Not All Marketplaces are Created Equal: Tales of a Marketplace Founder

My company Pared was recently included in an essay from Andrew Chen and Li Jin entitled What’s next for marketplace startups? Recently I’ve been seeing some great pieces about network effects coming from the insightful minds over at a16z such as 16 Ways to Measure Network Effects and The Dynamics of Network Effects. I always enjoy reading the analysis of smart people who are trying to synthesize and create themes about businesses like my own. James Currier at NFX Guild does a fantastic job of this and I’ve always been a fan of his take The Next 10 Years Will Be About “Market Networks” because it hits so close to home. The challenge is synthesizing so many disparate business models into a single tidy Medium post. Unfortunately, the downside of generalizing is that you miss out on the nuances and key differentiators to success of any one business or model. Andrew or James wouldn’t get many readers if they focused on very specific businesses or companies because it wouldn’t resonate with a broader audience.

It’s always fascinating to see the analysis and breakdown of businesses like your own from the outside perspective of an investor or academic, but we rarely see many insights shared from the inside. No one knows or sees the strength or weakness of a network or marketplace better than the people who are building them every day. I learn so much from spending time with fellow founders of marketplaces like my friends over at Honor and Trusted or friends who are executives at networks like LinkedIn. Although not all of their experience is directly applicable or relevant, I always come away armed with something new that inspires ideation and curiosity about my own business. The recent articles from the team at Andreessen Horowitz coupled with these conversations compelled me to reflect and share my experiences and insights with marketplaces.

One thing I’ve clearly learned over the years building marketplaces like Ebay, Luxe and Pared is that not all marketplaces are created equal. Marketplaces have evolved over time. As the VP of Marketing at Luxe, an on-demand valet service, I helped launch six new markets for a two-sided marketplace of valets and drivers. As the co-founder of Pared, a market network for hospitality workers and operators, I’ve led growth and marketing from day one. I’ve seen first-hand what it’s like to build both a marketplace and a market network. It is too easy and misguided to simplify all marketplaces into supply and demand. Marketplaces started as a venue for buyers and sellers to transact in person or offline (Wal-Mart, Costco, Sears). The internet allowed for platforms to bring together buyers and sellers online (Craigslist, eBay, Amazon). Adoption of smartphones untethered people to earn money from independently performing discrete tasks (Uber, Lyft Taskrabbit). Since then more independent workers have monetized their skills on marketplaces (Upwork, Thumbtack, Trusted). And now trained and certified individuals are starting to offer their professional services along with their identities and networks onto market networks (Pared, Atrium, Incredible Health). All of these evolutionary stages are viable business models that can clearly thrive and coexist in the modern world.

The evolution of marketplaces is tied closely to distribution and closing the information asymmetry gap between supply and demand (aka lead generation). Technology and new business models have essentially created transparency of opportunities. The first evolution was moving from traditional brick-and-mortar transactions, bulletin boards and classified ads to online exchanges for products and tasks. The second evolution was creating scaleable distribution networks for simple tasks such as dry cleaning, grocery shopping, taxi rides. The third evolution was enabling transactions that were traditionally brokered by agencies between skilled workers such as moving companies, childcare or office/administrative temp agencies. The final evolution is displacing the lead generation that professional services brokers such as law firms and nursing staffing agencies provide. Generally these firms and agencies have partners or leadership that sells to drive business and deal flow for the service professionals that they employ. These firms generally take a large cut of the profits for brokering these deals. New marketplaces are now cutting out the middleman and allowing the supply to keep more of the rake.

As someone who has experience in multiple stages of the marketplace evolution, I thought it would be an interesting exercise to break them down. In this post, I will address:

• The differences between two-sided marketplaces and market networks.
• How certain marketplaces benefit from network effects more than others.
• How different types of inventory that impact the defensibility of marketplaces.
• The importance of repeat transactions and reputation in marketplaces.

