding risk at least partially back to the sellers, by using

a platform mode.

The remainder of the paper is organized as follows. We present our basic model in

section 2, then we proceed to use variations of this model in order to formalize the respec-

tive eects of indirect network externalities, product complementarities/substitutability

and investment incentives on the tradeo between merchants and two-sided platforms in

sections 3,4 and 5. Section 6 uses the insights drawn from the formal analysis to discuss

several real world examples. Section 7 concludes.

2. Basic modelling framework

There is one intermediary which makes it possible for n identical sellers to deliver their

products to consumers (buyers). The intermediary can choose between two ways of func-

tioning: a merchant (one-sided) mode and a two-sided platform mode. Under the merchant

mode, the intermediary buys sellers products by oering a buyout bid BS for each seller

and resells the goods to consumers for an individual price pM (n) that it chooses. In the

two-sided platform mode, the intermediary charges each seller an "access" or "aliation"

fee PS , in exchange for which sellers can sell their goods directly to the consumers ali-

ated with the intermediary for an individual price pP (n). The price pP (n) is determined

by competition between n sellers for the consumers aliated with the two-sided platform.

We assume each seller is small enough (i.e. n is a continuum), so that it does not take

into account the eect of its price on overall user demand for the platform. We also im-

plicitly assume unaliated sellers cannot sell their products to consumers (aliated or

unaliated).

5

Regardless of the mode - platform or merchant - chosen by the intermediary, consumers

have to obtain access to (aliation with) the intermediary in order to be able to purchase

the sellers products: we denote by PC the consumer access fee charged by the intermediary.

Consumers buy either from the intermediary itself if the latter chooses to function in a

merchant mode, or directly from the sellers aliated with the intermediary, if the latter

chooses the two-sided platform mode.

The number of consumers which choose to aliate with the intermediary when the

price of each good is p and the intermediary charges PC for consumer aliation, is given

by:

NC = FV (n) np PC

where V (.) and F (.) are increasing in their respective arguments. Thus, we assume for

simplicity consumers are only horizontally dierentiated.

Each seller incurs a fixed cost f when making its product available for distribution

through the intermediary3. Thus, when selling through a merchant oering bid BS , each

seller makes total profits:

M = BS f

When selling through a platform, those profits are:

P (n) = pP (n)NC PS f = pP (n)FV (n) npP (n) PC

PS f

Thus, the key dierence in this very basic framework is that under the pure platform

mode, sellers care about the number of consumers patronizing the intermediary, whereas

3f is the same for all sellers. The substance of our analysis here is unchanged by allowing f to bedistributed according to a non-degenerate distribution H (.).

6

under the pure merchant mode, they only care about the buyout bid.

The timing of the general pricing/adoption game we consider throughout the paper is:

1) The intermediary announces seller access fee PS or buyout oer BS (dependingon the chosen mode) and, if it has the ability to credibly commit, it also announces

PC , the access price it will charge consumers in the third stage

2) Sellers decide whether or not to accept the intermediarys oer and those who doincur the cost f to make their products available through that intermediary.

3) If it has not announced PC in stage 1, the intermediary announces PC and con-sumers decide whether or not to aliate with the intermediary

4) Under the merchant mode, the intermediary chooses price pM (n) and under theplatform mode, aliated sellers choose price pP (n); in both cases, aliated con-

sumers decide whether or not to buy seller products.

Note that we implicitly assume that all bargaining power lies with the intermediary

when it makes pricing or buyout oers to sellers i.e. it will always charge a price that

makes sellers just indierent between accepting or not4. Finally, although we take n as

exogenously given, all the analysis below goes through when n is endogenously chosen by

the intermediary.

3. Indirect network externalities

In this subsection we assume for simplicity V (n) = v0 + n v and pP (n) = p, withv > p > 0. In other words, sellers products are identical but independent of one another.

