"NFX has proposed a similar idea called network bonding theory."

"Passive Participation: When Token Incentives Work"

"The first is Helium, a decentralized network that provides cheap, easy internet access to IoT devices “in the wild” — like e-scooters and sensors."

"The second example is Arweave, which is described as a decentralized, censorship-resistant storage network."

"Another example is Compound, a lending network. Lenders deposit their crypto assets into a lending pool for borrowers to access."

"They all require passive participation from users, particularly from the supply-side of their network (not very different from the concept of passive crowdsourcing in data networks)."

"That gives them financial upside and increases the utility of the network at the same time."

"The supply-side does not need to actively engage with the network to benefit from this financial upside."

"However, networks with passive participation also tend to be rare."

"Active Participation: The Limits of Token Incentives"

"One of the most important principles of bootstrapping a network is to start with the most underserved users."

"You also need the right type of users, i.e. those who feel the problem most deeply and would put up with any amount of friction to engage with your network."

"Tokens can be a blunt instrument to target this underserved niche because they can attract the wrong type of users — those drawn to financial incentives, and not the near-term utility of the network."

"When this happens, it becomes very difficult to reach the required density of the right kind of users."

"This shows what happens when there is a disconnect between financial incentives and network utility — instant growth out of the gates, followed by a painful decline."

"To overcome these network effects, Looksrare executed a “vampire attack” on Opensea, i.e. it distributed (or “airdropped”) LOOKS tokens for free to high volume Opensea users."

"This go-to-market (GTM) approach should have been enough for Looksrare to beat the cold start problem and scale its network. Unfortunately, financial incentives led to user behaviors that weren’t aligned with network utility."

"Interestingly, genuine trade volumes began to collapse as token payouts normalized."

"Sushiswap, a decentralized exchange, executed a vampire attack on Uniswap in August 2020. The results, again, followed a similar trajectory although perhaps less stark (and also plagued by governance issues)."

"It clearly can, but it will require active participation and engagement from its user base — with scaling tactics that are more likely to resemble active web2 networks like Roblox, rather than passive web3 networks like Helium."

"At a high level, the only way to address this problem is to link token incentives to network utility, i.e. ensure that users can only receive token incentives if they add value to the network. In other words, rewards need to be restricted to specific, desirable actions, not just adoption."

"Pay attention to the kind of network you’re building and how token incentives interact with network utility before relying on them as a bootstrapping solution."

"Facebook is among a handful of truly generational internet companies."

"First, it led to more complexity as there were now multiple ways for users to interact — Facebook posts (to all or some connections), posts to groups, direct messages, sharing pictures, commenting, likes, etc."

"And second, some users posted more frequently than others and began using Facebook for distribution, i.e. they were “broadcasting” to anyone they could reach, not just their close connections."

"The flip side of these changes is that they will dilute the social graph even further."

"This leaves a clear opening for a new wave of “unbundlers” to recreate a social graph — one for sharing photos with your real friends."

"Managed (or full-stack) marketplace is a loose term for a marketplace that attempts to improve its customer experience by being more involved in the execution of transactions."

"The food delivery space is a simple example of the transition to a “managed” model. The first wave of food delivery marketplaces was led by Just Eat, Grubhub, and Takeaway.com."

"These companies simply connected restaurants to consumers, leaving restaurants to handle deliveries themselves."

"Deliveroo, DoorDash, and Uber Eats were able to displace them by offering a superior experience to both restaurants and consumers."

"These challengers owned last-mile delivery in addition to connecting demand and supply. This made delivery times more predictable and also allowed consumers to track progress in real-time."

"Other managed marketplaces may own background checks, verification, customer service, etc. depending on the product or service in question."

"Based on their level of involvement, there are two types of companies that founders and investors have called “managed marketplaces”:"

"True “Managed” Marketplaces"

"In other words, the additional services provided by the marketplace are meant to enhance interactions between buyers and sellers."

