Making Markets

9 -8 1 8 -0 9 6
REV: FEBRUARY 28, 2018
THOMAS R. EISENMANN
SCOTT DUKE KOMINERS
Making Markets
Markets are everywhere—and where they’re not, you can build them!
Entrepreneurs have created and captured enormous value by designing and launching
marketplaces. Recent examples include product listing and exchange marketplaces like craigslist, eBay,
and Etsy; sharing economy marketplaces like Airbnb, Uber, and Upwork; crowdfunding and lending
marketplaces such as Kickstarter, GoFundMe, and Lending Club; and information marketplaces like
Gerson Lehrman Group and Quora. Internet giants like Alibaba and Google operate many
marketplaces at once—for products, information, media, and advertising.
New marketplaces address market failures. In many markets, high transaction costs and other
frictions combine with perverse incentives to produce suboptimal outcomes: parties transact when they
would have been better off doing so with others, and/or some opportunities for beneficial exchange
are missed altogether. When market failures occur, entrepreneurial opportunities arise; reshaping
markets creates value that can be captured. To fix markets, however, entrepreneurs must understand
how and why markets fail.
This note explains how to identify and capitalize upon marketplace design opportunities. The first
section defines markets and marketplaces and describes the basic functions of each. The second section
discusses attributes (e.g., heterogeneity of participants’ preferences and asymmetry in available
information) that determine whether and how marketplaces create value. The third section explains
common causes of market failure; the fourth presents a framework for designing marketplaces in
response. The fifth section discusses tactics for building trust and liquidity when launching new
marketplaces. The sixth and final section discusses challenges encountered as marketplaces mature
(e.g., congestion and disintermediation).
Markets and Marketplaces

What is a Market? A market encompasses all items that might be considered substitutes by
parties contemplating exchanges; all parties who, acting voluntarily in response to incentives, might
engage in such exchanges; and all settings in which those exchanges can occur.
Here, items can include goods, services, information, labor, or—as in the dating market—attention.
Substitutability means that when deciding whether to purchase an item, a buyer will compare it to her
best alternative; a seller will likewise take substitute opportunities into account. When exchange
Professors Thomas R. Eisenmann and Scott Duke Kominers prepared this note as the basis for class discussion. The authors appreciate the helpful
comments of Âriel de Fauconberg and Alexander Teytelboym.
Copyright © 2018 President and Fellows of Harvard College. To order copies or request permission to reproduce materials, call 1-800-545-7685,
write Harvard Business School Publishing, Boston, MA 02163, or go to www.hbsp.harvard.edu. This publication may not be digitized, photocopied,
or otherwise reproduced, posted, or transmitted, without the permission of Harvard Business School.
This document is made available to you compliments of Harvard Business School. Further posting, copying or
distribution is copyright infringement.
818-096 Making Markets
decisions hinge on the quantity and quality of available complements—as with computer operating
systems and compatible application software—then the items encompassed by the market are
substitutable sets of complements. a,1
Exchange implies that items are transferred from one party to another; in return, the recipient might
make monetary payment or transfer some other item(s). Parties may include intermediaries—for
example, brokers who introduce clients to potential transaction partners. The fact that parties act
voluntarily in response to incentives distinguishes market exchanges from transactions in planned
economies and from most resource allocation processes within corporations. Settings may include one
or more marketplaces, as defined below, as well as bilateral (i.e., one-on-one) interactions that occur
outside of marketplaces.
In exchanges, market participants must fulfill three functions in sequence. First, participants must
access each other to gauge interest. Next, the parties must agree upon terms of exchange. Finally, they
must process the exchange—transferring payment, for example, and arranging delivery. Completed
exchanges are called transactions.
What is a Marketplace? A marketplace gathers many potential transaction partners into a

venue—physical or virtual—and provides: (1) rules for participation, and (2) infrastructure to facilitate
transactions. Rules specify who can participate, which types of transactions are permitted and/or
executed, protocols for exchange and payment, how disputes are resolved, and so forth. Infrastructure
includes elements such as communication systems, procedures for verifying trading partners’
identities, payment processing, and logistical support.
Marketplaces help participants fulfill the three functions required to complete exchanges in
markets. Specifically, marketplaces help facilitate:
• Access to potential partners. Markets in which transaction partners are readily available are described
as being “thick”; formally, this means they have high liquidity and can support a high volume of
activity. b By gathering market participants into a single venue, marketplaces provide liquidity; they
make it less costly to discover new trading partners and to access past ones. Marketplaces may also
reduce the costs incurred in identifying and assessing potential partners by providing: (1) information
about demand-side partners’ requirements and supply-side partners’ pricing and item availability;
and (2) information about potential partners’ capabilities and reputations.
• Agreement on terms. Marketplaces may impose rules regarding permissible or required terms
of exchange, narrowing the scope of negotiations. (Amazon, for example, requires all third-
party vendors in its marketplace to accept returns.) Marketplaces may also help participants
select exchange partners or even constrain the set of possible exchanges. Additionally,
marketplaces may provide price-setting guidelines or mechanisms, such as specific auction
rules. Finally, marketplaces may provide communications infrastructure to facilitate
negotiations between potential transaction partners.
a Some items are unique; they are not available in multiple identical units. Unique items vary in the extent to which the existence
of similar (but not identical) items shapes potential partners’ assessments of the focal item’s exchange value. For example, a
purchase offer for Apartment 9E is likely to be influenced by the recent selling price for Apartment 10E in the same building,
which has the same layout and a slightly better view. Apartments 9E and 10E can therefore be considered to be in the same
market. By contrast, an item such as Da Vinci’s “Salvator Mundi”—which sold at auction in 2017 for $450 million, 2.5 times the
prior record for a painting—seemingly lacks substitutes that anchor its valuation, and thus comprises a market in and of itself.
b More specifically, liquidity reflects the degree to which an item can be quickly traded without affecting its price.
Making Markets 818-096
• Processing exchanges. Marketplaces leverage scale economies to reduce participants’ unit costs
for order and payment processing, shipping, and similar administrative components of
exchange. Marketplaces can spread the costs of processing infrastructure over the full volume
of participants’ transactions.
Not all marketplaces emphasize all three functions. Many marketplaces—for example, online
classified sites such as craigslist—focus only on partner discovery. Others eschew discovery and help
a set of participants who already know each other complete exchanges, as with the marketplaces used
to match medical school graduates to residency programs. Still others focus only on processing
exchanges that were previously agreed upon outside the marketplace, as with person-to-person
remittance systems like PayPal’s.
The canonical marketplace has a demand side that requires items and a supply side that offers them,
with multiple parties on each side, each of whom has broad discretion when setting exchange terms.
However, our definition of a marketplace allows for other configurations. For example, many auctions
have a single seller—or, in the case of request for proposal or procurement bidding processes, a single
buyer. Such auctions still have many bidders, and the auctioneer provides rules for participation and
infrastructure to facilitate transactions, consistent with our marketplace definition. Likewise, retailers
depart from the canonical model because their suppliers typically lack discretion over how their goods
are priced and presented. Nevertheless, retailers conform to our definition of a marketplace because
they gather many buyers and suppliers, and provide them with rules and infrastructure for exchange.
Not all markets have marketplaces. In some markets, participants engage only in bilateral
transactions. For example, there is no marketplace in which shippers book container ships; such
transactions are typically arranged by shippers contacting shipping lines directly or through shipbrokers.
Marketplaces may serve as venues for exchange in multiple markets. eBay, for example, offers
products across a huge range of product categories.
A marketplace will only emerge when a critical mass of market participants can expect to realize
greater value there than through relying solely upon bilateral transactions. Value in this context (often
called surplus) means total trading gains, net of expected transaction costs.
• Trading gains for buyers reflect the difference between their maximum willingness to pay for
items and the prices they pay; for sellers, trading gains reflect the difference between the prices
they receive and reservation prices—the lowest prices sellers are willing to accept. c
• Transaction costs include all costs market participants incur in identifying and negotiating with
potential transaction partners, and then processing the resulting exchanges. When participants
use a marketplace, their transaction costs include charges levied directly by the marketplace
(e.g., membership or transaction fees) plus other expenses required in using the marketplace
(e.g., training costs for employees).
c In unpriced markets, such as the dating market, trading gains can be construed to be based on “shadow prices,” that is,
measures of utility that actors would realize from transactions, compared to the status quo. Individuals will participate in dating
marketplaces if the expected utility from participation (i.e., the happiness derived from successful matches plus the
entertainment value of searching for a romantic partner) exceeds transaction costs (i.e., out-of-pocket costs for participation, the
opportunity cost of time spent searching, and the disutility derived from rejection).
Market Attributes
Several attributes of markets determine whether and how marketplaces can create value for
participants.
Heterogeneity. Some markets feature substantial heterogeneity in buyers’ requirements and/or

