Marketplaces

By Rohan Mahajan

Numerous marketplaces including Uber, Airbnb, and Ebay have had amazing success. Marketplaces are special instances of networks with two classes of nodes: supply and demand. Think drivers and riders for Uber or hosts and guests for Airbnb. Marketplaces can have differentiated and undifferentiated supply. In marketplaces with undifferentiated supply, demand has very few unique offerings they can choose from such as in Uber. In differentiated supply, demand has numerous unique offerings they can choose from such as in Airbnb. Most network principles (see Network Effects) still apply to marketplaces but marketplaces have unique properties. This article provides an introduction to marketplaces. It discusses the network effects of marketplaces, what value the marketplace business itself provides, marketplaces versus traditional businesses, monetization, jumpstarting marketplaces, desirable marketplace features, and finally the defensibility of marketplaces.

Network effects for marketplaces

With traditional network effects, an additional node provides additional value to the rest of the nodes. However, marketplaces have two distinct classes of nodes. Same-side network effects are how a node affects nodes of the same type while cross-side network effects are how a node affects nodes of another type. Demand same-side effects can either be positive or negative. Uber has slightly negative demand same-side effects because the direct effect of an Uber rider to a fellow Uber rider is more competition for a ride. Finite amounts of supply increase demand competition and the probability of negative demand same-side effects. Marketplaces with software or other virtual goods have infinite supply and thus don’t have negative demand same-side effects. In fact, Groupon and Kickstarter have positive demand same-side effects because more demand increases the likelihood that the campaign succeeds. Supply same-side effects are generally negative as increased supply increases competition and suppliers reately interact in a synergistic manner. However, certain marketplaces called marketplace networks allow suppliers to form connections with other members of supply. These suppliers work synergistically to fulfill the needs of the consumer. Honeybook, a marketplace for events, allows photographers, florists, caterers, and other event professionals to form connections and team up to offer successful events to consumers.

Cross-side network effects are almost always positive and bidirectional. More demand is better for supply and more supply is better for demand. These cross-side network effects help aggregate supply and demand and outweigh possible negative same-side network effects. As a supplier having another supplier onboard might slightly hurt your chances of winning one specific transaction, but more consumers will be willing to use the service. Winning 1% of a billion transactions is better than winning 50% of 100 transactions. Although cross-side effects are both positive, supply has a bigger cross-side effect on demand than demand has on supply. Generally, supply has dramatically fewer units than demand with each unit of supply participating in significantly more transactions than an corresponding unit of demand. Uber has significantly more drivers than passengers. Furthermore, an Uber driver participates in significantly more rides than an Uber rider.

Value marketplaces themselves add to their community

First, marketplaces aggregate supply and demand, allowing each consumer access to a huge amount of supply and each supplier access to a huge amount of demand. Second, marketplaces add trust through mechanisms such as reviews, identity verification/background checks, or insurance. Having a clear and transparent governance policy along with ensuring marketplace members write reviews is imperative for establishing trust. Trust required for supply and demand can be different. For instance, Uber requires much more stringent background checks for drivers than riders. Ensuring trust can increase friction and a marketplace must carefully navigate this trust-friction relationship and adjust it as the marketplace evolves. Trust is hard to build but can easily be broken. As trust increases, the number of transactions will increase.

Third, marketplaces can provide recommendations and help match demand and supply; for instance, Uber picks the closest driver for riders and Airbnb suggests certain rooms for guests. On a side note, marketplaces decrease the power of supply brands. Brands inspire trust and have good distribution capabilities. Marketplace recommendations allow consumers to learn about new unbranded products. Marketplace trust mechanisms such as reviews inspire user trust in these new unbranded products. Consumers talk about the marketplace they bought the product and not the product brand itself. In a certain perspective, the marketplace becomes the brand. Fourth, managed marketplaces (full-stack marketplaces) can help with conducting and fulfilling the the transaction: i.e. Doordash handles payments and also does delivery from the restaurant to the consumer’s door.

What value does being a marketplace provide over a traditional business?

First, marketplaces require significantly lower capital as suppliers generally provide most of the capital. For instance, Uber places the onus on drivers to provide the car. Second, marketplaces have to hire and manage fewer people. Suppliers also feel enhanced ownership compared to a normal employee. These features increase margins and allow the marketplace to grow rapidly and along with network effects, increase the probability the marketplace will be a monopoly. Third, organizations are good at creating standardized behavior but not differentiated and creative behavior. Marketplaces can crowdsource ideas and variations for supply. Airbnb itself would not have been able to create the millions of interesting and unique listings. Fourth, especially with the on-demand economy, marketplaces can easily adjust supply and demand. For instance, after a major event, Uber can offer surge pricing and temporarily scale up the number of drivers.