"As a result, this additional involvement does not disturb the network effect between buyers and sellers."

"Take Vestiaire Collective as an example — a second-hand marketplace for luxury fashion."

"Sellers first ship their product to Vestiaire Collective for authentication, before it is shipped to the buyer. Vestiaire Collective’s involvement helps establish trust with buyers and makes the fulfillment process easier for sellers."

"True “managed” marketplaces are also common in regulated services. For example, childcare marketplaces like Bubble, Kalendit, Koru Kids and Trusted need to ensure that all service providers are background checked and qualified before accepting them into their marketplace."

"However, they still have lower gross margins than unmanaged marketplaces because of their higher cost base."

"Marketplaces in Name Only (MINO)"

"In some cases, companies get so involved in the transaction that buyers and sellers no longer interact with each other. Instead, the company takes on the role of the supply side itself to control the customer experience."

"Since the demand and supply sides do not interact, network effects no longer exist. In other words, they no longer exhibit the defining feature of a marketplace."

"Take Opendoor for example. Home sellers don’t list their properties on Opendoor. Instead, they sell their homes to Opendoor (after an assessment)."

"What are the consequences of losing network effects? For one, capital requirements go up dramatically."

"In contrast, true marketplaces can leverage supply to attract demand organically and vice versa — this reduces customer acquisition costs over time leading to exponential returns on marketing investments."

"Instead, they simply need enough capital to acquire a “good enough” level of inventory in one geography."

"To summarize, network effects are the defining feature of marketplaces — a fact that is often lost in the discussion of how “managed” they are."

"True “managed” marketplaces aim to ease interactions between buyers and sellers, enhancing their network effects."

"In contrast, marketplaces in name only (MINOs) take on the role of the supply side themselves, removing all network effects."

"But at scale, MINOs tend to be more capital hungry, have weaker moats, and are more expensive to grow than true marketplaces (“managed” or not)."

"In order to bootstrap these interactions, marketplaces need to hit a critical mass of demand and supply. In other words, they need to have “liquidity”."

"Liquidity or critical mass is easier for 1-sided networks (e.g. social networks like Facebook) to achieve because the same users represent both demand (content consumption) and supply (content creation)."

"In general, the liquidity of any network is a function of the number of users and network density, i.e. the number of actual connections between users relative to the number of users."

"On marketplaces, these connections link demand with supply and, therefore, liquidity depends on the number of suppliers that can be matched with a given customer request, i.e. density of supply relative to demand."

"supply density can be broadly measured in two ways — the ratio of “match-able” supply to demand within a region or within a category."

"The relative importance of each type of supply density differs based on primary marketplace characteristics — geographic range and nature of supply for the marketplace in question."

"This means that Uber needs to maintain a specific supply to demand ratio, within the radius of a few miles, in order to hit that targeted wait time."

"As a result, hyperlocal marketplaces typically need to make significant investments in field operations to achieve and maintain liquidity."

"In contrast, Airbnb is a marketplace with cross-border network effects."

"Guests judge Airbnb’s product experience by their ability to find a high-quality property at a destination of their choice."

"Airbnb still needs to maintain a specific supply to demand ratio within a destination to achieve liquidity, but that can be spread across a larger geographic area."

"But in either case, geographic density is not a significant constraint to liquidity."

"Now let’s take a look at supply density from a different point of view. Uber has commoditized or interchangeable supply."

"This means that Uber riders are just looking for on-demand transportation, and are not sensitive to the brand of the vehicle or identity of their driver."

"But it also has a benefit, i.e. it is much easier to acquire a critical mass of supply across just a handful of categories."

"By extension, achieving liquidity across categories should be more challenging for marketplaces like Airbnb that have differentiated supply."

"This form of differentiated supply creates immense complexity in enabling matches and makes Airbnb more defensible, but at the cost of liquidity challenges, i.e. it is very challenging to acquire a critical mass of supply across each of the numerous categories."

