Platform leakage I

What factors make leakage worse, and what marketplaces can do about it

When we look at investing in platform startups, particularly marketplaces, often a key consideration is the extent of the leakage problem they face, and how they may deal with it. Leakage happens when participants meet on the platform but take their transactions off the platform to avoid paying the platform its cut. A typical example would be a buyer and a seller discovering each other on eBay, but then transacting directly in order to avoid eBay’s transaction fees. This is also oftentimes called disintermediation in the management literature, or showrooming in the economics literature.

Every marketplace that would like to charge some amount for transactions it enables faces an underlying leakage problem. However, the extent of the problem varies widely across marketplaces (from negligible to existential risk), depending on the nature of the transactions being enabled. After discussing the factors that determine the severity of the leakage issue, we lay out the various solutions that marketplaces have to deal with it.

First, to identify the exogenous factors that determine the extent of the leakage problem, we apply a series of questions:

1. What is the value of a typical transaction?

In the absence of leakage concerns, marketplaces typically find it optimal to charge higher fees (in dollar amount) for higher-value transactions. This means that leakage will inherently be more of an issue for marketplaces that facilitate higher value transactions. For instance, leakage is a serious issue for recruiting marketplaces that try to charge fees for successful matches (e.g. Glints), where the value of the resulting employment contracts can be hundreds of thousands of dollars, and the corresponding fees in the tens of thousands of dollars. This provides a big incentive for the parties to deal directly after getting matched. On the other hand, redbubble, a marketplace for gifts designed and sold by artists, typically deals in $1-$50 value items, and the fees it extracts are an order of magnitude less, so the underlying leakage problem is less severe.

2. Are marketplace users mainly seeking to conduct repeat transactions with the same counterparties or do they mainly discover and conduct one-off transactions with many different counterparties?

Repeated transactions between the same counterparties not only increase the amount of transaction fees the parties can save by completing transactions off the platform, but they also make it possible to get to know one another, build trust, and so be more willing to transact directly. Marketplaces for babysitters (e.g., dog walkers (e.g. Wag!), coaches (e.g. CoachUp), tutors (e.g. Preply), home cleaners (e.g. Handy), and gardeners (e.g. Lawn Love) all grapple with this problem, since most consumers prefer to work with a regular service provider once they find a good match. This is very different from marketplaces for on-demand services like Uber and Lyft, where consumers face a different service provider every time, and given the dollar amounts involved, it rarely makes sense to try to get around the platform fees.

3. To what extent can the typical transaction be well-defined and contracted upon without the parties meeting and exchanging information?

For some services, such as fixing a car or resolving electrical or plumbing issues, the parties have to meet and communicate before they settle on what the transaction will involve (e.g. inspect a problem and give a quote for the work required). In such cases, leakage becomes a lot easier as the parties share information, gain trust and can make payments directly before the platform is ever in a position to try to charge a fee for the transaction. Indeed, the platform may never know if a successful transaction even took place. Marketplaces for hiring (e.g. ZipRecruiter), home repairs (e.g. Thumbtack), and dating (e.g. Bumble and Tinder) inherently suffer from this problem since the two parties need to meet each other to decide if they are a good match. This is why none of the marketplaces mentioned in the previous sentence charges transaction fees: ZipRecruiter charges subscription fees to employers, all dating apps typically charge subscription fees, and Thumbtack charges referral fees to service providers. On the other hand, even if the parties eventually meet in person, such as in the case of Airbnb, if the transaction can be determined upfront, contracted on, and payment collected, then, all other things equal, the leakage problem will be significantly less severe.

4. Are the transactions conducted in person or remotely/online?

Other things equal, leakage is harder to avoid when transactions are conducted in-person. The reasons are obvious: in-person meetings make it easier for the parties to share information, gain trust, and coordinate payment in cash, or via one of the many available peer-to-peer payment apps (e.g. PayPal, Venmo or WeChat). In contrast, online transactions can be much more easily monitored by the platform, which provides a way for it to control many relevant aspects, such as the information shared, the way the parties interact and earn trust, the way payments are made, etc. This is why dealing with leakage is arguably easier for Fiverr than for TaskRabbit, both of which help consumers find freelancers. Because Fiverr deals with digital freelance work, freelancers have an incentive to keep everything on Fiverr to ensure they get paid (and customers do too since their payment is not transferred until they are happy with the work), something Fiverr can monitor directly. By contrast, TaskRabbit mainly enables services provided in-person at someone’s home or office, so there is a lot more scope for the two sides to circumvent the platform.

5. What is the scope to add benefits that increase the value of transacting on the platform?

The more frictions there are between buyers and sellers, the more scope there will be for the marketplace to add benefits that increase the value of transacting on the platform. For instance, asymmetric information (one party not knowing much about the other) makes it riskier to transact directly, which is something the platform can help resolve by providing escrow, insurance and reputation systems. As an example, compare hotel booking ( and private home booking (Airbnb). To the extent hotels or hotel chains are well-known, guests feel quite comfortable dealing directly with the hotel for their stay, even if involves paying them upfront. The situation is very different when it comes to staying at an unknown host’s private home, which is why Airbnb can add a lot of value. Thus, leakage is generally more of a problem for than for Airbnb. The scope for increasing the value of transacting on the platform also depends on how simple or complicated the transactions are. For example, there is not a whole lot of additional value that can be offered by marketplaces that help people find unused spaces (garages, a room, an office) to store their stuff (e.g. Neighbor and Stache). Once a consumer has found a suitable space, there may not be much inherent value of continuing to transact via the marketplace, especially if what they are storing is not particularly valuable (so insurance is not a critical need) and the rental is long-term (so there are no scheduling issues). This is very different from a shiftwork platform like CareRev, which helps hospitals hire independent nurses on-demand. CareRev not only matches hospitals with nurses, it also helps both sides manage their scheduling issues, allows hospitals to better manage their workforce, standardizes and handles all aspects of the work contracts, handles payments, etc.