Marketplaces vs. Market Networks

James Currier, co-founder and partner of NFX Guild, knows more about marketplaces and networks than most. He’s an investor or advisor to Honeybook, Houzz and AngelList and was the founder of multiple network and marketplace companies. He also coined the term “market network” in June 2015 in a TechCrunch article “From Social Networks to Market Networks.”:

“Marketplaces” provide transactions among multiple buyers and multiple sellers — like eBay, Etsy, Uber and LendingClub.

“Networks” provide profiles that project a person’s identity, then lets them communicate in a 360-degree pattern with other people in the network. Think Facebook, Twitter and LinkedIn.

What’s unique about market networks is that they:

• Combine the main elements of both networks and marketplaces
• Use SaaS workflow software to focus action around longer-term projects, not just a quick transaction
• Promote the service provider as a differentiated individual, helping to build long-term relationships

Currier defines a market network needing a SaaS workflow software to focus action around longer-term projects and not just brief transactions. I would take this a step further and say that the same effect occurs with a series of multiple repeat interactions on a marketplace platform. An example of this could be a cleaning service like Handy or a dog-walking service such as Wag where customers prefer to have repeat transactions with a specific service provider they trust or like. This differs from one-off transactions such as TaskRabbit or Uber.

In our world at Pared, a restaurant needs a cook for several days a week. Different cooks with the relevant skills will be offered and fill those open shifts. It’s important for the restaurant to build a good reputation as a desirable place to work otherwise no other cooks will be willing to work there. The same goes for a cook. If enough restaurant operators don’t like working with a cook, it will damage their reputation and ability to receive future offers from other venues. In a true market network, each node should be able to provide useful feedback or information on any other node. All of the nodes in a market network can be interconnected. People who work in the restaurant industry have ties to others in the industry because they’ve worked together in the past. Cooks also become restaurant owners. And occasionally restaurant owners and executive chefs go back to working for someone else if things don’t work out (as is often the case in the restaurant world). We’ve seen many executive chefs who used Pared as customers sign up to find work and additional income on the platform. This fluid mobility between both sides of the marketplace makes a market network even stronger because it creates more interconnectivity between the nodes. Some might say we are building the LinkedIn for the hospitality industry by mapping this network through our marketplace.

In a market network, where identity does matter, it is possible to capture returns for incremental marginal utility.

Benefits of a Market Network

Over the past decade or so, we’ve seen the incredible rise of marketplaces fueled by billions of dollars of capital. Scaling a two-sided local marketplace like Uber, Airbnb, Instacart and Doordash is expensive. It’s a race to dominate local markets that costs a ton in marketing dollars to win the hearts and minds of customers. The challenge is that the longer you are in a market, the more expensive it becomes to grow because of saturation. Your supply well begins to dry up and your cost of acquisition (CAC) starts to skyrocket. When you start to see television commercials, billboards and bus wraps, you can visually see the increasing CAC. I’ve heard that it costs a ridesharing company in San Francisco $2,000 to acquire a new driver now. That’s because everyone is already familiar with Lyft and Uber and if they haven’t already become a driver, it’s unlikely they ever will. Expensive referral bonuses and car leasing programs are all ways to entice new drivers to onboard. Activating a new driver in a mature market is extremely costly. Network effects should drive the cost of acquisition down over time as a market matures.

In a market network, where there are existing relationships, word-of-mouth leads to referrals and organic growth which lowers the CAC. We’ve seen this happen in our business where many people in hospitality have connections to other career hospitality workers. Uber drivers and Instacart shoppers have no prior community or network with other drivers and shoppers. Unlike these marketplaces where moms become taxi drivers and students become grocery shoppers, our pros have worked in this industry for years as this is their trade.Network effects thrive when there are connections between nodes in an actual network. Referrals should flow freely in a market network. At Pared, over 50% of our new users come organically or through word-of-mouth. There is a difference between virality and network effects. Currier explains the difference well, “Virality is about growth: people refer your product to others.Network effect is all about retention and defensibility of a business. Once you’ve built a strong network effect, it’s really difficult for others to compete with you.”