sellers’ offers. In the presence of heterogeneity, marketplaces that attract many participants improve the odds
of matches that can yield trading gains. Home buying is an example: buyers’ preferences vary considerably
regarding architectural style, layout, lot size, landscaping, street type, and distance to amenities; homes,
meanwhile, vary substantially with respect to those attributes. Consequently, a marketplace that provides
access to nearly 100% of homes for sale in a community—such as a regional real estate association’s multiple
listing service (MLS)—should attract potential buyers, who in turn will draw sellers.
Items in other markets differ across fewer dimensions; these dimensions can be used to categorize items
into different groups (for example, Hard versus Soft Red Winter Wheat; Brent Blend vs. Dubai Crude).
Homogeneous items—commodities—are exchanged by many buyers and sellers in very large volumes.
Commodity markets are often thick enough to allow continuous trading on exchanges—marketplaces that
organize double auctions where buyers and sellers regularly post bids and asks, respectively.
Synchronicity. Markets also vary in the extent to which participants are able and willing to
coordinate their evaluations of potential transaction partners. Sometimes, market participants can
synchronize activity without relying on a marketplace. U.S. colleges are an example: they have loosely
coordinated their admissions deadlines, and many use a common application form. In other cases,
instituting marketplaces helps ensure synchronicity. Auction houses aggressively promote their
offerings to gather a critical mass of potential bidders. Likewise, companies in industries with a high
volume of new product introductions often synchronize product launches at industry conventions.
(For example, many new toys are unveiled at TOY FAIR, held every February in New York City, which
attracts 1,500 manufacturers and distributors along with buyers from 5,000 retail outlets.) The
alternative—launching new products year-round, then conducting bilateral meetings to pitch those
products to retail buyers, would require high travel costs, and would forfeit opportunities to connect
with buyers who are too small to warrant a trip.
Synchronization is intrinsically difficult in some markets. For example, families require short-term
childcare services on different days—often with little lead-time. Likewise, the market for homes
generally lacks synchronization (although some homes are sold at auctions). Homebuyers arrive on the
market year-round as family circumstances change. In response, sellers may elect to keep their property
on the market for months, waiting for buyers whose preferences precisely match the property’s
attributes and who thus should be willing to pay high prices.
Sequential Evaluation. Markets that lack synchronicity often require sequential evaluation of
offers. For example, when hiring for specific management positions, companies typically make offers to
candidates one at a time; if a candidate declines, the company will continue searching or will extend an
offer to the next-best remaining candidate. In hiring searches, sequential evaluation makes sense because
finding candidates can take many months, and also because interest might be deterred by the reputational
consequences of a process that extended multiple offers simultaneously, then withdrew outstanding
offers after a candidate was hired. Sequential evaluation poses challenges for demand-side participants,
however, because those participants must decide whether passing on a “good-but-not-great” candidate—
and thereby extending the search—is likely to yield a better match in the long run. 2
Marketplaces can increase the efficiency of sequential search by bringing many candidates together
and aggregating their information into a uniform format. Additionally, marketplaces can sometimes
transform sequential search processes into simultaneous ones (as in online labor markets like Upwork,
which are so liquid that the possibility of simultaneous offers does not deter participation).
Information Asymmetry. Markets vary widely in the extent to which the information that
would be helpful to demand-side participants when evaluating supply-side offers is (1) readily and
freely available; (2) privately available at some cost; or (3) deeply “impacted,” because supply-side
participants have an incentive to not disclose it. The used car market provides an example. Potential
buyers can access free information services that provide data on the overall reliability of cars from each
model year; they can also purchase CARFAX reports that show whether specific vehicles had ever
deployed their airbags, been used as taxis, or failed state emission tests. But without thorough
inspections by experienced mechanics, most buyers are not able to gauge used cars’ reliability
accurately—and sellers are unlikely to volunteer information about potential problems. Used car
marketplaces can afford to provide more detailed quality certifications, and sometimes can even offer
insurance; this, in turn, makes buyers more willing to participate in exchanges.
Information asymmetry tends to be a more serious problem in markets for highly heterogeneous
items (such as homes, jobs, and romantic relationships). In such markets, marketplaces often focus on
information provision—either incentivizing participants to reveal their information truthfully, or
verifying participant information directly.
Price-Based Matching. In many markets, buyers and sellers focus almost exclusively on getting
the best prices possible when choosing counterparties. Often, matching on price entails either an
explicit or implicit auction. However, when a lack of synchronicity precludes an auction and instead
requires sequential evaluation of offers—as with individuals listing used cars on craigslist—some
sellers and buyers still focus solely on price when deciding when to halt their evaluation processes and
take an offer in-hand. Parties initiating transactions are more likely to rely on price-based matching
when (1) they have no concerns about counterparties’ abilities to reliably complete exchanges; and (2)
choosing one counterparty over another has no signaling consequences. d
In some markets, price is just one of many factors that determine matching outcomes. Unlike
physical goods, services cannot be inspected before parties commit to transactions; they must be
delivered afterward. Hence, buyers of services cannot be confident that low bidders will meet
specifications. For this reason, the winner of a request for proposal process (for example, to retain a
consulting firm) may not be determined solely on price; vendors’ reputations often play an important
role in selection. Likewise, a job seeker with multiple offers will not necessarily accept the one with the
highest salary; she also will consider each position’s responsibilities, location, advancement
opportunities, cultural fit, prestige, and so forth.
Finally, in some markets, pricing plays little or no role at all. In certain settings, such as college
admissions, applicant attributes that are of great concern to colleges are not correlated with candidates’
willingness and ability to pay. In other settings, such as the market for organ transplants, concerns
about fairness or social repugnance preclude price-based matching. There is special scope for
marketplace design when price is not a factor, as in such settings, parties lack a simple mechanism for
identifying which exchanges to undertake. And indeed, for example, organized marketplaces have
been quite successful in matching lower-income students to colleges (“The Posse Foundation:
Implementing a Growth Strategy,” HBS No. 309-056) and organ donors to recipients (“Kidney
Matchmakers,” HBS No. 908-068).
d With respect to the second criterion, sellers may prefer to transact with buyers who are well regarded by industry peers and
who thus can serve as references to help attract future customers—even if those buyers have not offered the best price.
Price Dynamism. In some markets, prices are dynamic—they adjust constantly based on the
relative balance of supply and demand. In other markets, prices are adjusted infrequently. Fixed pricing
may be a consequence of transaction costs—for example, the expense a grocery store would incur to
reprice items already on its shelves. e In other circumstances, concerns about fairness and reputational
consequences prevent suppliers from raising prices despite strong demand. For example, in the wake of
a disaster, many retailers of food, gasoline, and building materials refrain from price gouging. Many
entertainers likewise avoid pricing a venue’s best seats at rates that would offend loyal fans.
By bringing together many participants, marketplaces can enable dynamic price discovery, as with
commodity exchanges and rideshare surge pricing. Additionally, marketplaces can facilitate pricing
control, either by subsidizing participants temporarily (e.g., compensating drivers for picking up riders
in a snowstorm, while holding the rider price fixed), or by restricting participation (e.g., selling only
one ticket to each potential buyer, to prevent scalpers from acquiring many tickets).
Proximity. A market’s geographic boundaries may vary widely, depending on item attributes.
For example, the market for diamonds is global, because the low cost of transporting diamonds, relative
to their price, makes all diamonds of a given quality substitutes. By contrast, markets for babysitters
are local, being limited by caregivers’ commuting distances. Marketplaces can be especially valuable
in localized markets because they enhance liquidity—and they can sometimes expand market scope by
attracting participants from nearby locales.
Perishability. When items are perishable, sellers and buyers may find it advantageous to
congregate in a marketplace. For example, local farmers markets allow buyers to shop for produce from
many vendors. The alternative, visiting each farmer, would take far more time—especially if the buyer
wished to compare offers before making purchases.
Continuity. In some settings, market participants care about whom they transact with. If many
marketplace participants already have strong or contractually-binding relationships with preferred
trading partners, then partner discovery is not a priority, and marketplaces may principally add value
by reducing transaction processing costs.
Market Failure
Economists say that markets fail when they reach suboptimal outcomes—for example, by not
putting resources to their most productive (“efficient”) uses, or by allocating resources in ways deemed
socially unacceptable. Market failure sometimes entails an absence of trading, when mutually
beneficial exchanges are possible, but do not occur. Often, participants transact with the wrong
partners—for example, buyers who could have found a product that better fit their needs, or who could
have purchased the same item at a lower price from a different vendor. Other times, transactions occur,
but there is concern about unfairness or coercion in the transaction process.
Typically, markets fail for endogenous reasons—that is, for reasons that arise from within the markets
themselves—often with participants behaving rationally given the incentives they face and the
information at their disposal. Common causes of market failure include (1) high transaction costs that
tend to limit liquidity; (2) congestion—consequences of too much participation; (3) problems with
participants’ incentives; and (4) problems with intermediaries’ incentives.
e These sorts of costs are sometimes called menu costs, after the idea of having to reprint menus to change prices in restaurants.
Transaction Costs
Transaction costs are a common cause of market failure: 3 High transaction costs reduce the number
of participants who can trade profitably. Market thinness is self-reinforcing; an agent who expects to
find few suitable exchange partners will be less likely to participate in the market, in turn reducing the
odds that others will do so.
Transaction costs can be direct—as in the case of payment processing charges—or can arise indirectly
from market features such as the need to search for candidates. For example, consider the challenge
confronting parents who are seeking childcare providers soon after moving to a new neighborhood:
Without friends and family to offer referrals, parents need to identify and locate available candidates,
then confront the time-consuming task of scheduling face-to-face meetings—an imperative before
entrusting their children to strangers. More time might be spent checking references, without full
confidence in their veracity. The parents might feel an obligation to interview several candidates before
choosing one. Once vetted, hiring a childcare provider may entail stressful negotiations over
compensation, plus further messaging and/or phone calls for scheduling. Given the time required to find
and book a childcare provider, “date night” might become a casualty of market failure.
In many markets, the Internet has reduced transaction costs significantly. Nowadays—continuing
the previous example—parents might seek childcare providers through websites that aggregate
caregiver profiles, such as Care.com and Sittercity. Compared to pre-Internet options (e.g., flyers in
coffee shops and ads in newspapers), listing on websites is easier and cheaper for caregivers. Websites
can thus provide parents with access to more candidates, along with richer information about each;
this, in turn, can substantially reduce parents’ search costs. The Internet has had a similar impact on
market after market, reducing the transaction costs associated with exploring and comparing options.
However, the Internet is not a panacea: It does not eliminate search costs—it only reduces them. And
in some cases, the Internet can actually increase search costs, as in the cases of listings that are “stale”
or contain deliberate obfuscation or misrepresentation. 4
Congestion
Thickness is normally viewed as a positive attribute of markets, but high levels of participation can
lead to market failure if not managed properly.
First, there is concern about business stealing: Buyers or suppliers may worry about pricing pressure
(or simply being left out of exchanges entirely) if too many direct rivals or competitors also participate
in the market. The possibility of business stealing may deter entry, as prospective entrants may be
concerned that they will not be able to recoup their costs of participation. Additionally, concerns about
business stealing might compel market participants to contract early, before they have had
opportunities to evaluate all of their possible options. These issues were a major barrier to the
development of online business-to-business exchanges during the late 1990s.
If market participation is inexpensive, meanwhile, the cost of screening offers may itself deter
participation. Companies, for example, are often deluged with resumes from job seekers who can
respond, free of charge, to online “help wanted” queries. Likewise, online dating platforms that allow
free messaging can deter participation by individuals who attract an excessive volume of unwelcome
expressions of interest.
Participants’ Incentives
Many market failures arise when participants—acting in their own self-interest—simply respond
to incentives the market creates.
Information Asymmetry. In markets, information asymmetry—an uneven distribution of