4 Introducing variable degrees of bargaining power would simply make the dierence between merchantsand two-sided platforms less stark in this framework without changing the main insights.

7

We also abstract from a hold-up type problem that can arise through PC (cf. section

4 below) by assuming the intermediary commits to PC in stage 1, when it contracts with

sellers5.

a) Two-sided platform mode

In this case, sellers profits from aliating with the platform are:

(n) = pFv0 + n (v p) PC

PS f

Clearly, there are indirect network externalities among sellers, which means that mul-

tiple equilibria exist.

If expectations are "favorable", then every seller expects all other sellers to aliate

with (adopt) the platform whenever:

pFv0 + n (v p) PC

PS f 0

and in that case he will adopt the platform whenever this condition holds as well. It is

easily seen that this is an equilibrium strategy for sellers given PS and PC , therefore the

platform can charge:

PS = pFv0 + n (v p) PC

f

This leads to total profits:

PCNC + nPS =PC + np

Fv0 + nv np PC

nf

5Hagiu (2006a) shows that commitment might not always be optimal in the two-sided platform modewith unfavorable expectations, but we abstract away from that issue here.

8

which the platform maximizes over PC to obtain:

PF = maxhP

n ePF v0 + nv ePo nfIf expectations are "unfavorable", then every seller expects no other seller will aliate

with the platform unless:

pFv0 PC

PS f 0

It is easily seen that adopting if and only if this condition holds is also an equilibrium

strategy for sellers, sustained by unfavorable beliefs. In this case, the platform can only

charge:

PS = pFv0 PC

f

and attracts all sellers. Total platform profits are then:

PNF = maxPC

PCF

v0 + vn np PC

+ npF

v0 PC

nf

and clearly:

PNF < PF

b) Merchant mode

The intermediary behaves as a merchant in the sense that instead of trying to "attract"

sellers, it simply buys their products and resells them to consumers. Given our assumption

that the intermediary has all the bargaining power in both modes, a seller will accept the

bid if and only if:

BS f

9

Therefore, in this case, there are no indirect externalities to speak of. Each seller only

cares about the bid she is being oered, not about the number of consumers that the

merchant will be able to attract. Nor does any seller care about what the other sellers do.

Merchant profits are then simply:

M = maxPC ,p

PC + np

Fv0 + nv np PC

nf

We have thus proven the following proposition:

Proposition 1 The merchant mode is strictly preferred to the two-sided platform

mode whenever there is a positive probability that seller expectations are unfavorable:

M= PF> PNF

The insight contained in this result is quite straightforward: it is easier (cheaper) to

convince sellers to sell their products outright than to aliate to a platform and sell the

products to consumers themselves, because the first option eliminates coordination issues.

On the other hand however, a merchant generally incurs higher costs per seller, cor-

responding for instance to inventory and risk costs undertaken when taking possession of

sellers products, and to more complex management costs. In that case, the expression of

total merchant profits may become:

M = maxPC ,p

PC + np

Fv0 + nv np PC

nf C (n)

where C (n) is the cost of taking possession of n goods and is increasing (possibly convex)

in n. Therefore, the tradeo between the merchant mode and the two-sided platform

10

mode in this simple setting is between higher costs of managing more products and higher

costs of "convincing" sellers to aliate. This suggests that intermediaries, especially for

new goods, will generally start under a merchant format and, as sucient sellers become

aliated, move towards a more "open", platform mode, which is cheaper per seller and

thus allows intermediaries to oer a broader variety of products.

4. Pricing distortions, product complementarity/substitutability

Another factor which typically makes the merchant mode more desirable for the interme-

diary is the existence of pricing distortions.