Suppose the answers to the questions above indicate there is a significant leakage problem. What can platforms/marketplaces do to minimize leakage?

There are two types of solutions: the choice of business model and the design of incentives.

Business model choice

The most obvious way to address leakage via the choice of business model is to eliminate transaction fees, which is what creates the problem in the first place. Indeed, if the leakage problem is severe enough (based on the questions above), this may be the only reasonable way to avoid it. Instead, marketplaces can monetize in other ways. One possibility is to charge product or service providers for the referrals of new customers by the platform. This is the approach chosen by Thumbtack, which help consumers find local handymen, housecleaners and other service professionals, and by Capterra, which helps businesses find suitable business software from thousands of different providers. These companies would have a hard time monitoring the actual transactions because in most cases, the parties involved would rather deal directly rather than channel all their communications and negotiations through the marketplace. Another option is to charge subscription fees to one or both sides of the platform: the hiring platform ZipRecruiter mentioned above has several tiers of subscription fees for recruiters (each tier allows for a maximum number of job postings) and dating platforms (e.g. Bumble, Hinge, Tinder) charge subscription fees to all users. Yet another option is to charge for listing, advertising and/or for prominent placement on the marketplace. The last of these options is Taobao’s business model, which allowed it to overtake eBay in China (eBay insisted on charging transaction fees for far too long, because that’s what was working in the United States).

Rather than give up on monetizing via transaction fees (which may be the most efficient way to monetize), marketplaces can consider another, more drastic way to avoid leakage. This involves taking control over the transactions themselves, thereby moving away from the platform business model altogether. For example, Hello Alfred provides at-home services (e.g. cleaning, laundry, grocery shopping, clothes alterations), delivered by assistants who are called “Alfreds” and are employees of the company. When they launched Hello Alfred in 2014, its founders did consider the marketplace business model, in which the Alfreds would have been independent contractors (like drivers on Uber or cleaners on Handy), but in the end decided that employing the Alfreds would allow them to offer a better customer experience.Among other issues, leakage would have been a serious problem under the marketplace model. Of course, this avenue is oftentimes simply not realistic. We wouldn’t advice Thumbtack to try to hire the huge variety of local service providers on its marketplace just to eliminate the potential for leakage.


The second type of solution to the leakage problem is the most generally useful, and it involves providing clear incentives for the relevant parties to keep transactions on the platform---thereby making it possible to charge transaction fees. These incentives can take the form of carrots or sticks. We greatly prefer (and advise companies to mostly use) carrots, though sticks can sometimes be helpful.

Offering carrots means the marketplace provides a lot of value added in terms of reducing frictions, such that participants find it a lot more appealing to transact on the platform than somewhere else. There are many value added services platforms can offer: the ability to track and share performance (particularly relevant for coaching and tutoring); the ability to manage dynamic schedules for service providers; insurance, payment escrow and dispute resolution to give both parties peace of mind; standardized contracts and protocols to streamline transactions; rewards for both sides tied to conducting transactions on the platform (e.g. appearing higher in search results, discounted merchandise and preferential access to desirable features); data analytics and recommendations based on past transactions, etc. Marketplaces who invest in such carrots have a much easier time convincing their participants to pay higher fees to conduct transactions through them. In other words, if users still prefer going around the platform, then either the marketplace is not creating enough value, or its fees are too high.

Sticks are punishments or restrictions that some marketplaces impose in order to fight leakage. For instance, Airbnb and eBay threaten to hide, demote or ban users that take transactions off the platform. A less drastic “stick” that marketplaces can use is to hide relevant identifying or contact information, which would allow users to communicate outside the platform. This makes it harder for parties to engage in leakage and can be especially effective when transactions are conducted fully online, so there is no opportunity to meet in person (e.g. Fiverr, Upwork).

As mentioned earlier, we much prefer investing in businesses that can deal with leakage via carrots rather than having to rely on sticks. Indeed, carrots are more sustainable in the long run. And using sticks to prevent leakage creates user resentment and does not really address the underlying leakage problem.

That being said, in some circumstances sticks can be quite effective. Some large platforms, including,, Expedia and Sabre, have traditionally forced their respective sellers to accept price parity clauses, which penalize sellers that attempt to take buyers off the platform with lower prices by banning them from the platform altogether. These restrictions have helped these platforms build their market positions. However, under regulatory pressure, or in some cases, new laws, these clauses have been removed, either partially or fully in Europe and Australia, and in the case of Amazon, in the U.S. Still, given the power of their recommendation engines, platforms can still discipline sellers that are pricing lower in alternative channels by demoting them in search results (as and Expedia seem to do).

Finally, we should mention an important pitfall that startups (and their investors) need to be aware of when evaluating the risk of leakage. Initially, when a marketplace is growing fast and users on either side are mainly discovering new transaction partners on the other side, rather than doing repeat transactions with existing partners, leakage does not appear to be a big issue and may lead to complacency, i.e. under-investment in features designed to minimize leakage incentives. Once users have settled down on a few counterparts they transact with often, then the leakage problem can become a lot more serious. Consequently, marketplaces should try to think about getting the right incentive systems in place from the start.

If you enjoyed reading this, sign up now so you will receive our posts directly in your inbox as they are published.

In the meantime, tell your friends!