information about an item’s true quality—can take two forms, each of which can contribute to market
failure.
Private information may be distributed unevenly among rival participants—as is often the case in
“common value” auctions, in which the items being sold would have the same final value for all
bidders (e.g., Treasury bills or oilfield leases), but bidders have their own forecasts of the final value,
based on the information at their disposal. The bidder who wins a common value auction could
conceivably have superior information that truly justifies the amount they bid, but more typically, the
winner has overestimated the item’s value in making predictions based on incomplete information;
this tendency to overpay is called the “winner’s curse.” Savvy bidders may anticipate the winner’s
curse and shade their bids accordingly. However, research shows that even experienced bidders—for
example, construction companies and telecommunications carriers—are often vulnerable to the
winner’s curse. 5 Market efficiency suffers when firms are bankrupted by overbidding.
Asymmetric information between sellers and buyers also can lead to market failure—especially
when an item’s quality: (1) can vary widely; (2) is known to one party in the exchange; but (3) is difficult
for other parties to discern. Used cars are an example, as explained in Professor George Akerlof’s classic
paper, “The Market for Lemons.” 6 In used car markets, buyers must offer prices that average the values
of high- and low-quality cars. Potential sellers who know they have high-quality cars—”peaches”—
will not transact at the average price, so they leave the market. Eventually, only lemons are offered,
and the market fails.
Unraveling. Intensely competitive markets can “unravel” when demand-side participants make
preemptive—and sometimes “exploding”—offers to supply-side candidates based on limited quality
indicators available when the offers are made. Examples of markets that have unraveled in the U.S.
include college admissions; the recruitment of college athletes and consulting firm summer interns; the
placement of medical school graduates into residencies; and the appointment of law school graduates
as judicial clerks. Unraveled markets fail to the extent that (1) some candidates who accept preemptive
offers have strong early quality signals, but turn out to be weaker than expected; and (2) some
candidates accept early offers even though those candidates could secure more preferred offers if they
were to wait.
Unraveling is more likely to occur when markets: (1) do not employ price-based matching; (2) lack
enforceable rules on timing; (3) are loosely synchronized; and (4) feature strong preference ordering on
both sides of the market. According to Professor Alvin E. Roth, unraveling in labor markets is often
driven by recruiters who rank just below those that are most preferred by job seekers; these “second-
best” recruiters make preemptive offers to candidates who, based on early signals, show potential to
be at the top of recruiters’ preference orderings. 7 The candidates, unsure of their true quality levels,
must make “bird-in-the-hand-versus-two-in-the-bush” choices between accepting early offers from
“good” recruiters and waiting to see if better offers from “great” recruiters are forthcoming.
Holdout Risks. The value of some assets hinges crucially on the control of complements.
Examples include bundles of related patents, radio spectrum in contiguous geographic markets, and
adjacent parcels of land required for major real estate developments. In the presence of complements,
failure to acquire any one element would substantially reduce the value of all the other assets.
Consequently, when complementary asserts are acquired in sequence, there is a risk that a potential
seller will hold out late in the process, demanding a price that would capture most of the rent available
from assembling the full set of items. Due to holdout risk, some potentially profitable projects will be
abandoned or passed over entirely—a market failure. 8
Externalities. An externality is a benefit or harm experienced by some party due to the actions of
some other party, with no compensating payment between the parties. In the case of a positive
externality, B pays no price for a benefit provided by A (e.g., B admiring flowers planted in A’s yard).
With negative externalities, B receives no compensation for harm caused by A (e.g., B suffering stench
from A’s upwind pigpen). Absent intervention—for example, by government regulators—that
“internalizes externalities” by compensating (or charging) the responsible parties for the benefit (or
harm) experienced by others due to their actions, markets with externalities fail to reach socially
optimal outcomes. 9
Network effects in markets entail externalities. When a new participant joins a market, she boosts
the value of affiliation for potential trading partners—a “cross-side” network effect—and she typically
is not compensated for this benefit. “Same-side” network effects—in which participants affect the value
of participants on the same side—are likewise externalities. When a new participant joins a market, for
example, she reduces the value of marketplace affiliation for potential rivals, due to concerns about
business stealing—without bearing this cost.
Absent a way for participants to internalize positive externalities, new participants are less likely to
join, and a market’s growth could stall. Participants on each side of multi-sided markets thus often
confront a “Catch-22:” they delay entry until they are confident there will be enough participants on
the other side, but this leads the other side to delay entry in response. Furthermore, when each side of
a market is highly fragmented, participants cannot easily signal their intentions to join, so they cannot
be assured that others will do so. No one moves unless everyone moves, so no one moves. Economists
Joseph Farrell and Garth Saloner labeled this scenario the “penguin problem:” Hungry penguins gather
at the edge of an ice floe, reluctant to dive into the water. There is food in the water, but a killer whale
might be lurking, so no penguin wants to dive first; hence, all remain hungry. 10 In the presence of
“penguin problems,” behavior that is rational for each individual agent may lead to market failure.
Public Goods. Public goods are items that can be consumed by one party without reducing their
availability to other parties (i.e., they are “non-rivalrous”). Examples include national defense, law
enforcement, and radio broadcasts. People systematically under-contribute to the provision of public
goods because they can “free ride”—that is, people can enjoy the benefits of others’ contributions, and
so are incentivized to contribute less themselves. f
Intermediaries’ Incentives
Some markets fail due to the behavior of intermediaries who operate marketplaces or serve a
brokerage role in connecting market participants.
Insufficient competition. A market may be dominated by one or just a few marketplaces due
to entry barriers, in particular, the combination of strong network effects and high switching and/or
multi-homing costs. g Monopolies and oligopolies can lead to market failure when they increase the
f Due to free rider problems, public goods are often provided by governments and funded by taxes.
g Homing costs encompass investments, out-of-pocket expenses, and inconveniences borne by participants due to affiliation with
marketplaces or brokers; multi-homing costs reflect the sum of homing costs for participants who affiliate with multiple
marketplaces or brokers. Switching costs are out-of-pocket expenses and inconveniences incurred by participants when they
cost of marketplace participation, driving some participants from the market. Likewise, in the absence
of pressure from rivals, dominant marketplaces may fail to innovate in ways that would reduce
transaction costs. 11
Intermediaries’ conflicts of interest. Brokers play important roles in many markets. Examples
include real estate agents, independent insurance agents, stockbrokers, travel agents, talent agents,
headhunters, independent sales representatives, and personal shoppers. Brokers are retained by clients
who participate in at least one side of the market. Brokers help their clients identify potential
transaction partners, and sometimes help clients negotiate exchanges.
Conflicts of interest are pervasive in brokerage relationships. Some conflicts are engendered by
pressures or incentives that lead brokers to serve the interests of participants on the other side of the
market, at the expense of their clients. For example, stockbrokers are expected to steer their clients to
investments that match their clients’ preferences—yet stockbrokers may be instructed to push
securities underwritten by their employers. Stockbrokers may also be tempted to engage in the illegal
practice of “front running,” in which they trade in advance of a client’s big order, anticipating that the
order will move the market price. Likewise, airlines sometimes offer travel agents special bonuses
(called “overrides”) for selling tickets on specific routes; this creates incentives for travel agents to
suggest travel plans that may not be ideal for their clients, in terms of cost and/or convenience.
Other conflicts stem from the combination of principals’ inability to monitor intermediaries’
behavior, combined with a lack of incentive alignment. A residential real estate broker who represents
a home seller, for example, will split a 6% commission (paid by the seller) with the buyer’s broker.
When deciding whether to accept an offer or wait for a better one, the seller stands to gain 94% of any
improvement. The broker will keep only 3%, and must balance this potential upside against the time
required to show the property to more potential buyers.
Marketplace Design
Sometimes, it is possible to mitigate (or even eliminate) market failures through the design of new
marketplaces—that is, new rules and/or infrastructure for exchange. The optimal design response
depends crucially upon the type of market failure: Marketplace design often addresses structural
failures that have arisen endogenously as markets developed—such as perverse incentives—through
transaction guidelines and monitoring, or through mechanisms that determine which exchanges to
execute. Meanwhile, marketplaces often respond to underlying attributes of markets that cause
failure—such as transaction costs and information asymmetries—by developing new infrastructure to
support participation.
Design choices not only determine how well marketplaces respond to market failures, but also
govern how and when marketplaces can succeed. Sometimes, a marketplace designer may have to give
up on (or defer) a design that would be ideal at scale in exchange for one that can attract initial
participants who create liquidity.
To think about marketplace design, we use a simple conceptual framework, pictured below: Market
attributes determine market structure, and both can lead to market failure. Marketplace design
addresses market failures, and the chosen design determines the challenges that arise at launch and
those that are ongoing (the latter of which also depend on the process and outcomes of the launch).
abandon one marketplace or broker in favor of another. For more background on homing and switching costs—and discussion
of differences between them—see the Appendix to “Winner-Take-All in Networked Markets” (HBS No. 806-131).
10
Market Launch
Attributes Challenges
Market Marketplace
Failure Design
Market Ongoing
Structure Challenges
Source: Casrewriter analysis.
Marketplace designers have many different design levers to select from. Here, we introduce the “7
Ps,” a mnemonic for analyzing marketplace design choices:
• Pairing reflects mechanisms used to match transaction partners, for example, auctions;
algorithmic recommendations; and filtering, sorting, and search systems.
• Permissions are rules regarding who can participate in the marketplace and what types of
transactions are allowed.
• Profiling encompasses information presented by the marketplace about participants (e.g.,