The first type of pricing distortion is the one introduced by independent seller pric-

ing in a two-sided platform mode. Indeed, whenever sellers products are complementary

(substitutable), by pricing independently, sellers fail to internalize complementarity (sub-

stitutability) eects, leading to too high (too low) prices (i.e. pP (n) pM (n)).To see this, assume PC is exogenously fixed to P in both the merchant and the platform

mode - this assumption helps us abstract from the hold-up issue analyzed below. In this

case, under a two-sided platform mode the intermediary sets:

PS = pP (n)FV (n) npP (n) P

f

leading to:

P =P + npP (n)

FV (n) npP (n) P

nf

whereas under the merchant mode:

M = maxp

P + np

FV (n) np P

nf

11

Clearly, M > P because the merchant has the flexibility to internalize the comple-

mentarity (subtitutability) between seller products.

Note that if P were not exogenously given and could be chosen freely by the intermedi-

ary, then the two modes lead to the same level of profits. The additional degree of freedom

that the merchant possesses in choosing p is not neceesary for reaching the first best level

of profits in this simple model. Of course, in reality, that additional degree of freedom can

make a big dierence, as soon as the interval in which PC can vary is limited (for example,

Internet digital music stores usually cannot charge access fees to users) or consumers are

vertically dierentiated so that using a razor-and-blades pricing strategy is optimal. This

is especially true when there are other variables that the merchant can choose (advertising

levels, bundling strategies, store layout and design, etc.), which can create value above and

beyond what the sellers can create by acting independently. Again, the tradeo is that the

merchant mode may incur higher operational costs than the platform mode.

The second type of price distortion is the one which arises when the platform cannot

credibly commit to PC in the first stage, upon contracting with sellers. In this case, as

shown in Hagiu (2006b), a hold-up problem arises between the platform and sellers. Indeed,

in the second stage the platform sets:

PC2 = argmaxPC

PCFnV (n) npP (n) PC

which fails to take into account seller profits npP (n)F

nV (n) npP (n) PC

. As a

result, PC will be set too high relative to the price which maximizes joint profits, i.e.:

PC2 > PC = argmax

PC

PC + npP (n)

FnV (n) npP (n) PC

12

This is ex-ante reflected in PS: the platform has to lower the aliation fee for sellers in

order to compensate them for the platforms subsequent failure to internalize their profits:

PS = pP (n)FnV (n) npP (n) PC2

< pP (n)F

nV (n) npP (n) PC

By contrast, the merchant mode allows the intermediary to fully internalize seller profits

(it buys them out!) in the second stage:

M2 = maxPC ,p

PC + np

FnV (n) np PC

yielding total merchant profits:

M = maxhP

n ePF v0 + nv ePo = PC + npP (n)F nV (n) npP (n) PCFor example, with linear consumer demand F (u) = u, we obtain:

P =1

4

h(V (n))2

npP (n)

2i nf < M = 14(V (n))2 nf

Thus, the merchant mode does strictly better than the platform mode, even in the

absence of seller coordination problems.

One could argue that an easy way for the platform to get around this hold-up problem

is to charge sellers variable fees (or royalties), which is often the case with real-world inter-

mediaries. Indeed, assume now that under the two-sided platform mode the intermediary

can charge both the fixed fee PS and a royalty , proportional to the price pP (n) charged

13

by sellers to consumers, 0 1. In the second stage the platform sets:

PC () = argmaxPC

PC + npP (n)

FV (n) npP (n) PC

And in the first stage, the plaform can therefore charge:

PS = (1 ) pP (n)FV (n) npP (n) PC ()

f

leading to total platform profits:

P = max

PC () + npP (n)

FV (n) npP (n) PC ()

nf

It is then easily seen that setting = 1 achieves the first-best level of profits, so that:

P ( = 1) = M

But = 1 simply means that the platform is the residual claimant of all second-stage

seller revenues (from selling to consumers), which is equivalent for all practical purposes

to a merchant mode, i.e. the intermediary taking possession of sellers products).