Marketplaces that don’t benefit from network effects that drive CAC down over time are in a race to raise capital to grow and grab market share while developing a cost efficient model where the unit economics are positive. In the case of ridesharing, the rising acquisition costs and increasing driver churn in saturated markets require inordinate amounts of capital to buy time to reach the goal of self-driving cars to replace drivers.

In a marketplace where identity doesn’t matter, the upside on returns you can capture are limited. You can capture increasing returns on scarcity (surge pricing) and time. Lyft and drivers can make more during a peak traffic times or high demand times like New Year’s Eve or Outside Lands. Instacart can charge a higher fee for an earlier delivery time slot if you want your groceries badly enough. But outside of those parameters, there aren’t many ways to drive revenues or earnings other than volume. You can’t raise P any more, so you have to increase Q to drive growth.

In a market network, where identity does matter, it is possible to capture returns for incremental marginal utility. A parent would be willing to pay a premium for a sitter on Trusted because they have used them in the past. We’ve had customers pay a premium to get a specific Pro they’ve liked working with before. A rider doesn’t really care who their driver is on Uber, or who their delivery person is on Postmates, so they wouldn’t be willing to pay more for the option to choose someone in particular. Prior experience or interaction is one way to capture higher returns, but so are specialized skills. Customers on Honor would be willing to pay a premium for a Care Pro who was certified for seizure recognition and response to care for their elderly loved one who suffers from seizures. Restaurants on Pared are willing to pay us more for a high-skilled line cook who has experience working at Michelin-starred establishments or for a sushi chef. The upper limit to what returns you can capture on a market network are much higher than that of most marketplaces. I will speak more in a later post about the how differentiated supply increases defensibility.

Network Effects

Many marketplaces benefit from network effects. Not all marketplaces benefit equally, based on the various network effects that occur. The amount of time it takes for network effects to kick in also varies from marketplace to marketplace. NYU professor Arun Sundararajan classifies network effects into five broad categories:

1. Direct (Same-Side) Network Effects: Increases in usage lead to direct increases in value.
2. Indirect (Cross-Side) Network Effects: Increases in usage encourage consumption of complementary goods, which increases the value of the original product.
3. Two-Sided Network Effects: Increases in usage by one set of users increases the value to a different set of complementary users, and vice versa.
4. Local Network Effects: Increases in usage by a small subset of users increases the value for a connected user.
5. Compatibility and Standards: The use of one technology product encourages the use of compatible products.

Instagram benefits from direct network effects as it becomes more valuable as more users join and create more content. Google benefits from indirect network effects as users of G Suite (Docs, Sheets, Slides, Mail, Calendar and other Google productivity software) increases the value to all users as more users join and collaborate. Uber benefits from two-sided network effects as more riders join, it becomes more valuable to drivers for earning more income. The more drivers on the platform, the faster it is to get a ride and more valuable Uber becomes to riders (this value plateaus at some point making it an asymptotic network effect). Trusted benefits from local network effects because unlike drivers, caregivers aren’t commoditized. The more caregivers that can speak a certain language or care for special needs children, the more value there is for a specific population of parents but not to all parents. And finally, the compatibility and standards network effect has helped propel companies like Adobe to success with the mass adoption of PDF, Illustrator and Photoshop file formats. People are forced to purchase Adobe software licenses in order to use these universally accepted standards. Salesforce is another company that has benefited from the fact that most salespeople are trained and familiar with the de facto standard for CRM software. Many enterprise companies are practically forced to choose Salesforce as their CRM because it is the software that almost all salespeople know.