"What makes marketplaces defensible and scalable also has a direct impact on how easy or difficult it is to reach critical mass."

"Since these are inter-dependent, we can layer liquidity drivers on top of the marketplace matrix to identify broader patterns."

"Clearly, marketplaces that have both commoditized supply and cross-border network effects (Tier-2A, bottom-right quadrant) face the lowest hurdles to liquidity."

"Conversely, liquidity is most challenging for hyperlocal marketplaces with differentiated supply (Tier-2B, top left quadrant)."

"In the absence of deep pockets that can fund a “brute force” approach, Tier-2B marketplaces have two options to achieve liquidity."

"The first approach is to use a “come for the tool, stay for the network” approach."

"In the absence of SaaS integration, the only other way to bootstrap marketplaces that are both hyperlocal and have differentiated supply (Tier-2B) is by relying on “side switching”."

"Side switching occurs when marketplace participants do not have fixed roles, i.e. a buyer can become a seller and vice versa (think Freaky Friday applied to marketplaces)."

"This is an approximation of what happens on social networks, where most users simultaneously act as demand (content consumption) and supply (content creation)."

"Once they create liquidity and grow, they can expand to other categories and leverage the existing base of users to bring in more valuable network participants with fixed roles."

"In order to maximize the liquidity advantage from side switching, marketplaces need users to switch back and forth from the demand to the supply side frequently, almost on par with social networks."

"Rather, marketplace startups need to embed side switching as a core part of their value proposition."

"Poshmark is a great example of a Tier-1 marketplace built on side switching."

"Their network effects are natively layered, i.e. every project combines multiple forms of network effects."

"Their network effects are also weaker and less defensible compared to web 2.0 variants, at least so far."

"However, the addition of more nodes increases capacity, and therefore value, for buyers of the Ether token."

"More token buyers increase the value of the Ether token and therefore make it more valuable for nodes to validate transactions."

"In other words, these are cross-side network effects on a 2-sided interaction network, with side switching built-in (token buyers can also be validators and vice versa)."

"However, this network structure also creates a couple of challenges: (1) a unique form of negative network effect, and (2) commoditization risk."

"So beyond a point, the addition of a token buyer decreases the value of the network for all other token buyers."

"This is partly because blockchains are identity agnostic, i.e. the identity of each node does not matter to other nodes or to token buyers."

"And so, each new node adds less incremental value to the network as it scales."

"Unlike blockchain protocols, the telephone network was identity focused, i.e. you could not contact a specific person if they did not have a telephone connection, even if others did."

"Competing blockchain protocols merely need to become “large enough” relative to their transaction volume or activity to compete — which has led to a sea of competing blockchain protocols and tokens."

"First, it doesn’t have the “matching” (or “app store”) component that we see in Web 2.0 platforms (e.g. iPhone, Salesforce, Shopify, etc.)."

"Second, there is no underlying product here other than the Ether token."

"These two factors have led to a wave of new Layer 1 blockchain protocols, beyond Bitcoin and Ethereum — from Cardano to Solana and more."

"Think of this as another network effect (interaction network) layered on top of the platform — the more smart contracts there are for a protocol, the easier it is for developers to build new ones. However, cross-chain composability, i.e. composability of smart contracts built across protocols, is likely to dilute its impact on defensibility."

"However, it is difficult to classify their network effects because their value and utility are still unclear. This isn’t unusual at the earliest stages of a technology cycle — experimentation and evangelism always precede utility."

"As you can see in the animation above, Axie Infinity combines four different network effects:"

"The Axie Marketplace is the second layer of Axie Infinity’s network effects."

"The resulting marketplace network effect reinforces the interaction network effects of the game."

"Thanks to the differentiated nature of its supply, Axie Infinity’s marketplace component is highly defensible, i.e. the Axie Marketplace is likely to remain the go-to destination for buying in-game items."

"My fellow Atomico Angel Sarah Drinkwater describes DAOs as “group chats with a shared goal and money” — which is an accurate summary."