It goes without saying that in reality it is rarely feasible to charge = 1, i.e. to extract

all revenues from sellers. There are several important reasons for this: the need to preserve

seller investment incentives (when the latter need to invest in enhancing the quality of their

products after having contracted with the platform) and asymmetric information about

product quality, which requires the sellers to bear at least some of the risk associated with

their products ( = 1 transfers all the risk to the platform-intermediary).

In the following subsection, we show how the tradeo between the need to resolve the

14

hold-up problem and maintaining sucient seller investment incentives plays out and that

in such a context the platform mode can be preferred to the merchant mode.

5. Hold-up vs. preserving innovation incentives

Assume now that the quality of seller products is variable. The variant of the model that

we use here draws upon Hagiu (2006b). Under the platform mode, each seller can produce

a product of quality q by investing cq2

2 . If the intermediary opts for the merchant mode

and takes possession of seller products, it can invest itself in improving product quality, at

a cost Cq2

2 , C c.

The timing of the game is now slightly dierent to reflect investments in product quality:

1) The intermediary contracts with sellers (announces PS or BS)

2) The sellers or the intermediary (depending on who owns the products at this stage)

invest in product quality

3) The intermediary sets the access price PC to consumers

4) Sellers or the intermediary sell products to consumers

By symmetry, all sellers products will be of the same quality, regardless of the mode

chosen by the intermediary. We assume that when the common quality is q, consumer gross

surplus from the n goods is V (nq), where V (.) is increasing and concave. This implies that

pP (n, q) = qV 0 (nq). In what follows, we will use V (x) = Ax, 0 < 12 . Finally, we also

assume consumers demand for aliation with the platform/merchant is linear: F (u) = u.

The following proposition contains the key result of this section.

Proposition 2 The total profits obtained under a platform mode, P , are higher than

15

the total profits obtained under a merchant mode, M , if and only if:

(1 ) (1 + )1+1

cC

1

< 1 (5.1)

Proof Under the merchant mode, the intermediary can buy the goods at cost f in

the first stage, invest in quality q and the sell them to consumers. In the second stage it

makes total profits:

maxPC

PC + nqV 0 (nq)

V (nq) nqV 0 (nq) PC

=(V (nq))2

4

Therefore, total merchant profits from the perspective of stage 1 are:

maxq

((V (nq))2

4 nCq

2

2 nf

)= max

q

A2n2q2

4 nCq

2

2 nf

leading to qM =hA2

2C

i 122

n2122 and:

M = (1 ) A2

4

A2

2C

1

n

1 nf (5.2)

A two-sided platform sets PS and in the first stage, which determine the quality

q chosen by sellers in the second stage, anticipating the price PC that the platform will

charge consumers in the third stage. Third stage platform profits are:

maxPC

PC + nqV 0 (nq)

V (nq) nqV 0 (nq) PC

=1

4

V (nq) (1 )nqV 0 (nq)

2

16

and consumer adoption of the platform is:

NC =V (nq) (1 )nqV 0 (nq)

2

In the second stage, all sellers choose the same quality q (by symmetry), given by6:

(1 )V 0 (nq)NU = cq

or:

(1 )V 0 (nq) V (nq) (1 )nqV0 (nq)

2= cq

Using V (nq) = A (nq), we obtain qP () =hA2(1)

2c

i 122

n2122 , where = (1 ).

The platform can then charge:

PS = (1 ) qPV 0nqP

V nqP (1 )nqPV 0 nqP 2

cqP2

2 f

This leads to the following expression of platform profits:

P = nPS +1

4

VnqP

(1 )nqPV 0

nqP

2=

1

4

hVnqP

2 (1 )nqPV 0 nqP 2i nc qP 22

nf

Plugging in the expression of qP () above, straightforward calculations yield:

P () =

1 (1 )1

1A2

4

A2

2c

1

n

1 nf

6Recall our assumption that sellers are small enough so that they ignore the eects of their price andquality investments on total consumer demand for the platform.