Pared has the fortune of tapping into multiple network effects. Indirect network effects: as we expand into new roles such as bartenders and baristas, we increases the value to our operators. It also increases value to our pros who have diverse skills and experience because it opens up new earnings opportunities. Two-sided network effects: the more restaurants and cities we add, the more people want to work on the platform; the more talented and skilled people that work on the platform, the more restaurants will want to access that pool. Local network effects: As we add new markets and cities, the platform becomes even stronger. We’re already seeing Pros working across markets between San Francisco and New York. The restaurant industry is a transient one and people move around all of the time. Being in multiple cities is very attractive to the supply side. It is also compelling to the demand side because many restaurant groups and chains have multiple locations across cities. We work with customers like Proper Food and others that have operations in both New York and San Francisco. And finally if we can get to the point where our brand is trusted for vetting quality hospitality professionals and penetration is high enough, we aspire to one day become the industry standard for staffing and hiring.

Leveraging Existing Networks

Spinning up a marketplace from scratch is very, very hard. Most people run into the chicken or the egg conundrum. When I was working on building Luxe, an on-demand valet service, I learned first hand how hard it was. As a new marketplace, you don’t benefit from network effects until you hit a certain critical mass. I helped launch six new markets in New York, Seattle, Chicago, Austin, Boston and Philadelphia. As the VP of Marketing, it was up to me to acquire supply and demand in these nascent markets. There are many levers to pull and they aren’t the same in each city. On the supply side, we recruited valets via Craigslist and Indeed postings, referral bonus programs and even canvasing college campuses. On the demand side, I tried everything including Facebook and Google ads, referral programs, corporate perks partnerships radio ads, podcast ads, postcard mailers, billboards and more (read: expensive). In our most mature market, San Francisco, we had brand recognition because of our blue jackets and scooters. But in a city like Los Angeles which is much more spread out, we didn’t have that same recognition. Network effects working in one market has no bearing on them working in another market. Needless to say, when you don’t have an existing network to take advantage of, building a network can be daunting and ultimately very capital intensive to get to the point where network effects kick in. When you don’t have an existing professional to build upon, you are forced to boil the ocean to find supply and demand which can be extremely costly. Instacart has to market to anyone who needs groceries or anyone who could shop for groceries (everyone). Uber and Lyft have to market to anyone who needs a ride (everyone) or who drives a car. Spending marketing dollars reaching these large and broad audiences is not for the faint of heart or wallet. Compare this to Incredible Health, a market network for nurses. Nurses have existing professional networks of other nurses, medical professionals and hospitals that they’ve trained with or worked at in the past. Targeting nurses is a lot more cost effective than trying to target anyone who shops for groceries. And once you reach one of those nurses, they can spread the word to their colleagues and peers in the healthcare industry.

My experience at Luxe prepared me well for creating a marketplace. But this time I had the benefit of starting with an industry that had strong built-in networks. At Pared, we are not afraid to go into new markets and dominate because of our existing network and relationships in the hospitality industry. This is a direct reflection of the power of real world market networks and network effects. Just as Facebook started with the existing network of the Harvard students, Pared started with our network of San Francisco restaurateurs. My co-founder Will Pacio worked with some of the best chefs in the world under Thomas Keller at per se and The French Laundry. His experience and industry connections add credibility and trust that gives Pared access into an exclusive fraternity of restaurateurs. Will is so ingrained in the community that he is on the board of directors of the Golden Gate Restaurant Association. Without the validation of referrals and testimonials from the network, we probably would not be working with some of the most well-respected restaurant groups in the world such as the Crenn Dining Group (Atelier Crenn, Bar Crenn, etc.), Thomas Keller Restaurant Group (Bouchon Bakery), Jean Georges Group (Jean Georges, ABC Kitchen, etc.), and Major Food Group (Parm, Dirty French, etc.). Every one of these esteemed restaurants we add to our demand side increases the desirability and value of our network to both supply and demand. Social proof and reputation attracts more talented supply and prominent demand. Because restaurant groups have locations across multiple markets and chefs are nomadic and have connections around the world, we don’t have to start from scratch each time we launch a new market. The compounding cross-side, two-sided and local network effects will continue to drive growth and defensibility in our business.