reputation scores and transaction histories) and participants’ requirements or offers (e.g.,
product attributes and availability).
• Processing entails activities undertaken by the marketplace to facilitate transactions, for

example, transferring payment and arranging shipping.
• Pricing reflects the means through which participants compensate the marketplace, for
example, subscriptions, space rental, and transaction fees.
• Promotion makes potential participants aware of the advantages of trading in a marketplace

and types of transactions that are possible.
• Priming entails actions through which a marketplace prepares items for exchange, instead of
relying solely upon the items’ suppliers to do so.
Pairing
Marketplaces can take four broad approaches to matching: relying upon (1) price-based pairing
through auction bids; (2) prescribed pairing, with human matchmakers or algorithms specifying a single
match; (3) proposed pairing, with marketplace recommendations that provide leeway for participants to
select other partners; or (4) passive pairing, with participants screening options using data and tools—
but not recommendations—provided by the marketplace.
Each pairing approach can be combined with partitioning, in which the marketplace interposes itself
between supply- and demand-side users, so that participants on each side transact with the
marketplace, rather than interacting directly with individual participants on the other side. When
partitioning marketplaces, the marketplace organizer may implicitly or explicitly restrict the set of
participants that can transact with each other.
11
Price-Based Pairing. Some marketplaces match exchange partners solely based on prices set
through price discovery processes—explicit or implicit auctions. The main varieties of auctions include:
• Ascending auctions (also called open or English auctions), in which a price visible to all
bidders is successively raised (by either the auctioneer or the bidders, depending on the
auction’s format), until a single bidder remains and wins the item at the final price. Note
that the final price is typically just a bit higher than the maximum price that the runner-up
was willing to bid.
• Descending auctions (also called Dutch auctions because the format is used to sell flowers in
Holland), in which the auctioneer starts with a high price, which is successively lowered until
a bidder accepts the current price; that bidder wins the item and pays the accepted price.
• First-price sealed-bid auctions, in which bidders simultaneously submit bids without seeing
each other’s bids, and the highest bidder wins the item and pays the price she bid.
• Second-price sealed-bid auctions (also called Vickery auctions after an economist who studied
them), which work like first-price sealed auctions, except that the winner pays a price equal to
the second-highest bidder’s stated bid.
• Double auctions (used in continuous limit order books), in which both buyers and sellers state
“bid” and “ask” prices, and whenever some buyers’ bids exceed some sellers’ asks, those buyers
and sellers transact at a price between the bid–ask spread.
Variations of these auction formats exist. For example: eBay employs ascending auctions, but the
auctions have a fixed closing time, and bidders often delay their bidding until the last minute in hopes
that other bidders will not have an opportunity to respond. 12
It can be difficult to predict which auction format will maximize either total trading gains or the
auctioneer’s revenue. Upon first inspection, descending and first-price sealed-bid auctions seem like
they might have an edge, because their winners pay the amount bid, whereas winners of ascending
and second-price sealed-bid auctions pay the amount the runner-up bid. However, if bidders know the
winners will end up paying their bids, then they will never bid their true values—if they did, then they
would realize no trading gains. Ascending and second-price sealed-bid auctions, by contrast, make it
safer for participants to bid their true values, since whoever wins will always net a profit equal to the
range between her bid and the second-highest bid. h
In ascending auctions and double auctions, meanwhile, participants can become aggressive after
they observe other participants’ bidding behavior. Sealed bid and descending auctions, by contrast, do
not convey information about other bidders’ valuations. But ascending auctions can pose problems
attracting bidders, because potential participants may assume that the party with the highest valuation
will always win. Double auctions may also have trouble attracting participants, due to concerns that
they reveal too much information to prospective competitors. If many rivals avoid the auction, then
the expected winner may be able to acquire the auctioned item at a bargain price. Sealed bids, by
contrast, offer weaker bidders the possibility that the expected winner will shade her bid in hopes of
realizing trading gains; in that scenario, a strong bid by a weaker player might be successful.
h In private value auctions, in which each bidder derives her own personal value for acquiring the item, independent of others’
values (as with event tickets), it is optimal for bidders to bid their true values in both ascending and second-price sealed-bid
auctions.
12
Competition among bidders typically drives up the prices in auctions; hence, collusive agreements
among bidders often seek to reduce participation. George Washington Plunkitt, a powerful member of
the Tammany Hall political machine in late 1800s New York, described one particularly colorful
example of what he called “honest graft” but today we would call illegal collusion:
[T]he city is repavin’ a street and has several hundred thousand old granite blocks to
sell. [... A] newspaper [...] got some outside men to come over from Brooklyn and New
Jersey to bid against me.
Was I done? Not much. I went to each of the men and said: “How many of these
250,000 stories do you want?” One said 20,000, and another wanted 15,000, and [the] other
wanted 10,000. I said: “All right, let me bid for the lot, and I’ll give each of you all you
want for nothin’.”
They agreed, of course. Then the auctioneer yelled: “How much am I bid for these
250,000 fine pavin’ stones?”
“Two dollars and fifty cents,” says I.
“Two dollars and fifty cents!” screamed the auctioneer. “Oh, that’s a joke! Give me a
real bid.”
He found the bid was real enough. My rivals stood silent. I got the lot for $2.50 and
gave them their share. 13
Even in the presence of modern antitrust enforcement, auctioneers must be careful to design their
marketplaces in ways that avoid providing subtle opportunities for collusion. In ascending auctions
for wireless spectrum, for example, participants have sometimes attempted to use early bids to signal
their intentions to rivals: In 1997, U.S. West, which was eager to acquire wireless spectrum in Minnesota
in auction lot 378, bid $313,378 and $62,378 in earlier lots for spectrum in Iowa—a priority market for
U.S. West’s main rival for lot 378, and a market where U.S. West previously had shown no interest in
acquiring spectrum. The interpretation was that U.S. West was attempting to communicate its
intentions regarding lot 378 to its rival, in hopes the rival would stay out of that auction in exchange
for U.S. West reducing competition elsewhere. 14 Messaging through bids can be prevented by
requiring bidders to follow prescribed increments, or bid round numbers.
Prescribed Pairing. With prescribed pairing, a marketplace owner uses a non-price mechanism
to identify exchanges deemed to fit the agents’ preferences. The exchanges are typically presented to
partners as “take-it-or-leave-it” choices (as in school choice programs) or sometimes even binding
assignments (like in the assignment of cadets to branches of military service); sometimes, participants
are instead presented a small list of several recommended matches.
Prescribed pairing can be based on human curation, with marketplace staff making matches based on their
understanding of parties’ preferences. For example, Dating Ring relies on human matchmakers (assisted by
algorithms), as does Poppy, a service that connects families with childcare providers, on demand (“Updating
Dating,” HBS No. 818-052; “Poppy: A Modern Village for Childcare” HBS No. 818-075).
In other cases, prescribed pairing is based purely on algorithmic results. When an individual requests a
ride on Uber, for example, Uber contacts a single nearby driver; if that driver declines the trip, then Uber
contacts another. To each rider, Uber likewise presents only a single driver who has accepted the match.
Some marketplaces employ market-clearing algorithms to simultaneously match pools of participants.