17

Optimizing over (implicitly over ), we obtain the optimal royalty rate chosen by the

two-sided platform:

=

1 +

Clearly, 0 < < 1. The optimal royalty rate for the platform trades o the need to

provide sucient investment incentives to sellers (which requires low ) against the need

to overcome the hold-up problem induced by third-stage pricing (which requires high ).

Total platform profits are then:

P =

1

(1 + )1+1

A2

4

A2

2c

1

n

1 nf (5.3)

Comparing the expressions of profits under the two dierent modes ((5.2) and (5.3)),

we have P > M if and only if condition (5.1) holds.

First, note that when C = c, i.e. the merchant is equally ecient in investing in

product quality as the sellers themselves, then the merchant mode is always preferred since

(1 ) (1 + )1+1 > 1 for all > 0. This is simply because the merchant mode avoids the

hold-up problem while still providing the optimal quality level.

If on the other hand C is suciently high relative to c, then the platform mode is

preferred by the intermediary. Indeed, in that case, the gains from devolving control

over and providing incentives for quality investments to sellers outweigh the drawback of

incurring the hold-up problem.

It should be clear that all of the above results hold when n is not exogenously given,

but chosen by the intermediary implicitly through its choice of PS.

18

6. Discussion

The formal analysis in the previous three sections makes it clear that the tradeo between

the merchant mode of intermediation and the two-sided platform mode runs far deeper

than the presence of indirect network externalities or absence thereof. In order to get a

sense of how the distinction between the two modes works in real-world examples, it is

useful to start with a brick-and-mortar retailer, such as Walmart.

In its simplest form, Walmart is a merchant. It first acquires many dierent products

from its suppliers and then resells them to consumers at its local store. Though it would

seem that the transactions Walmart conducts with suppliers, respectively consumers, are

independent of one another, whether or not there exist indirect network externalities de-

pends on the nature of the contracts with suppliers. If these simply specify a fixed buyout

fee at which Walmart acquires a given quantity of suppliers products, the latter do not

care about how many people visit Wal-Marts stores; they are only interested in the price

Wal-Mart bids for their products and the quantity it buys. Similarly, consumers care only

about the prices at which Wal-Mart sells these products to them, not the prices at which

it buys from suppliers, nor how many suppliers it contracts with.

In reality, Walmarts contracts with suppliers are more complex: they usually specify

a price per unit, a quantity it commits itself to buy, and also a price at which the sup-

plier has to re-purchase unsold units. Such contracts bring Walmart closer to a two-sided

platform mode by introducing indirect network eects. Suppliers must now factor into

their calculations consumer trac at Wal-Mart stores, upon deciding whether or not to

be "aliated" with Walmart. At the same time, this type of contracts improve eciency

through risk-sharing: Wal-Mart can order larger quantities upfront, rather than restricting

itself to a minimum in order to avoid accumulating unsold inventories.

19

But Wal-Mart can become two-sided in a more substantial way. Imagine that instead

of buying all products and reselling them, thus eectively taking ownership, Wal-Mart

rents shelf-space to some suppliers, say to Kellogg for its cereals, to Coke for its cans and

to Sony for its electronic devices. The suppliers are responsible for supplying, displaying,

pricing and advertising their merchandise, within the space allocated by Wal-Mart, and

they receive the revenue from sales to consumers. The eciency gains come from providing

suppliers with incentives to use price, advertising, and display to maximize profits, part of

which Wal-Mart can extract through higher rents. On the other hand, some eciency is

lost because suppliers do not fully internalize the eects of their in-store actions on other

suppliers (cf. section 4). And here too, Kellogg, Coke and Sony are more than a little

interested in the trac Wal-Mart generates, since this determines how many consumers

are likely to stop by their stands and eventually buy their products. It makes sense to

think of them as on board the Wal-Mart two-sided platform.