"This is yet another form of interaction network, where identity influences defensibility. In this case, the importance of user identity depends on the scale of the network."

"In order to play Axie Infinity, players need to purchase 3 Axies from the Axie marketplace — the cheapest of which costs ~$200."

"To lower this entry barrier, “scholarship” programs have cropped up that “rent” Axies to aspiring players."

"The first possibility is that true network effects are no longer a meaningful source of structural defensibility in this era."

"Instead, defensibility will depend on the tribal and psychological switching costs of each project community."

"The second possibility is that we are still too early in the web3 cycle for sustainable defensibility to emerge — similar to Yahoo during the early years of web 1.0 and Myspace in web 2.0."

"In other words, most projects are still experimenting with the capabilities of web3."

"And the long-term winners — with stronger and more defensible network effects — will only emerge after this phase of experimentation."

"The Interaction: How do users interact with others?"

"The Network Effect: Does the addition of a user increase value for all users?"

"Scalability: In what way does every new user affect value? Are there any limitations?"

"Defensibility: How does this change as adoption grows?"

"These questions will be crucial to evaluate the potential of web3 projects. Those that combine the capabilities of web3 with stronger network effect layers could turn out to be the biggest winners of this era."

"In fact, combining multiple forms of network effects, like Slack, Carta, and Poshmark have done, is one of the most effective ways to strengthen both defensibility and scalability."

"Crucially, these network effect models can be combined or layered on top of each other. Doing this can mitigate existing weaknesses and strengthen scalability and defensibility even further."

"There are two broad approaches that entrepreneurs can use to layer network effects:"

"Add new types of network participants"

"The first and simplest way to layer network effects is to introduce a new type of network participant and connect them to existing users."

"Slack is a 1-sided SaaS-enabled network that connects users within an organization. In 2015, Slack added a new network effect layer to its product by creating a developer program and app directory."

"This allowed third-party developers to create integrations for Slack users and layered a user-developer network effect (platform) on top of the existing user-user network effect (interaction network)."

"This not only enhanced Slack’s value proposition for customers but also increased switching costs."

"Adding a developer program is the most basic way of introducing a new network effect to a business model, and one that has been used by a range of B2B companies."

"A few years later, Amazon layered a marketplace on top of this data network by adding third-party sellers. This form of layering helped Amazon create a brand new business and also insulate itself from the downsides of data network effects."

"In order to introduce new network effects, added network participants need to take on roles that did not exist before and enable new types of interactions (or transactions)."

"Another facet to keep in mind is that companies can add multiple types of participants to successively layer new network effects. Shopify is a great example here, as it has layered two successive forms of network effects on top of their core product — by adding two different types of participants."

"This approach is fairly simple to conceptualize but can be challenging to execute as it involves acquiring new participants (which can be expensive) and integrating them into the product."

"Create new types of connections between existing network participants"

"AngelList is another example of this approach. It began as an interaction network connecting investors and entrepreneurs. Then in 2013, AngelList launched syndicates that connected those investors and entrepreneurs in a marketplace to raise funds (in addition to providing a SaaS workflow for managing the process)."

"Creating new connections between existing network participants is more complex than simply adding new network participants because it requires a deeper understanding of customer behavior. However, it is also easier to execute because it does not involve acquiring and onboarding new participants. Rather, it purely relies on product development and effective communication of the value."

"It is also worth noting that these two approaches — adding new types of network participants and creating new types of connections between existing network participants — can be combined. Poshmark is a perfect example of this."

"Poshmark began its journey as a social, C2C commerce app that allowed users to buy and sell their clothes."

"it also added a social layer by allowing users to communicate and follow each other on a social feed. In other words, it connected all users on an interaction network."

"In 2015, Poshmark launched a wholesale portal that allowed sellers to buy clothes directly from brands to list on their Poshmark boutiques. In other words, it added brands and layered a new B2C(2C) marketplace on top of their product."