Unless carefully designed, such algorithmic matching systems can be prone to gaming behavior that can
13
lead to market failure. Consider, for example, the problem of matching all medical school graduates to
training residencies. Each graduate has preferences regarding hospitals, based on the hospitals’
reputations, as well as personal factors such as the desire to live near family. Likewise, each hospital has
preferences regarding graduates, perhaps based on their academic performance and medical schools’
prestige. A straightforward algorithmic solution is to (1) ask each side to rank their preferences; then (2)
match all parties who have listed each other as their #1 choice; and then (3) apply rules that give
remaining parties as high a choice as possible (e.g., after mutual #1 matches are made, match graduates
to their #1 choice whenever that hospital lists the graduate as their #2 choice, and so forth).
Now consider the incentives of a (soon-to-be) medical school graduate whose true #1 choice is the
Mayo Clinic, but who believes that Mayo may prefer other strong candidates. She ranks her true #2
choice, Dana-Farber, as her #1 pick—aiming to ensure a spot there. If, consistent with her true
preferences, she had instead listed Mayo #1 and Dana-Farber #2, then she would have risked being
passed over by Mayo. She likewise would have risked Dana-Farber filling its spots with candidates
who listed it #1; in that scenario, the graduate would have ended up—at best—with her true #3 choice.
The graduate’s ranking strategy is not truthful—but it is rational, given the uncertainty and
incentives she faces. But if, unbeknownst to the graduate, Mayo has ranked her #1, then her strategy—
and the market—has failed to achieve the optimal outcome. Moreover, if Mayo were to call the
graduate after the assignments had been made, it is possible that the graduate would back out of her
assigned job with Dana-Farber, creating instability in the market, and forcing Dana-Farber to scramble
and look for a replacement.
In the context of residency matching, the deferred acceptance algorithm, first formalized in 1962 by
economists David Gale and Lloyd Shapley, 15 yields a better outcome: one with stable matches—such
that no agent A can find a new partner X who A prefers to its assigned partner Y, while X in turn would
also give up its assigned partner B in exchange for A, if A were to offer.
The deferred acceptance algorithm works as follows:
1. Graduates and hospitals privately and simultaneously submit preference lists to the
marketplace organizer.
2. On behalf of each hospital, the algorithm (internally) extends provisional offers to the hospital’s
top choices, in descending order, until all of the hospital’s spots are filled. Then, on behalf of each
graduate who received at least one offer, the algorithm tentatively accepts the offer that the
graduate had ranked highest; these provisional offers and tentative acceptances are not visible to
either side (they are just used as inputs by the algorithm for subsequent rounds of the algorithm).
3. In each subsequent round, for every hospital that had an offer rejected in the prior round, the
algorithm extends an offer for that spot to the graduate who: (a) is ranked highest by the
hospital, but (b) has not yet received an offer from the hospital. Then, on behalf of every
graduate who has one or more new offers, the algorithm tentatively accepts their highest ranked
offer—even if doing so entails rejecting an offer tentatively accepted in a prior round.
4. The algorithm proceeds until no graduate would wish to accept a new offer. At that point, the
tentatively accepted matches are finalized, and the resulting assignment of doctors to hospitals
is stable. i
i As an exercise, consider why deferred acceptance results in a stable outcome. To simplify, start with the case in which each
hospital has only one open position.
14
Proposed Pairing. Many marketplaces propose potential transaction partners—based on

analysis of participants’ attributes, stated preferences, and/or transaction histories. Amazon, for
example, recommends products in many different ways, including, on product pages, lists that show
items “Frequently Bought Together”; “Sponsored Products Related to This Item”; “Customers Who
Bought This Item Also Bought”; and “Customers Who Viewed This Item Also Viewed.” Elsewhere,
Amazon provides recommendations based on a customer’s transaction history.
Algorithms, human curators, or a combination of the two may make recommendations. Often, multiple
recommendations are proffered, with no expectation that participants must select from among them.
Passive Pairing. Some marketplaces provide no specific recommendations regarding potential

exchange partners. Instead, they may provide access to directories or databases that include profiles
and reviews of marketplace participants and their offerings, along with tools for searching and
screening this data; participants are expected to use these tools to find exchange partners for
themselves.
Marketplaces can provide information that steers participants to potential partners without
explicitly recommending matches. Lists of “top sellers” play this role, as do prominently displayed
“featured items” (for example, at checkout in a grocery store). Participants may pay a premium for
prominent placement. Likewise, some marketplaces charge an extra fee to suppliers to highlight an
offer (e.g., adding a logo or additional information), to certify certain offer features, or to move an offer
higher in users search results.
Partitioning. Each of the pairing approaches described so far can be combined with partitioning,
through which the marketplace interposes itself between supply- and demand-side participants. In
partitioned marketplaces, participants are required each to transact with the marketplace, rather than
transacting with counterparties directly.
Examples of marketplaces that partition include Uber, which presents (via prescribed pairing)
participants on each side of the market with at most one potential partner at a time; Poppy, which does
the same with families and caregivers; and Shift, thredUP, and Gazelle, which acquire used goods—
cars, apparel, and electronic devices, respectively—then recondition and (via passive pairing) resell
them (“Shift Technologies, Inc.,” HBS No. 818-002; “thredUP: Think Secondhand First,” HBS No. 817-
083; “Gazelle in 2012,” HBS No. 711-446). Traditional retailers, which buy new goods from vendors
and then resell them to consumers, likewise partition their markets.
Partitioning may offer ongoing benefits to marketplace participants. There may, for example, be
economies of scale from consolidating physical inventory. Likewise, by taking responsibility for selling
items, a marketplace may offer a better buyer experience: When partitioning, a marketplace can allow
buyers to compare products from different suppliers, present information about products consistently,
facilitate comparison-shopping, cross-sell and bundle complementary goods, and offer “one-stop”
shopping and delivery. 16 However, the benefits of partitioning must be balanced against the
incremental costs and risks of holding inventory—as well the possibility that a marketplace may not
be as effective at selling products as suppliers with deep product knowledge are.
15
Permissions
Marketplaces may have rules about permitted types of participation, permitted types of
transactions and/or transaction times, and protocols for completing exchanges.
Participation. Permission to participate in a marketplace may be limited in many ways, for

example, to paying members or parties with specified qualifications. For example, a regional real estate
association’s multiple listing service is only accessible to licensed brokers. Poppy requires its childcare
providers to have at least five years of relevant experience.
Some marketplaces grant exclusive rights to participants in certain categories. A grocery store
signing a lease as an anchor tenant in a strip mall, for example, might insist that no other food retailers
be allowed as tenants.
A crucial participation decision is whether the marketplace’s owner is permitted to trade as a

principal, in addition to organizing and operating the marketplace. In many settings, the marketplace
owner will have access to information on other participants’ transactions that could give the owner a
trading edge. (During the 1970s, for example, Salomon Brothers created the market for mortgage-
backed securities and was the leading trader in that market.)
Transaction Types. Types of transactions may also be limited. On eBay, for example, trade in
more than fifty items is prohibited or restricted (including firearms, gift cards, plants and seeds, slot
machines, stamps, tobacco, and charitable fundraising). Kickstarter has a similar list; among other
items, it prohibits offering genetically modified organisms as rewards.
Hours of Operation. Predictable hours of operation are important for coordinating

marketplace activity. Marketplaces may limit hours to ensure adequate liquidity when they are open.
A farmer’s market, for example, may be open once or twice a week; opening more frequently would
reduce the number of farmers willing to travel to the marketplace to present their wares at any given
time.
Exchange Protocols. Many marketplaces have detailed rules regarding processes for initiating
and completing transactions. For example: Airbnb hosts can pre-approve guests who inquire about
specific dates. A pre-approved guest has 24 hours to respond, but hosts can pre-approve multiple
potential guests for the same date; the first to respond gets the booking.
Profiling
Marketplaces often publish information about participants and their offers; this information can be
essential for helping potential exchange partners determine fit, and can help engender trust in the
marketplace by making new participants more comfortable joining.
Depending on participants’ requirements, information in profiles can be extensive or limited. In

commodity markets, for example, traders are often anonymous and information on their offers is
confined to the price and volume of their bids and asks. In dating marketplaces, by contrast,
participants provide profiles with detailed information—some might even say, too much
information—about their background, habits, and preferences.
In many marketplaces, information about participants and offers is structured in a consistent

format, which facilitates search and filtering. Marketplaces must decide whether to allow participants
to enter information directly, which can reduce costs but poses potential concerns about information
16
quality (e.g., confusing wording; low-resolution photos) and veracity. j In some cases, information may
be entered by participants but verified by the marketplace owner or certified by a credible third party.
Care.com, for example, allows either families or caregivers to pay for third-party reports on prospective
caregivers’ motor vehicle records and/or criminal histories. The Chinese dating site Jiayuan gives
participants the opportunity to submit copies of academic and other credentials; the site places a
“verified” marker on the profile of any user whose credentials it can authenticate.
Profiling may require the integration of information systems by the marketplace and participants, for
example, to view available inventory—a crucial capability when inventory is perishable, as with air travel.
Some marketplaces allow anonymous participation specifically because participants demand it. In
financial markets, for example, sophisticated buyers and sellers routinely divide large trades into
multiple small transactions to avoid alerting other traders to transactions big enough to move a
security’s price; without anonymity, such a strategy would be impossible.
In many marketplaces, profiles include reputation scores and other types of user reviews; such data
can be crucial as participants make decisions about potential exchange partners. In structuring
reputation systems, marketplaces confront several choices: 17
• Should reviewing rights be limited? To help ensure that reviews are accurate and useful, some
marketplaces limit reviewing to verified exchange partners; this can reduce noise introduced
by negative reviews posted by competitors, as well as excessively positive self-reviews or
postings by accomplices (“shills”) who pose as enthusiastic customers. But restricting
reviewing also makes it harder to build an initial review base.
• Should reviews be anonymous? Anonymity can encourage honesty if a transaction partner is