As Walmarts logistics have become more ecient and it has integrated more infor-

mation technology into its systems (for example its industry-leading Retail Link system,

enabling managers to monitor point-of-sale data and inventory store by store), it has man-

aged to shorten the time it holds inventory, devolve a higher share of risk back to its

suppliers and provide them with more flexibility in choosing the way they distribute their

products. The retailer has thereby significantly evolved towards a two-sided platform.

Interestingly, Amazon, the worlds leading online retailer, has followed a similar trend.

When Je Bezos started the company in 1995, it was very close to a pure merchant mode,

with the exception of certain contractual arrangements designed to transfer some inven-

tory risk back to book wholesellers. After developing a sophisticated e-commerce and

database infrastructure, Amazon started its "marketplace" initiative in 1999. Under this

20

initiative, the company began allowing some of its suppliers7 to operate their own store-

fronts on the Amazon.com website. Today, its contractual relationships with merchants

fall roughly into 5 categories: zShops for small merchants, Merchants@Amazon.com, Mer-

chants.com, syndicated stores and marketing deals, the last 4 targeting large business

sellers. These relationships dier in the share of inventory risk and control Amazon is tak-

ing on. For instance, under the Merchants@Amazon.com arrangement, suppliers can sell

products through Amazons website while still performing many of the commerce functions

themselves, such as maintaining ownership of and setting prices on inventory. By contrast,

under the Merchants.com arrangement, Amazon operates third-party supplier websites

(e.g. www.Target.com) but takes inventory in its distribution centers and completes order

fulfillment functions itself. In both cases, Amazon charges a fixed fee and a commission per

item sold, ranging from 5% (when Amazon does not take significant inventory functions

and risks) to 15% (when it does).

Turning now to digital music, as pointed out in the introduction, although at first glance

the iTunes digital music store looks like a two-sided platform, Apples almost absolute

control over music and movie pricing ($0.99 per song, $1.99 per TV series episode, $9.99

and $14.99 per movie downloaded), as well as over the consumer interface, makes it quite

merchant-like. The same holds true for may of the recently emerging Internet video portals,

Google Video being a case in point. Most of the videos available are free and for those

which are sold at a positive price, Google restricts providers to charging $1.99 or $3.99.

It is interesting to ask what would need to change in order for these sites to move closer

to a two-sided platform mode. An obvious part of the answer is the transfer of control over

pricing from the intermediary to the content providers. Presumably, this will happen over

7Amazon refers to its suppliers as "merchants", but in order to maintain consistency and avoid confusion,we will call them "suppliers".

21

time as consumers become more accustomed to such services, so that there is less value

created by uniform and centralized pricing. In fact, the trend towards decentralization of

control and therefore towards more two-sided platform-like intermediation models has al-

ready started taking place in the Internet video market. Some sites such as Brightcove.com

dierentiate themselves from Google through the flexibility they oer content providers in

choosing their revenue models - pay-per-view, pay-per-download or advertising-supported.

Brightcove even allows its content providers to have store fronts on its website from which

users can "jump" directly to the content providers websites.

Last, it is instructive to compare the respective modes of distribution of music con-

tent and of videogame content. Even at the very beginning of a new console generation,

videogames are almost never sold under a merchant mode by the console manufacturers8.

The reason has less to do with product variety (consumer demand for variety is arguably

equally high in both markets) than with the significant asymmetry of information and

need to preserve developer incentives to invest ressources in enhancing game quality in

the videogame case. By contrast, at the time music labels contract with digital distribu-

tors such as Apples iTunes or Real Networks Rhapsody, the "quality" of the songs (i.e.

their popularity) is usually known, therefore the intermediary mode chosen is determined

by other considerations, such as the ability of a merchant to extract higher profits from

consumers through creative pricing and bundling schemes.

The key insight which emerges from the preceding discussion and formal analysis is

that the distinction between the two modes of intermediation - merchant vs. two-sided

platform - is contractual. Where a given intermediary is located along the merchant-

8The exception is of course the one or two hit titles that every console buys out in order to ensure their

exclusivity and solve the initial "chicken-and-egg" problem (consistent with our section 3 above).