"Poshmark created an incredibly sticky experience by combining both a C2C marketplace and a 1:many interaction network."

"Users did not just come to Poshmark to buy or sell clothes, but also to discover new brands and styling ideas."

"The identity of influencers and users on Poshmark became a core part of the experience which was very difficult to replicate."

"Often, incumbent networks are disrupted in the same way other types of businesses are. Startups create new value propositions that initially target novel, low-value markets, and gradually encroach on the incumbent’s market."

"Startups following this approach have to contend with a near-term risk, i.e. the risk that there is no market for their product."

"In addition to this, there is one more way of disrupting network effects — by making the incumbent’s network redundant and creating a better customer experience in the process."

"Networks typically need to introduce some element of friction to create a network effect."

"For example, users of an interaction network like Skype can only call other Skype users. In a marketplace, users can only transact by interacting with a seller present on the marketplace."

"Eliminating these points of friction can dramatically improve the customer experience but at the cost of sacrificing network effects."

"Since the incumbent has already validated the market, this approach trades the near-term risk associated with establishing product-market fit for the long-term risk associated with defensibility."

"Startups following this approach need to have the foresight to identify this risk and address it in other ways."

"Zoom attacked Skype by turning its competitive advantage — its network — into a weakness. It accomplished this by allowing its users to call anyone, even if they did not have a Zoom account."

"The combination of superior user experience and the virality embedded into the product helped Zoom rapidly overtake Skype."

"But even in the absence of those failures, Skype could only compete with Zoom by abandoning what made it successful in the first place — its closed network."

"It was only in 2018 (7 years after it was founded) that Zoom attempted to address this weakness by layering on an app marketplace — the hope was that it would introduce new barriers to switching."

"But by this time, video calling was largely considered a utility, and bolting on integrations did not create a deep enough moat. This gave Google Meet and Microsoft Teams (which is replacing Skype) an opening to clone Zoom and catch up."

"Zillow was one of the first companies to introduce network effects to a complex, high friction industry — real estate. The company aggregated property listings from agents and connected them to prospective buyers or renters."

"Like Zoom did to Skype, Opendoor turned Zillow’s competitive advantage — the network effect between agents and buyers — into a weakness."

"It did this by pioneering the iBuyer model — it purchased homes from sellers, refurbished them, and sold them to buyers."

"Its “home flipping” model was far more capital intensive and had much lower margins than Zillow’s marketplace business. This made it difficult for Zillow to respond when Opendoor was starting out."

"Writers were compensated based on how paid Medium subscribers engaged with their (paywalled) articles."

"Substack turned Medium’s model on its head by making its network redundant."

"Substack has grown rapidly since it was founded in 2017, but Medium has had trouble responding to the threat. This is because cloning Substack’s product experience would involve dismantling, or at least disintermediating, its primary competitive advantage — its network of writers and readers."

"While Substack had to abandon network effects to challenge Medium, it had another effective form of defensibility baked in from the start — switching costs."

"Writers would lose subscribers if they abandon Substack. This is important because a writer’s relationship with subscribers is stronger than with Medium followers. And it is doubly important when they are paid subscribers."

"disruption could even be caused by weakening a network effect instead of abandoning it. For example, TikTok’s decision to abandon the follower graph and distribute content algorithmically helped it scale despite the presence of Facebook, Instagram, and Snapchat. However, it did so at the expense of defensibility which allowed copycats like Triller to gain traction."

"Making networks redundant is a valid approach to scale a startup when competing against incumbents with strong network effects."

"it is only one half of a successful business model."

"As you scale, you need to build in some form of defensibility to sustain your position. In the absence of this, you risk losing the market back to the incumbent or to other copycats."

"With this in mind, you need a roadmap to regain what you have sacrificed as you scale. Building strong switching costs, like Substack, is one way of doing this. Layering new network effects on top — as early as possible — is another avenue."