worried about a bad review provoking retaliation by a counterparty—especially if the marketplace
features reciprocal reviews. However, anonymous reviews can raise concerns about accountability
(and exacerbate the risk of false reviewing) unless appropriate safeguards are in place.
• How much information should be collected? Soliciting input on multiple dimensions of

satisfaction can provide information that is useful to participants, but asking for more
information will tend to depress participants’ willingness to submit reviews.
• Should reviews be reviewed? Amazon, Yelp, and many other marketplaces allow users to vote
on the usefulness of reviews. Vote results are used to give prominent placement to helpful
reviews—but such a system again provides opportunities for abuse.
• Should reviewing be incentivized? A common problem with reviewing systems is that they
disproportionately attract reviews from marketplace participants who have had extremely
positive or negative experiences and thus are motivated to share their views. Offering
incentives for reviews can ensure that reviews reflect a more complete distribution of user
experience. However, individuals who respond to incentives for reviewing may not have
preferences comparable to those of average marketplace participants. Additionally, if the
marketplace organizer offers incentives, it is often important to incentivize reviews equally—
independent of valence—to avoid skewing the review distribution directly. 18
j According to a study conducted by the dating site OkCupid, people on dating sites on average represent themselves as being
two inches taller than they actually are (see Rudder, “The Big Lies People Tell in Online Dating,” OkCupid Blog, Jul. 7, 2010,
accessed on Jan. 20, 2018 at https://theblog.okcupid.com/the-big-lies-people-tell-in-online-dating-a9e3990d6ae2/).
17
Processing
Marketplaces that facilitate transactions often assist exchange partners with transaction processing.
Processing assistance can take many forms, including transferring payment, holding payment in
escrow pending confirmation of a successful exchange, offering guarantees to one or both sides that
exchanges will be successful (or insurance that serves the same purpose), arranging for shipment,
completing government paperwork (e.g., registering a car with the Department of Motor Vehicles, or
filing customs forms for international shipments), or tracking shipment status.
Due to economies of scale, marketplaces often have a cost advantage over individual participants
in providing processing services, and can profitably pass savings along to participants. Providing
processing services can also attract participants to a market, by making exchange easier and by
reducing concerns about potential partners’ trustworthiness. When marketplaces charge extra for
transaction processing services, they can be a source of additional profit.
Pricing
Marketplaces earn revenue in different ways, which can broadly be categorized as follows:
• Transaction Fees. Many marketplaces charge flat fees per transaction, or fees that vary with
transaction size. eBay, for example, collects 8% of the first $50 in auction proceeds; 5% for
proceeds from $50 to $1,000; and 2% thereafter.
• Trading Spread. Some marketplaces profit by buying and reselling items. For example, market
makers are broker-dealers who, on behalf of a stock exchange, assume responsibility for—and
the risks associated with—holding a certain number of shares of a given security, to ensure
liquidity and orderly trading. Market makers earn a spread between the bid prices they pay to
sellers and the ask prices received from buyers. (Notably, the spread can be negative if the
market maker holds a large position after a decline in market prices.) Marketplaces that match
through partitioning also often earn trading spreads, as with Gazelle, Shift, and thredUP.
• Lead Generation Fees. Marketplaces that help partners discover each other but do not oversee
complete transactions sometimes charge a fee for each lead sent to a marketplace participant.
For example, Chegg amasses profiles of college-bound high school students, and then forwards
leads to colleges interested in marketing to students who meet specified criteria.
• Listing Fees. Marketplaces may charge fees for each item listed. eBay, for example, charges
listing fees that vary with the reserve prices set in its auctions; the fee is $1.00 for an item with
a reserve price between $50 and $200. eBay also offers “optional advanced listing upgrade fees,”
which allow sellers to enhance their listings with larger photos, extra text, and other features.
Listing fees typically reduce the number of items made available in a marketplace, but this can
have positive effects: in addition to providing a direct source of revenue, listing fees provide
suppliers with an added incentive to list only high-quality items that are likely to sell.
• Membership Fees. Participants in marketplaces sometimes pay membership fees that are not
explicitly linked to transaction volume. Membership fees may be levied upfront or on a
subscription basis; they might also vary with the quality and scale of marketplace presence that
participants desire (as with companies paying a premium for large, favorably located booth
space at a trade convention). It is important to note that membership fees may deter
participation in a new marketplace, due to uncertainty about whether the marketplace will
create enough value to warrant the investment.
18
• Processing Fees. As already remarked, marketplaces that assist with transaction processing
often charge ancillary fees for such services—for example, an online retailer charging a
premium for expedited shipping.
The listed revenue sources are not mutually exclusive; many marketplaces earn revenue in multiple
ways. As noted later (in the section “Building Liquidity”), the level and mix of fees—and which side of
the marketplace pays them—can have a significant impact on a marketplace’s ability to attract
participants and boost transaction volume.
Demands for revenue shape incentives in other ways. Consider the motivations of marketplaces
that help potential trading partners discover each other, but do not help them complete transactions.
Dating and recruiting sites, for example, are motivated to amass more subscriptions and job listings,
respectively—that is how they make money. They do not make more money if their users successfully
match; in fact, matching can lead to less revenue, as users who find partners or jobs end their
subscriptions and/or remove their listings. Likewise, dating and recruiting sites have incentives not to
prune stale profiles and job listings, respectively, as long as those stale postings contribute to revenue
(as they do, for example, if they make the marketplace appear thicker, and thus help keep current users
more engaged).
In sum: Marketplaces that earn revenue only from facilitating customer discovery will focus less on
delivering successful outcomes for participants than marketplaces that facilitate end-to-end
transactions and collect transaction fees. 19 But at the same time, transaction fees can create incentives
for marketplaces to complete transactions that may not be in participants’ best interests—for example,
encouraging participants to purchase an item when a substitute not available in the marketplace would
better fit the buyer’s preferences.
Promotion
Marketplace owners often undertake promotion efforts that make potential and existing
participants aware of types of transactions that the marketplace supports. Promotion can leverage
external channels or can occur within the marketplace itself.
External promotion—for example, paid marketing or public relations efforts—can be especially

important during a marketplace’s launch phase.
Internal promotion generally aims to inform existing participants of potentially attractive offers. It
can take the form of prominent placement or direct communications to marketplace participants. As
already noted, marketplaces may charge participants for internal promotion. But in offering paid
placements, marketplaces must weigh tradeoffs: Does the revenue upside justify the possibility of
perceiving promoted offers as spam or disingenuous?
Priming
Priming entails making something ready for use, as with a pump or a firearm. In the marketplace
context, priming involves actions through which a marketplace owner prepares items for exchange—
instead of relying solely upon the items’ suppliers to do so. By priming, the marketplace owner seeks
to boost items’ appeal and thereby increase trading gains for marketplace participants; the marketplace
owner aims to capture a share of trading gains that more than offsets additional costs incurred.
19
Priming helps ensure that items to be exchanged have consistent quality—which can be important
when supply-side participants might be inclined to ignore or hide their items’ shortcomings (as in the
“lemons” market for used cars). Priming can improve trust and reduce search costs for demand-side
participants. Shift, for example, takes physical possession of used cars and then reconditions them after
conducting a detailed inspection. The second-hand apparel reseller thredUP does the same. Poppy
conducts extensive background checks, reference checks, and interviews of caregiver candidates.
Employees of onefinestay visit homes and clean them to ensure they meet the company’s standards for
vacation rental (“onefinestay: Building a Luxury Experience in the Sharing Economy,” HBS No. 515-072).
Compared to a marketplace that follows a laissez-faire approach to managing supply, one that
follows a priming strategy—sometimes referred to as an “augmented” or “managed” marketplace 20—
typically requires suppliers to cede considerable control and incur upfront costs in anticipation of
greater trading gains in the future. Shift’s sellers, for example, forego use of their car while Shift holds
it in inventory. The peer-to-peer car rental company Turo (formerly known as RelayRides) initially
required suppliers to install electronic keyless entry systems to allow renters to access their cars.
IguanaFix, which matches home improvement contractors to homeowners in Latin America, completes
background checks and interviews contractors, trains them to follow IguanaFix’s service delivery
standards, prohibits them from charging more than originally quoted, and closely monitors contractors
during a probation period if they receive a customer rating of 3 or lower on a 5-point satisfaction scale
(“IguanaFix,” HBS No. 817-056). If too few suppliers are willing to cede control and incur costs, a
priming strategy can backfire.
Priming activities overlap with some of the other “Ps” already described. In particular, partitioned
pairing, through which participants transact with the marketplace owner rather than directly with each
other, is an important element of many priming strategies. For example, partitioning by Shift not only
facilitates reconditioning, but also allows Shift sellers and buyers to avoid the nuisance of dealing with
each other to schedule test-drives and negotiate prices.
Likewise, when priming, the marketplace owner often takes responsibility for profiling. By doing so,
the marketplace owner ensures quality and consistency of presentation. For example, Dating Ring’s
matchmakers interview clients, carefully curate photos they submit, and then draft profiles that
enhance clients’ appeal. Similarly, after reconditioning used goods they resell, Shift and thredUP
photograph the goods and create marketplace listings. Auction houses like Christie’s and Sotheby’s
invest considerable resources in publishing glossy, informative catalogs that showcase items in
upcoming auctions: a priming activity that bridges both profiling and promotion.
Launching Marketplaces
Launching a new marketplace poses special challenges. Foremost, the marketplace must convince
potential participants that they can realize greater value by switching away from other exchange
opportunities or by adding a new marketplace to their stable. Potential participants will be concerned
with whether the new marketplace will attract enough other participants to deliver value—that is,
whether liquidity will be adequate. Liquidity, in turn will hinge on many factors—in particular,
whether the new marketplace and its participants are deemed to be trustworthy.
20
Building Liquidity
Liquidity’s Impact. Liquidity creates value in a marketplace by reducing participants’ costs of
identifying transaction partners. With a large number of potential partners, it is easier to find one with
an acceptable offer. Increased liquidity can also impact average prices realized in a marketplace if the
ratio of supply to demand changes. Finally, greater liquidity should tend to reduce price volatility,
because potential trading partners will be able to evaluate a larger set of offers and review prior
transaction data before they commit to an exchange. Hence, to the extent that risk averse participants
value price predictability, they should value increased liquidity.
Barriers to Adoption. The major barriers to potential participants embracing a new