22

platform continuum depends on the allocation of control rights over the decision variables

impacting the sale of products to consumers, on the sharing of economic risk and on the

allocation of consumer ownership between suppliers and the intermediary.

7. Conclusion

In this paper we have shown that the tradeo between the merchant form of intermedi-

ary organization and the two-sided platform form is aected by several fundamental eco-

nomic factors: indirect network eects between buyers and sellers; asymmetric information

between sellers and the intermediary; investment incentives and product complementari-

ties/substitutability. The following table summarizes all of the relevant factors and their

eects on the desirability of each of the two modes of intermediary organization.

23

Note that this table and the analysis above hold true for a monopoly intermediary.

With competing intermediaries, more subtle strategic issues may arise which are beyond

the scope of this paper. For example, Hagiu and Lee (2006) studies the strategic eects

of intermediaries devolving control over content pricing to content providers, in a context

with intermediaries competing for content and consumers.

Clearly, this paper is only a preliminary treatment of the economic tradeos enumer-

ated above. The main contributions have been to establish a connection between the recent

literature on two-sided markets and the one on market intermediaries and to identify the

factors driving the choice between the two polar forms of market organization for interme-

diaries. This should provide the initial starting point for future economics research on this

topic.

References

[1] Armstrong, M. (2006) "Competition in Two-Sided Markets," forthcoming, Rand Jour-

nal of Economics, 2006.

[2] Biglaiser, G. (1993) "Middlemen As Experts," Rand Journal of Economics, Vol. 24,

pp. 212-223.

[3] Caillaud, B., and B. Jullien (2003) Chicken and Egg: Competition Among Interme-

diation Service Providers, Rand Journal of Economics, 34(2), 309-328.

[4] Evans, D. S. (2003) The Antitrust Economics of Multi-Sided Platform Markets,

Yale Journal on Regulation, 20(2), 325-82.

[5] Evans, D. S., A. Hagiu and R. Schmalensee (2006) "Invisible Engines: How Software

Platforms Drive Innovation and Transform Industries," MIT Press, October 2006.

24

[6] Hagiu, A. (2006a) "Pricing and Commitment by Two-Sided Platforms," forthcoming,

Rand Journal of Economics, 2006.

[7] Hagiu, A. (2006b) "Two-Sided Platforms: Pricing, Product Variety and Social E-

ciency," mimeo, Harvard Business School, 2006.

[8] Hagiu, A. and R. S. Lee (2006) "Exclusivity and Control," mimeo, Harvard Business

School, 2006.

[9] Leschly, S., M. J. Roberts, W. A. Sahlman (2003) "Amazon - 2002," Harvard Business

School case 9-803-098

[10] Rochet, J.-C., and J. Tirole (2003) Platform Competition in Two-Sided Markets,

Journal of the European Economic Association, 1(4), 990-1029.

[11] Rochet, J.-C., and J. Tirole (2004) "Two-Sided Markets: Where We Stand," forth-

coming, Rand Journal of Economics, 2006.

[12] Rubinstein, A. and A. Wolinsky (1987) "Middlemen," The Quarterly Journal of Eco-

nomics, Vol. 102, pp. 581-593.

[13] Schmalensee, R. (2002) "Payment Systems and Interchange Fees, Journal of Indus-

trial Economics, Vol. 50, 103-122.

[14] Spulber, D. F. (1996) "Market Microstructure and Intermediation," The Journal of

Economic Perspectives, Vol. 10, No. 3, pp. 135-152.

[15] Stahl, B. (1988) "Bertrand Competition for Inputs and Walrasian Outcomes," The

American Economic Review, Vol. 78, pp. 189-201.

[16] Yoe, D. B. (2005) "Walmart, 2005," Harvard Business School Case 9-705-460.

25