Monetization

Marketplace participants or the marketplace itself can set prices. With undifferentiated supply, the price should be the same and so logically the marketplace should always set it. In differentiated supply, the price differs and thus some marketplaces might let suppliers set their price: i.e. airbnb. Furthermore, suppliers can further extract value through innovative mechanisms such as auctions in Ebay. These type of features add value and potentially more income to suppliers at the expense of increased friction on the demand side.

Regardless of who sets the price, the platform needs to monetize. As written in Network Effects, marketplaces can charge for advertisements, access for third parties, and for enhanced privileges. However for marketplaces, the easiest and most frequent monetization strategy is transaction fees. This transaction fee can be taken from the demand side, the supply side, or both. Factors such as price-sensitivity, magnitude of cross-side effect, and difficulty of acquisition impact how to price the different sides. Involvement in the payment flow increases the ease of taking this transaction fee. If the majority of the transaction happened offline and the platform has no way to verify what happened in the transaction, then a transaction fee may not be feasible. For instance, the platform Thumbtack matches consumers with local professionals such as plumbers. Thumbtack does not know what the plumber exactly did. Therefore, Thumbtack charges local professionals a fee to be listed on the market. Listing fees encourage supply to ensure that their listings are high quality but may deter supply from joining the platform because they don’t want to pay an upfront cost with no guarantee of a transaction.

How high of a fee should a marketplace charge? Raising fees increases the revenue and margin of the company. However, this high transaction fee may deter users to join and decrease growth. If two marketplaces are able to aggregate demand evenly, supply will choose the marketplace which takes less of a cut. At the same time, this extra revenue and better financial performance (possibly leading to more capital) may allow for the company to invest in distribution tactics to increase growth. Because of network effects, growth and capturing the market is imperative.

Certain marketplaces, especially those with supply being businesses, have two monetization models: the merchant model and the agency model. In the merchant model, the platform pre-buys large quantities of supply directly from the merchant and then sells this supply itself on the marketplace. In the agency model, these businesses are able to list their supply on the marketplace directly. Generally, the marketplace charges a higher fee in the merchant model than in the agency model. However, the marketplace must negotiate with merchants directly and supply growth is much slower, especially if there is a long tail of supply. As mentioned before, sacrificing monetization opportunity for growth is often necessary in marketplace businesses. For instance, Expedia dominated the market with a merchant model where they bought blocks of hotel rooms and sold them on their platform. Booking.com used an agency model where hotels could list rooms directly and quickly gained market share. Expedia eventually relented and offered the agency model.

Good and bad characteristics for a marketplace

Certain characteristics are desirable for a marketplace. Higher transaction values mean the marketplace can charge more. Higher frequency of transactions not only means more frequent revenue but also helps create a habit and thus increases retention. More frequent transactions increase opportunities for word of mouth growth and viral loops. Marketplaces with infrequent transactions might lose the customer and have to regain them. These marketplaces can acknowledge this and focus more on SEO to keep reacquiring the users. Marketplaces can try to keep engaging users even if they don’t transact frequently through mechanisms such as outbound email campaigns or making it fun/interesting to browse supply. Supply fragmentation is desirable. Big suppliers may think they can aggregate demand without paying fees to the marketplace.

Differentiated supply is generally beneficial as it is harder for a competitor to copy. The marketplace asymptotes later as each unit of supply makes a materialistic difference. However, marketplaces should only differentiate supply if demand cares about the differentiation and should not unnecessarily differentiate supply as it creates unnecessary complexity for the demand side. Imagine if Uber had different products for each make of car. Riders would have to use a super complicated user interface. Matching drivers with riders would become super difficult and wait times would dramatically increase. Irregular unscheduled on-demand usage helps because it decreases the probability of hiring a dedicated resource and really requires aggregated supply. Additionally, the combination of differentiated supply with regularity means the demand side will probably use the same unit of supply and have no need for the platform.

Transactions where users can not afford to make a mistake are more difficult for marketplaces. The demand side will spend significant time vetting a dedicated resource. Because of the high severity of a failed interaction, the demand side needs extremely high trust but will be reluctant to participate in transactions that help build trust. For instances, most parents would be more hesitant to use an on-demand child care marketplace than an Uber. Managed marketplaces attempt to solve this problem by offering extensive services to increase trust. Trusted, a marketplace for child care, performs detailed background checks/ interviews potential providers and also provides extensive tools to monitor the child-care providers.