marketplace are listed below. 21 Note that some of the barriers correspond to specific causes of market
failure.
• Setup Costs. If affiliating with a new marketplace requires significant upfront investments—
for example, system integration or credential certification—then potential participants may
choose to “wait and see” if the marketplace will gain scale before they join.
• Expectations. New marketplaces confront a Catch-22: participants on one side will not join unless they
expect to be able to access a critical mass of potential partners on the other side—and vice versa.
• Trust. The concerns described so far can be amplified by a lack of trust: Potential participants
may have concerns about the quality of offers available in the new marketplace, asking, “How
can we be sure counterparties will deliver what we expect, when we have never traded with
them before?” Likewise, they may doubt whether the new marketplace itself can meet its
promises—or, if it is a startup—even survive, asking, “Can we entrust this unproven entity with
mission-critical transactions? With our data? What will happen if this startup fails?”
• Dominance. There are also scenarios where participants avoid a proposed marketplace because
they fear that its owner may be too successful; this may happen when the owner has significant
resources and a track record of dominating other markets. For example, Apple has had
difficulty extending its Apple Store franchise for downloading and streaming media to include
a subscription television offering, in part because cable network programmers refused to
participate, having witnessed Apple’s ability to dictate terms to music companies in the iTunes
Store years earlier.
• Business Stealing. Given uncertainty about the supply/demand balance that will prevail in a new
marketplace, participants may worry about pricing pressure if too many direct rivals also participate.
• Favoritism. Potential participants may fear that marketplace owners could skew the terms of
exchange in one side’s favor. For example, rivals sometimes cooperate to create a new
marketplace, aiming to strengthen their position in trading with traditional partners (as with
Covisint in online auto parts procurement and Orbitz in online travel).
Tactics for Building Liquidity. Owners of new marketplaces can pursue a range of strategies
for building liquidity. Aggressive promotion is an option, if the marketplace’s owner has resources to
fund marketing. Partitioning can offer a different, more market-driven solution: the marketplace can
create liquidity (or smooth intertemporal liquidity shortfalls) by buying items itself and then reselling
them.
21
Other tactics listed here are described in more detail in the technical note “Mobilizing Networked
Businesses” (Coles & Edelman, HBS No. 911-048). Tactics for building participant trust in new
marketplaces are summarized later.
• Segment Focus. New marketplaces can aim to deliver high value to participants in a narrow
range of market segments, with the possible goal of broadening the range over time.
Segmenting allows a marketplace to focus early product development and marketing on the
unique needs of segment members, thereby accelerating participant acquisition. However,
when a marketplace is tailored to the specific needs of a segment, it can prove difficult to attract
new segments without diluting value for all participants and muddling marketing messages.
• Marquee Participants. Exclusive contracts with marquee participants—attractive trading partners

for other potential participants—can help build marketplace liquidity. For example, when Priceline
was building its “name-your-own-price” marketplace for air travel, it signed up Delta Airlines as a
charter partner, giving Delta an equity stake. 22 Likewise, shopping mall owners routinely seek a
department store as an anchor tenant, often discounting the store’s rent as an incentive.
• Exclusivity. As already noted, potential participants may be reluctant to join new marketplaces
if they anticipate serious problems with business stealing. One strategy for addressing this
concern is to deliberately exclude all but a single participant from one side of the marketplace,
then charge that sole participant high fees for the privilege of trading with the other side. (The
online car buying service Carpoint, which forwarded consumers’ queries to a single dealer in
any given geographic territory, employed this strategy (“Carpoint in 1999,” HBS No. 800-328).)
Of course, the marketplace must ensure that participants granted exclusivity do not abuse their
monopoly positions; otherwise, participants on the other side will avoid the marketplace.
• Standalone Value. A marketplace that needs to solve a Catch-22 before it can leverage network
effects can attract participants on one side by providing sources of value that are independent
of network size. Transaction cost savings are one possible candidate: Parties can initiate
transactions off of the new marketplace—perhaps with past transaction partners—and then
bring their exchanges to the marketplace for processing. By leveraging economies of scale and
passing along savings, the new marketplace can start to build a participant base. Over time, the
marketplace may attract enough participants to host earlier phases of exchange, as well.
Rentjuice, for example, tapped a source of standalone value when mobilizing a new
marketplace for apartment rentals: It initially built a suite of SaaS tools for rental real estate
brokers. Once a critical mass of brokers was using the tools, Rentjuice leveraged those
relationships to begin to amass apartment listings, which in turn attracted potential renters
(“Rentjuice,” HBS No. 811-069).
• Penetration Pricing. Sponsors of a new marketplace can build liquidity by subsidizing early
adopters. Giving early adopters free (or deeply discounted) access essentially compensates
them for the risks and costs they incur in supporting the marketplace before its viability is
assured. Once the marketplace gains momentum, the subsidy is removed.
• Permanent Subsidies. Alternatively, a marketplace owner may choose to permanently subsidize

participants on one side of the market—for convenience, “Side A”. The subsidy attracts more
participants to Side A, which boosts Side B participants’ willingness to pay for marketplace
affiliation, due to network effects. The marketplace owner exploits this by charging a premium to
B that more than offsets subsidies to A. Generally, it makes sense to subsidize the marketplace’s
more price sensitive side and to charge the side that more strongly increases demand in response
to the other side’s growth, due to a preference for transaction partner variety. 23
22
• “Pay-As-You-Go” Pricing. Marketplaces can encourage participation by reducing startup costs

for new participants; one way to do so is to avoid upfront, fixed fees for marketplace affiliation
and instead rely more heavily upon transaction fees. With “pay-as-you-go” pricing, participants
incur costs only when they successfully transact.
• Synchronicity. With a limited number of potential trading partners available at the outset, a new
marketplace will be more valuable if participants are active in the marketplace at the same time.
Hours of operation can be limited to help ensure a critical mass of participants is available when the
marketplace is open. Alternatively, the marketplace can provide tools to facilitate asynchronous
exchanges that unfold over time, with clear communication regarding transaction status.
Tactics for Building Trust. In parallel with efforts to build liquidity directly, new marketplaces
can pursue different tactics for building the trust of potential participants, including:
• offering guarantees that participants will be compensated if exchanges are not completed to their
satisfaction (or insurance that serves the same purpose);
• putting payment in escrow pending confirmation of successful transactions;
• providing processes for resolving disputes between exchange partners;
• certifying participants or their offers;
• publishing reputation-based ratings and reviews;
• publishing testimonials from satisfied participants;
• offering other “social proof,” such as the involvement of marquee participants; and
• pursuing priming strategies to ensure the quality and consistency of offers.
Maintaining Marketplaces
After marketplaces achieve critical mass, their owners may face challenges in managing them in
ways that deliver adequate ongoing value for marketplace participants.
Balancing Supply and Demand. A marketplace must keep the ratio of active supply-side to
demand-side participants within a target range. If the ratio drifts too high or low, participants on the
overcrowded side may, due to excessive rivalry, derive insufficient value to continue using the
marketplace. Marketplace owners can keep supply and demand in balance through price changes and
promotional efforts.
Congestion. A successful marketplace may attract so many participants that it becomes

congested —especially if the underlying market is itself congested. Solutions for congestion include:
(1) raising prices to drive off low quality participants; and (2) refining pairing mechanisms to make it
easier to identify good matches. A related risk in congested markets is unraveling—and again, refining
pairing mechanisms is a potential solution.
Disintermediation. If multiple rounds of direct communication are required between buyers

and suppliers (e.g., over requirements, delivery dates, or price terms), then participants may try to
complete their exchanges away from the marketplace in order to avoid fees. Likewise, for marketplaces
that connect trading partners who will engage in recurring transactions (e.g., marketplaces for house-
23
cleaning), there is a risk that once connected, transaction partners will conduct future exchanges
outside the marketplace.
Tactics for managing disintermediation are described in the technical note “Disintermediation in
Two-Sided Marketplaces” (Edelman & Hu, HBS No. 917-004). In brief, they include:
• Detecting patterns that suggest disintermediation has occurred (e.g., noticing that parties who
regularly transacted at a specific time are no longer doing so), then threatening deactivation.
• Withholding partners’ contact information and requiring participants to use marketplace

messaging channels to communicate (and then monitoring those channels to make sure
participants do not share contact information).
• Offering hard-to-replace features (e.g., project tracking tools used by clients of Upwork freelancers)
that are highly valued by at least one side.
• Providing volume-based benefits (e.g., a loyalty program).
• Simply conceding that disintermediation is likely and collecting more revenue upfront before parties
transact—while recognizing that higher upfront fees may deter marketplace participation.
Marketplace Dominance. Some marketplaces enjoy winner-take-all structural attributes,