Starting a marketplace is hard. Starting a marketplace where there is no existing network to leverage is even harder. Existing networks allow a market network to map on top of connections and relationships that can instantly take advantage of network effects to grow. These network effects help drive increasing LTV and decreasing CAC which ultimately lead to a healthy business.

Defensibility of Supply

The simplest form of a marketplace has one product or service that is sold through a transaction between a buyer and a seller. Most marketplaces evolve from selling one “thing” to selling a variety of “things” by leveraging their technology and platform to expand and diversify. Amazon started selling books when they were founded in 1994 and then started to sell everything else in 1998. Uber started with ridesharing and moved into food delivery with Uber Eats and logistics with Uber Freight (after a failed attempt with Uber Rush). Some marketplaces like Rinse, Wag or Handy never diversify and stick to doing one thing very well to capture market share. There’s something to be said about having a simple product offering and marketing message. The problem with selling one “thing” is that there are few barriers to entry unless that product or service is very difficult to copy.

There are three distinct types of marketplaces that are separated by the type of supply or inventory:

1. Commoditized — A marketplace where the product or service offered is fungible. Lyft, Bird and Instacart are examples of commoditized marketplaces. A ride from point A to point B on Lyft is an easily replaceable service offering from one driver to another, the only difference is the time it takes for one to get to you. All shoppers on Instacart are essentially the same to you as you can’t differentiate one shopper from another, you just want the first shopper who is available.
2. Differentiated — A marketplace where the products or services are differentiated. AirBnB is an example of a differentiated marketplace where inventory is varied by sized and amenities. The number of bedrooms, pool or no pool, these are all factors in the decision of the demand side. This also limits the demand for different types of inventory. A party of 8 is only considering a handful of available options on the marketplace.
3. Multi-Faceted — A marketplace where the products or services are differentiated, but have a breadth of offerings or skills. In a differentiated marketplace such as AirBnb, a one bedroom can only be a one bedroom and a house either has a pool or it does not. In a multi-faceted marketplace, each product or service has multiple dimensions. These dimensions are independently valuable to a single customer or to a broader set of customers, extending the potential earnings for the supply side.

Commoditized Marketplaces

A commoditized marketplace is the fastest to start and grow, because you can focus on one product offering and often throwing gobs of capital can help accelerate that to market dominance and crushing competition. That being said, because it’s so simple, it also makes it very easy for others to copy and join the fray. The current electric scooter wars are evidence of this. If you have the capital to purchase a significant number of aftermarket Segway scootersand procure off-the-shelf hardware to enable geolocation and locking functionality, you just need to build an app to scan a QR code to unlock and process payments, and you can start a viable scooter company. Obviously this is a gross oversimplification, but the number of scooter companies that flooded the market in a short period is evidence of this: Bird, Lime, Spin, Scoot, Skip and that’s just the in the States. Most consumers don’t really see any differentiation between scooter companies because it’s so difficult to differentiate. When I am looking to go home, I don’t care which scooter I use, because essentially they’re all the same (and sometimes literally because they’re made by the same manufacturer). I’m just optimizing for finding the closest available scooter. The company that wins my business is the first one I can find. The same can be said about ridesharing. How many times have you opened up both your Lyft and Uber apps to look for a ride. Many riders are consider the two companies interchangeable, just as the drivers do. They optimize for price and convenience (time/distance). Commoditized marketplaces are the least defensible until they reach a point of market dominance which is why so many are so well funded. Reid Hoffman and Chris Yeh refer to this as first-scaler advantage in his book Blitzscaling: The Lightning-Fast Path to Building Massively Valuable Companies: “The most frequent offensive reason for blitzscaling is to achieve a critical mass that confers a lasting competitive advantage.” Unfortunately, even if you dominate a market, competitors and new upstarts can still threaten to erode your share just as Lyft did to Uber in San Francisco. When there’s no stickiness or reason to be loyal to a brand, switching costs are negligible.

