Network effect is defined as the phenomenon wherein the value of a product or service increases corresponding to the number of consumers of that product. In essence, the network effect creates a positive feedback loop that make the network more valuable as more users join the network. This often creates “winner takes all” dynamics in the market; but even without that, network effects help strengthen the product/service as the network gains traction.
Network effects have been instrumental in creating value even before the start of the Internet era. Network effects were amongst one of the four or five “moats” highlighted by Warren Buffet and later systematically validated by Morningstar. In the Internet era, the importance of network effects has only gone up. NfX, one of the US-based investment funds, analyzed companies that had scaled beyond $1 Billion valuation (in 1994–2017 period) and attributed 68% of value creation to network effects. (link)
However, it turns out that network effects did not provide the defensibility they were supposed to provide. Consider a few examples: Uber has been unable to benefit from network effects to the extent its earliest supporters had imagined; Apple Music wasn’t able to defend its turf against Spotify and Pandora; PinDuoDuo was able to compete with established players such as Taobao, JD.com, VIP.com, and T-Mall to become the second largest e-commerce platform in China in merely three years. Similarly, network effects didn’t prevent Facebook from beating MySpace and, subsequently, didn’t help Facebook to defend its core usage (i.e., communications) against upstarts like by WhatsApp and Instagram. In fact, there are examples where startups without network effects were able to compete with incumbents with large network effects and carve out a niche for themselves. For example, we have seen rapid growth of Lime / Bird (in US) and Bounce / Vogo (in India) despite the presence of larger mobility players such as Uber, Lyft, and Ola.
So, one is forced to ask: what are the fundamental characteristics of network effects? What defines relative strengths of network effects? Is it possible to increase the strength of network effects to defend against competitors? These questions are important to understand if/how new technologies (such as crypto-assets, IoT and smart cities, ML & AI, etc.) will be able to disrupt incumbents and create new value.
In the subsequent sections, we present an overview of a new theory underlying network effects. Existing approaches (such as Berkshire Hathaway, Morningstar, A16Z, NfX, etc.) are facing problems because they have explored network effects from investor-centric (for capital allocation) or company-centric (for building defensibility) perspective. This is exemplified by “winner takes all” phrase that is most popularly associated with network effects.
As opposed to these, we present a customer-centric approach to define and understand network effects. Using this, we outline a theory for building network effects in a holistic and systematic manner. Our customer-centric framework and network effect theory solves the problems and challenges faced by earlier attempts.
A customer’s journey with any product has three phases: Invite, Engage, and Connect. This is true for any kind of product — digital (such as consumer and SaaS products for businesses) as well as non-digital (such as cars, apparel, fashion, entertainment, etc.).
In the Invite phase, customer explores the product to evaluate if value proposition offered by the product resonates with them. In the next phase, customers Engage with the product (before and after conversion/purchase) to derive value from the product. If the product delivers the promised value, customers start developing an association with the product — which, after repeated and consistent value delivery by the product, evolves into emotional Connect with the product (that often transcends the functional attributes of the product).
The company building the product or service has to cater to the three phases of customer journey distinctly. During the invite phase, the company needs to trigger initial positive outlook towards the product amongst the right sets of users and ensure that the users have good initial experience with the product. During the engage phase, company needs to ensure that users have a good experience (both before and after value delivery). Finally, during the connect phase, company needs to proactively work towards establishing emotional connect with the users.
The figure above shows customer journey as a loop with Invite phase leading to Engage phase and Engage phase leading to Connect phase. The figure also shows Connect phase leading back to Invite phase. Why should this be the case? This is because happy and emotionally connected users inevitably talk about a newly discovered product/service with their family members, friends and colleagues. What are the implications of Connectleading back to Invite? Thanks to social networks and influence networks (social media, i.e.), the word-of-mouth buzz generated by Connected users spreads wider and persists longer. These digital footprints left by users, therefore, helps to improve the Invite for the next set of users.
In fact, since every iteration helps improve subsequent iterations, customer journey is not merely a loop, it is actually a “spiral” that keeps improving and growing bigger as the products goes through Invite — Engage — Connect phases with every customer.
The spiral nature of customer journey gives rise to network effects. The network effects, in fact, are present by default; companies only need to ensure that they are able to amplify them and enhance them by taking customer-centric perspective.
Indirect Network Effects
In addition to the word-of-mouth buzz created by happy and connected users, companies can also collect valuable data and metadata related to customer’s preferences and usage patterns across different stages of the spiral. These signals are useful to improve user experience for the current as well as future sets of users. We refer to these network effects, which arise due to indirect involvement of users, as “Indirect Network Effects”.
As an example, usage metadata from the past and current users can be used to invite users more efficiently and to onboard interested with better users experience (by providing visitors and prospects with more relevant content and user experience). Moreover, companies can improve engagement with prospects/customers by using metadata from the past usage and engagement patterns to build ever-improving personalization or recommendation engines. As you can imagine, these are weak forms of (indirect) network effects.
A slightly stronger form of indirect network effect arises when the companies are able to leverage data itself (in addition to the metadata) to improve customer’s journey through different phases. An example of this is ability of the lending companies to improve their credit risk models based on the behaviour of the past and current users (so that credit can be made available to more users).
The above correspond to different forms of platform-based indirect network effects. A variant of these are process-based indirect network effects: accumulated learnings within organizations often translates into new and better processes that the companies are able to either deliver better services or build better products. Example of the former is Infosys’ “global delivery model” (which involved codification of delivering value via a combination of on-site and off-site value delivery teams) while an example of the latter is Toyota Production System (which was codification and productization of Kaizen approach of continuous improvement). (link) Another process-based indirect network effects is the co-creation process that can be used by companies to build better products by direct customer involvement in the product definition as well as product iterations.