Being a global marketplace is more desirable than being a local marketplace. Uber is a local marketplace as passengers only care about drivers in their city. However, Airbnb is a global marketplace as guests can stay at places all across the world. Similar to the clustering coefficient of traditional networks, local marketplaces are harder to grow as establishing a local marketplace does not help establish the next one. Furthermore, local marketplaces are less defensible as competitors can attack each local marketplace at a time.

Jumpstarting marketplaces

Jumpstarting marketplaces is a unique challenge because it requires acquiring both demand and supply. As mentioned in Network Effects, focusing on a smaller market and encouraging creators(i.e. supply) to bring in demand is always a good strategy. Additionally, marketplaces can add independent value to one side, so that the marketplace can aggregate it without the other side. Once that side is aggregated, the marketplace can focus on getting the other side. For instance, the iPhone added value to the initial consumers even though there were no other apps. Developers saw so many consumers using the iPhone and decided to build apps. Opentable built tools for restaurants to internally manage their open tables. After aggregating all these restaurants’ open tables, Opentable allowed the demand side to make reservations online.

Furthermore, marketplaces can artificially subsidize sides of the market to incentivize them to join the network. In new markets, Uber gives drivers bonuses to encourage them to drive. Marketplaces can also artificially provide supply. For instance, Ebay itself could buy some items from some other stores and then sell them on Ebay. Moreover, marketplaces can use the multihoming strategy where they entice the supply side with a low risk offering that has no upfront costs and limited opportunity costs and still allows it to participate in its previous economic arrangements. Doordash and Uber Eats allowed restaurants to operate as normal and use their additional kitchen capacity to potentially earn extra revenue. Finally, marketplaces can convince people on one side to also perform roles on the other side. Ebay actively encouraged buyers of used goods to also sell their used goods.

Defensibility of Marketplaces

Craigslist seemingly hasn’t been updated in a decade but still has 90 million unique monthly visitors due to its aggregated supply and demand. In addition to the aggregated supply and demand, people can be reluctant to switch marketplaces due to reputation earned and preferences learned by the marketplace. However, competitive marketplaces can use almost any technique in the jumpstarting marketplaces section such as multihoming. Marketplaces such as Doordash sign supply to exclusive contracts to prevent multihoming. Numerous marketplaces are attacking specific vertical subsets of existing horizontal marketplaces such as Craigslist. Airbnb caters specifically to housing. Eventbrite focuses on tickets. Lyft focuses on ridesharing. These marketplaces build tools and design their communities specifically for their vertical.

One interesting phenomenon is the competition between different services marketplace for undifferentiated supply. For instance, Lyft competes with Uber but also Doordash, despite demand wanting entirely different items. Service workers have a fixed amount of hours they can work and can’t physically perform multiple tasks at the same time. These service workers care about total revenue, which equals pay per hour * total number of hours. Paying more per hour definitely gives a marketplace an advantage. Offering more hours of work is helpful for a marketplace, but with multihoming, users could possibly use an app that has high hourly pay but doesn’t offer that many hours of work and then another app for the rest of the hours. It would be interesting to quantitatively understand how willing supply is to multihome to maximize revenue, how loyal they are to marketplaces that offer less revenue, the marginal value of money once they hit their earning target, and how quickly they hear and comprehend that another marketplace is more lucrative. Why aren’t all drivers on the same platform? (Any comments here would be helpful.)

Conclusion

Marketplaces can be analyzed systematically. Marketplaces can have negative or positive demand same-side effects but have negative supply same-side effects and positive cross-side effects. Marketplace business themselves aggregate demand and supply, add trust, help match supply and demand, and can help conduct transactions. Compared to normal businesses, marketplaces have less capital costs, can grow quicker, can have more differentiated and creative supply, and can more scalably adjust supply and demand. Marketplaces can monetize through transaction fees, ads, charging for third-party access, and enhanced user privileges. Higher transaction fees, more frequent transactions, irregular transactions, supply fragmentation, and being global are all desirable features for a marketplace. Marketplaces can be jump started through focusing on small clusters, virality, providing value to one side, subsidies, artificially creating supply, multihoming, and convincing people to be both supply and demand. Marketplaces such as Craigslist have succeeded due to network effects, but they can still be disrupted.

Sources

Almost all of the ideas/ content were found in the following people’s blogs/writing/twitter/videos. This post is more of a summary/aggregation of all these posts than an original post.

Sangeet Paul Choudary 
Nx effects blog
Andrew Chen
Casey Winters
Bill Gurley
Jeff Jordan
Brian Balfour
Li Jin
Bros Wertz
Angela Tran Kingenys

Acknowledgements

I would like to thank Li Jin, Deepak Narayanan, James Thomas, Andrew Liang, Karthik Jagadeesh, and Varun Mohan for feedback.