Differentiated Marketplaces

A marketplace that has a variety of product offerings that meets different customer needs is a differentiated marketplace. Zappos is a differentiated marketplace because it sells all types of shoes from boots to dress shoes to sneakers and not just one shoe. Caviar, Uber Eats and Doordash have a broad selection of restaurants to choose from. Customers choose between multiple delivery services based on the options available to them. But if I’m craving Cheesecake Factory, I have to use DoorDash, if I want a Big Mac, I have to use Uber Eats. These marketplaces both provide food delivery, but they’re not completely interchangeable because of exclusive agreements with restaurants. And although there is no exclusivity, Airbnb, VRBO andHomeAway all have a unique portfolio of vacation rental options which differentiate them from one another by inventory. Differentiated inventory is what sets these marketplaces apart and makes some more defensible than others.

Differentiation is powerful because the more sticky a platform becomes for the demand, the less likely they will leave. If they come for one product and you introduce another valuable product, they are less likely to churn because they rely on you. This was the case at Yahoo!, where I took my first tech job as a product manager in 1999 (yes I’m dating myself here). Long before the days of BI tools like Looker or Periscope, we discovered that if we could get a user engaged with three Yahoo! products across the platform (My Yahoo, Finance, Fantasy Sports, Shopping, etc.) that they were going to stick around. Having differentiated product offerings increases engagement and decreases the chance of churn.

Multi-faceted Marketplaces

Finally we have multi-faceted marketplaces. The inventory or supply in these marketplaces are differentiated other supply in multiple ways. This could apply to physical products that are multi-purpose like Storefront which rents out physical spaces for pop-ups for retail, events or showrooms. This generally applies to service or skill-based marketplaces where a professional has different abilities or skills that are valued on the platform. Ridesharing, delivery or personal shopper platforms only require one skill to perform a transaction. Compare that to Upwork, where a pro has a variety of skills from language to programming to software proficiency. On Pared, a skilled line cook can also work as a prep cook, dishwasher, or busser because they have prior experience in those roles. A seasoned server has previously worked as a bartender, barista and cashier. These multi-dimensional skillsets make the supply side much more valuable to the marketplace because it unlocks more demand. Before we were focusing on restaurants, but as we’ve added roles like bartender and barista (skills that our many on our supply side already had), we’ve been able to serve new categories of customers including nightclubs/bars and cafes. Additionally, a professional can increase her earnings potential with each skill she acquires. The reason LinkedIn tried to bridge the skills gap of recruiters and candidates with the acquisition of Lynda.com was to help candidates attain new desirable skills to meet hiring requirements. This was an effort to level up supply to meet unmet demand and create successful recruiting matches which drive their Recruiter business.

The restaurant labor force is highly differentiated with a broad spectrum of skills and experience. Pastry, oyster shucking, butchery, sushi, etc. are all very specialized skills. A sushi restaurant is willing to pay a premium for a sushi chef because of scarcity. Li Jin and Darcy Coolican from A16Z explain the power of this differentiation, “Platforms/marketplaces with more differentiated inventory have stronger and longer-lasting network effects, because they have a diversity of inventory that suits the unique preferences of customers (while maintaining just-enough substitutability across that inventory as well).” The more sushi chefs added to the supply pool, the more valuable it becomes. Curation and matching are increasingly critical with differentiated supply. The team that can do this best will build a deep moat of defensibility.

The importance of how difficult curation and matching are to the success of a marketplace cannot be underestimated. If you can’t meet your demand with quality supply, then you will end up with a churned customer. I am often asked if Pared will expand into other verticals outside of the hospitality industry. There are enough nuances within hospitality alone that getting that right is challenging enough (that and the hospitality industry has a huge TAM). When you try to be all things to all people, you end up not doing anything very well. TaskRabbit is the cautionary tale of a marketplace that struggled to find an identity. Initially called RunMyErrand.com, TaskRabbits would could be tasked with just about anything you could think of, which was likely its biggest weakness. Ultimately after raising $38M to reach a $120M valuation, TaskRabbit sold to IKEA for $40M. TaskRabbit went from an app that could find people to run any number of errands to assembling flat-packed Swedish furniture. Having a focus and identity is important for a company because customers need to know what you do.