including strong network effects and high multi-homing costs (for analysis of these attributes, see
“Winner-Take-All in Networked Markets,” HBS No. 806-131). When a marketplace achieves a
dominant share, it must contend with risks of regulation and antitrust intervention.
Closing Note
Markets frequently fail due to high transaction costs, perverse incentives, and other factors.
Whenever markets fail, potential entrepreneurial marketplace design opportunities arise.
The Internet’s rapid growth in the mid-1990s spawned a flurry of new marketplaces, as did the
subsequent spread of social networks and smartphones. In coming years, technological change—for
example, the adoption of augmented/virtual reality and blockchain technologies—seems certain to
create yet more opportunities for marketplace entrepreneurs.
As entrepreneurs pursue marketplace design, they should remember that details matter—or more
colloquially, “The Devil is in the details.” As this note demonstrates, markets are complex, and the causes
of their failure vary. Shifts in incentives that address one problem can sometimes create others. Given this
complexity and the plethora of choices available when designing pairing mechanisms, profiling systems,
pricing approaches, and other principal components of marketplaces, it is crucial for entrepreneurs to
truly understand markets they aim to serve, and the solutions and mechanisms they intend to use.
24
Endnotes
1 Substitutes exhibit positive cross-price elasticity. Cross-price elasticity measures the change in volume sold of Product A in
response to a change in the price of Product B, holding all else constant—in particular, A’s price. When products are
substitutes, if B’s price is increased without changing A’s price, customers buy more A. However, defining markets as sets of
transactions involving goods with positive price elasticity—the criterion often applied in antitrust proceedings—excludes
markets where demand and supply forces operate but explicit prices are not a deciding factor—for example, in most dating
markets or public school choice programs.
2 For discussion of strategies for (and pitfalls of) sequential search, see Christian & Griffiths, Algorithms to Live By: The Computer
Science of Human Decisions, Holt, 2016, ch. 1.
3 Coase, 1960, “The Problem of Social Cost,” Journal of Law and Economics, 3.
4 Ellison & Ellison, 2009, “Search, Obfuscation and Price Elasticities on the Internet,” Econometrica, 77:2.
5 Kagel & Levin, Common Value Auctions and the Winner’s Curse, Princeton University Press, 2002.
6 Akerlof, 1970, “The Market for Lemons: Quality Uncertainty and the Market Mechanism,” Quarterly Journal of Economics, 84:3.
7 Roth, Who Gets What, and Why?, Eamon Dolan, 2015.
8 Kominers & Weyl, 2012, “Holdout in the Assembly of Complements: A Problem for Market Design,” American Economic
Review Papers & Proceedings, 102:3.
9 This section is adapted from “Platform-Mediated Networks: Definitions and Core Concepts,” Eisenmann, HBS No. 807-049.
10 Farrell & Saloner, “The economics of horses, penguins, and lemmings,” in Gable, L. ed., Production Standardization and
Competitive Strategies, North-Holland, 1987. Farrell and Saloner showed that fragmented demand might also lead to excess
momentum, which they labeled a “lemming problem.” If I expect most other users to quickly join a new marketplace that seeks
to displace an old one, I may prematurely abandon the old network and in doing so incur unwelcome switching costs—even
though I’d be happier if a critical mass of users stuck with the old marketplace. Unbeknownst to me, due to our collective
inability to communicate and coordinate behavior, many others may hold the same preferences and expectations.
11 Segal & Whinston, 2007, “Antitrust in Innovative Industries,” American Economic Review, 97:5; Raskovich & Miller, 2011,
“Cumulative Innovation and Competition Policy,” U.S. Department of Justice Working Paper.
12 Roth & Ockenfels, 2007, “Last-Minute Bidding and the Rules for Ending Second-Price Auctions: Evidence from eBay and
Amazon Auctions on the Internet,” American Economic Review, 92:4.
13 Riordon, Plunkitt of Tammany Hall: A Series of Very Plain Talks on Very Practical Politics, Signet Classics, 1995, pp. 4-5;
excerpted in Kominers, “Plunkitt: Won / Auctioneers: $2.50,” Journal of Political Economy, 120:1.
14 Klemperer, Auctions: Theory and Practice, Princeton University Press, 2004, p. 105.
15 Gale & Shapley, 1962, “College Admissions and the Stability of Marriage,” American Mathematical Monthly, 69:1; see also Roth, “The
Evolution of the Labor Market for Medical Interns and Residents: A Case Study in Game Theory,” Journal of Political Economy, 92:6.
16 Hagiu & Wright, “Do You Really Want to be an eBay?” Harvard Business Review, 91:3.
17 For an overview of issues with review and rating systems, see Luca, 2016, “Designing Online Marketplaces: Trust and
Reputation Mechanisms,” Harvard Business School Working Paper 17-017. This section also draws upon Coles & Edelman,
“Market Design in Online Businesses (Abridged)” HBS No. 915-016.
18 Li, Tadelis & Zhou, “Buying Reputation as a Signal of Quality: Evidence from an Online Marketplace,” 2016, NBER Working
Paper No. 22584; see also Kominers, “For Better Online Product Reviews, Pay the Reviewers,” Bloomberg View, Sep. 21, 2016,
accessed on Jan 25, 2018 at https://www.bloomberg.com/view/articles/2016-09-21/for-better-online-product-reviews-pay-
the-reviewers.
19 Levine & Syed, “The Emergence of the End-to-End Marketplace,” Forbes, Dec. 14, 2014, accessed on Jan. 6, 2018 at
https://www.forbes.com/sites/valleyvoices/2014/12/15/the-emergence-of-the-end-to-end-marketplace/#3913b98e7af9.
20 Galston, “Anatomy of a Managed Marketplace,” TechCrunch, May 25, 2017, accessed on Jan. 6, 2018 at
https://techcrunch.com/2017/05/25/anatomy-of-a-managed-marketplace/
21 This section is adapted from Eisenmann, “Managing Networked Businesses: Network Mobilization Module,” HBS No. 808-079.
22 Martha Brannigan and Carrick Mollenkamp, “Delta's Stake in Priceline.com Presents Some New Challenges,” Wall Street
Journal, June 14, 1999, accessed on Jan 11, 2019 at https://www.wsj.com/articles/SB929311017970408400
23 Eisenmann, Parker, & Van Alstyne, 2006, “Strategies for Two-sided Markets,” Harvard Business Review, 84:10; see also Weyl,
2010, “A Price Theory of Multi-Sided Platforms,” American Economic Review, 100:4.
25

Making Markets

Uploaded by

Copyright:

Available Formats

Making Markets

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Making Markets

Uploaded by

Copyright:

Available Formats

What are the main functions of a marketplace?

What are the main functions of a marketplace?

What attributes determine whether marketplaces create value?

What attributes determine whether marketplaces create value?

9 -8 1 8 -0 9 6

REV: FEBRUARY 28, 2018

SCOTT DUKE KOMINERS

Markets and Marketplaces

What is a Marketplace? A marketplace gathers many potential transaction partners into a

Heterogeneity. Some markets feature substantial heterogeneity in buyers’ requirements and/or

Information Asymmetry. In markets, information asymmetry—an uneven distribution of

Source: Casrewriter analysis.

• Profiling encompasses information presented by the marketplace about participants (e.g.,

• Processing entails activities undertaken by the marketplace to facilitate transactions, for

• Promotion makes potential participants aware of the advantages of trading in a marketplace

Some marketplaces employ market-clearing algorithms to simultaneously match pools of participants.

The deferred acceptance algorithm works as follows:

Proposed Pairing. Many marketplaces propose potential transaction partners—based on

Passive Pairing. Some marketplaces provide no specific recommendations regarding potential

Participation. Permission to participate in a marketplace may be limited in many ways, for

A crucial participation decision is whether the marketplace’s owner is permitted to trade as a

Hours of Operation. Predictable hours of operation are important for coordinating

Depending on participants’ requirements, information in profiles can be extensive or limited. In

In many marketplaces, information about participants and offers is structured in a consistent

• Should reviews be anonymous? Anonymity can encourage honesty if a transaction partner is

• How much information should be collected? Soliciting input on multiple dimensions of

External promotion—for example, paid marketing or public relations efforts—can be especially

Barriers to Adoption. The major barriers to potential participants embracing a new

• Marquee Participants. Exclusive contracts with marquee participants—attractive trading partners

• Permanent Subsidies. Alternatively, a marketplace owner may choose to permanently subsidize

• “Pay-As-You-Go” Pricing. Marketplaces can encourage participation by reducing startup costs

• putting payment in escrow pending confirmation of successful transactions;

• providing processes for resolving disputes between exchange partners;

• certifying participants or their offers;

• publishing reputation-based ratings and reviews;

• publishing testimonials from satisfied participants;

• pursuing priming strategies to ensure the quality and consistency of offers.

Congestion. A successful marketplace may attract so many participants that it becomes

Disintermediation. If multiple rounds of direct communication are required between buyers

• Withholding partners’ contact information and requiring participants to use marketplace

• Providing volume-based benefits (e.g., a loyalty program).

Marketplace Dominance. Some marketplaces enjoy winner-take-all structural attributes,

7 Roth, Who Gets What, and Why?, Eamon Dolan, 2015.

You might also like