First-scaler advantage is critical to the success of a commoditized marketplace such as ridesharing or scooters. That being said, it doesn’t guarantee that second-movers can’t come and steal away share which ends up costing more capital to fend off. On a differentiated marketplace, you can lock in customers with exclusive partnerships for supply on food delivery platforms. But once those expire, you might experience attrition. If I want a Chipotle burrito but they’re not on DoorDash anymore, I probably won’t be using DoorDash. This is the risk of customers who are loyal to the brands you carry and not your brand. The same can be said for Netflix or Hulu. If a show like “Friends” is no longer on that platform, some people will unsubscribe, so Netflix is forced to pay $100 million to keep it. When you map your marketplace to an existing network, it’s truly winner takes all. Users are unlikely to leave a network or platform that all their peers are already on. It is very difficult to knock off the incumbent if they own a network. Facebook and LinkedIn are prime examples of the power and longevity of mapping existing networks. All of their relationships and connections live on the platform and they’ve invested too much time into building up their identity there. The switching costs are extremely high. Successfully building a marketplace on top of an existing professional network has long-lasting rewards such as lower cost of acquisition, higher frequency of transactions and decreased churn.

Repeat Interactions and Reputation

In a commoditized marketplace like Uber or Lyft, the chances of getting the same driver matched again are very low (unless you live in a much smaller market). The consequences of getting a bad review from a rider does not necessarily limit your pool of opportunity until you get enough negative ratings to drop you below the threshold. Commoditized marketplaces are generally one-time interactions between supply and demand. Repeat interactions with the same service provider generally only occur on a platform where reputation (identity or brand) matters because as a customer you want to have a similarly awesome experience and control for quality. The term “repeat transactions” or “repeat interactions” in this post refer to multiple interactions with a specific service provider on the supply side vs. multiple transactions with the marketplace itself. In a market network where repeat transactions occur, it behooves a service provider to perform well because not doing so will result in fewer opportunities and consequently lower earnings. A nurse working on Incredible Health can’t afford to perform poorly at a hospital because there are only so many hospitals that can provide work. A market network keeps supply accountable because their reputation and identity matters to every node in the network. A freelancer on UpWork depends on their reputation in order to get future contract jobs, so they do everything they can to keep their ratings high for each job.

Repeat transactions, new connections and interactions build the network value over time. For a business like HoneyBook, people ideally only have a wedding once and won’t ever interact with those vendors again. The same thing happens with Thumbtack, although you build up an identity and reputation on the platform, the frequency that someone needs a flat screen wall-mounted is pretty small. These connections become one-time service transactions where the consumer is unlikely to need those services again once the transaction is completed. Compare this to Trusted, where parents are matched with childcare professionals. These are repeat transactions where building a relationship and reputation are important for both sides of the market. For a market network to truly thrive, the nodes in an N-sided marketplace needs to find value in the connections. Frequency of interaction and relationships matter. That is difficult to do outside of a vertical wherenetworks already exist but are not yet codified.

When reputation matters on a platform because opportunity is determined by repeat transactions like Upwork, Trusted or Pared, there is a stickiness that doesn’t exist in commoditized marketplaces. How many cars do you see with both Lyft and Uber stickers on the road? Almost all drivers switch between both services because they’re trying to maximize their earnings and volume of transactions. The reputation on both platforms is a simple rating scale. They can erase a bad ride with more good rides because they won’t get that rider who rated them negatively again. Compare this to a Pro on Pared who gets a negative rating for a gig as a dishwasher, he or she will not get offers for that establishment again and that limits their opportunity for earning potential. On the flipside, the more a Pro builds their reputation on the platform, the more opportunities they will unlock as they gain new skills. The switching costs are much higher for them to start over and build that reputation again on another platform. Every testimonial from a node on the network whether it’s a restaurant operator or another Pro improves their standing. That pro or restaurant operator has their own network and reputation that it imputes with a positive rating or review. As a rider, the identities of other riders who rated the rider are inconsequential to me.

The biggest hurdle for any skills-based marketplace is quality. It doesn’t matter how much supply you have in a marketplace if they can’t actually perform the tasks required. One the most valuable things a marketplace can bring to the table is skills verification. This can come in the form of interviews, work history verification or taking exams to prove their capabilities. Meeting people in person and interviewing or auditioning them is probably the easiest way to find out if they can get the job done. Verifying their work history with references is another way to confirm that they are capable. Another common and standardized measure of abilities is certifications. Whether it’s completing a medical board exam or the bar exam for lawyers, these tests certify a minimum of knowledge and expertise. While they don’t guarantee a good doctor or lawyer, exams filter quality to a degree. A professional marketplace such as Gigster which matches tech companies with freelance programmers who work at Google, Amazon or other reputable tech companies and are looking for projects on the side (they definitely don’t need the extra cash!). On top of work history, Gigster vets through development and design screenings and then finally a personality phone screen. Coding challenges are an excellent and scaleable way to measure a developer’s abilities. Customers are paying a premium for this tedious upfront screening so they can have peace of mind. All of these hurdles are essentially a filter for quality. If a company can find scaleable ways to screen for quality, they can build supply and grow faster than their competitors.

Although subscriptions and SAAS business models have much lower churn rates than most marketplaces, repeat transactions can drastically improve retention. Repeat transactions with a service provider means that the customer trusts and wants to work with that person again and again. If you build the features to create loyalty with the service provider on the platform, then you in turn will keep the customer loyal to your platform as well. The more value you generate for a service provider, the better your retention for both sides of the marketplace will be. Therefore it’s critical to keep supply happy on any marketplace, because once they leave, the customer is likely to leave with them. Alternatively, if you are able to curate high quality supply and prove that to the customer, you may be able to salvage their business. The only way to do that is to consistently provide them excellent matches and a phenomenal customer experience. The upside of a commoditized marketplace is that one-off transactions means that a customer isn’t likely to churn if any one service provider leaves the platform because there’s no attachment. I won’t stop using Lyft if a driver I once took a ride with stops driving, because I likely won’t even know. If the stylist I always see leaves StyleSeat, then I will most definitely know and I will likely leave with them unless I have had previously good experiences with other stylists on the platform.

The biggest threat to a repeat transaction and reputation-based market network is disintermediation. If a customer finds someone with a great reputation and works with that service provider consistently, it only makes sense for them to hire them directly outside of the platform. This was the downfall of home-cleaning marketplaces like Homejoy. Disintermediation leads to a leaky bucket of both supply and demand. The only way to avoid disintermediation in a transactional market network is if the platform is compelling and sticky enough for both parties. Pros prefer the flexibility and higher wages on Pared and operators are constantly struggling with turnover and recruiting regardless of hiring new employees. Trust and convenience are critical to retention. Jonathan Golden, now a partner at NEA, talks about the importance of trust and convenience for his alma mater Airbnb in his post “Four Questions Every Marketplace Startup Should Be Able to Answer”.

Building a company is like assembling a Cessna. Building a marketplace is like assembling a Boeing 747.

Conclusion

There are academics that have researched and written white papers about network effects and marketplaces that are far more intelligent than myself. I can only share the insights from my personal exposure to building marketplaces and the learnings from my brilliant friends who have started and worked on marketplaces. Reid Hoffman likens building a startup to jumping off a cliff and assembling an airplane on the way down. Building a company is like assembling a Cessna. Building a marketplace is like assembling a Boeing 747. If I’ve learned anything, it is that every marketplace has its own unique nuances and challenges (legislation, geography, unions, etc.). Anyone considering starting a marketplace business should be aware of the types of marketplaces and the potential network effects that they could benefit from. Those who are already in the thick of building a marketplace or market network should create products and features that enhance and accelerate those network effects that can propel their success forward.