The above two, in turn, enable companies to attract better talent and build stronger teams (e.g., Pixar with the animation talent) and, thereby, continuously improve the products/services provided to customers. (link)
Direct Network Effects
In order to unlock stronger network effects, companies have to build “direct” network effects. Direct network effects can be unlocked by enhancing the three phases of customer journey with one important addition: direct user involvement. If products can get users directly involved during Invite, Engage, or Connect phases, customer journey spiral continues to grow in a self-sustaining manner.
What does “direct user involvement” correspond to? Direct user involvement requires that customers participate or contribute in the customer journey beyond the tasks and activities they do to derive value for themselves. For example, users perform a number of activities for buying books from e-commerce retailers (such as search, comparison and selection, payment, etc.). These are needed to purchase the book and, therefore, for buyers to derive value for themselves. In addition, the buyers might get directly involved if they can be encouraged to attract new users, to provide review about the book for the benefit of subsequent buyers, or to form groups in order to help each other to make purchases together.
We represent user’s involvement during the three phases of customer journey separately as the “Involve” phase:
Note that as the number of users increase, user involvement is likely to increase automatically. This is the lure and the strength of the direct network effects: they promise ever-improving product (and, therefore, ever-improving customer experience)! Better still, network effects become more active and stronger with the addition of increasing number of users.
Direct users involvement during the three phases of customer journey results in three different kinds of network effects:
1. Direct user involvement in Invite gives rise to “Viral Networks”
2. Direct user involvement in Engage gives rise to “Exchange Networks”
3. Direct user involvement in Connect gives rise to “Connected Networks”
Let’s look at them each of these in more depth so that we can understand their main characteristics.
Viral networks are built when the current users invite new users to join the network. There are two types of viral networks: (1) acquisition-based viral loops and (2) engagement-based viral networks.
Acquisition-based viral networks are the online incarnations of the traditional word-of-mouth buzz creation. New users join the network because they hear about it from someone they know and/or can trust. Acquisition-based viral loop with some form of incentives is popularly known as viral marketing.
In the engagement-based viral networks, on the other hand, current users invite new users to join the network because it improves the engagement experience for the current users.
There has been some debate whether acquisition-based viral networks (or viral marketing) are network effects or not. We prefer to include these as the weakest form of direct network effects because, often, companies pass on the reduced customer acquisition costs back to customers (for example, in terms of lower subscription costs).
Engagement-based viral networks are stronger than the acquisition-based viral networks because they directly help increase the value extracted by the users. Invision is a good example of this because designers invite other team members (such as product managers and engineers) to collaborate with them in a more efficient manner.
Exchange networks are built when the current users engage with each other to improve the experience for everyone. There are three types of exchange networks: (1) marketplaces & market networks, (2) platform-based networks (including metadata networks and SaaS-enable Marketplaces — SeMs), and (3) platforms with same-side network effects (including content & data networks).
Marketplaces and market networks are the most common form of network effects — they involve direct value exchange amongst the network participants. Typically these are online incarnations of the networks that already exist in the real world.
Platforms (which can be software-based on hardware-based) enable reimagining how interactions and value exchange between different stakeholders can be facilitated. Typically, platforms promise lower friction, higher reach, more variety, etc. to the participants. In addition to providing value to the participants, platform-based networks enable more efficient value exchange between different sides of the network.
SeMs are a variant of these — where one side of the marketplace (typically, sellers or service providers) derive value from a SaaS offering, which helps to build one side of the marketplace. After securing one side of the marketplace, SeMs unlock higher value with the help of network effect dynamics.
Metadata networks are a weak form of platform-based networks: participants are able to derive value merely by virtue of their participation in the network.
Platforms with same-side network effects rely on information exchange amongst the participants and, typically, are powered by the pay-it-forward outlook. These are typically secondary networks because the information exchanged is related to the value exchanged in the primary network (which can be a marketplace itself). As an example, consider TripAdvisor that enables travelers to share their experiences with others (which improves everyone’s experience); the primary value exchange is between the hotels and travelers. Data and content networks are a prototypical example of networks that complement the primary networks.
Connected networks are built when the current users help to build and deepen emotional connect for everyone. There are three types of connected networks: (1) social & collaboration networks, (2) community-based networks, and (3) marketplaces with collaboration & same-side network effects.
Social networks are, typically, online incarnations of the networks that already exist in the real world. Online variants increase frequency of usage and, therefore, help increase emotional connect amongst the participants.
Community-based networks facilitate interactions that enable users to share their experiences, insights, knowledge, etc. with other (same-side network) users without any explicit expectation in return. In some sense, this is a natural progression from the “platforms with same-side network effects” in the altruistic dimension. User involvement in these cases is over-and-above the transactional value exchange; by caring and sharing with each other, users form a community. This is triggered by user’s desire to develop an identity and associate with a larger purpose in order to develop a feeling of belongingness.
AirBnB is a celebrated recent example of a company that has consciously developed a community by understanding the traveler’s needs as well as wants. While traveler’s needs correspond to standardized accommodation at affordable prices, their wants correspond to their desire to understand different cultures and to appreciate the oneness of the humanity (and, therefore, the desire to stay with locals and to get first-hand feel of the local culture) while, at the same time, to feel special (and, therefore, the desire for unique experiences).
It is useful to leverage the most active, engaged and experienced users (“super users”) — to build and maintain the community. Companies can encourage super users to take co-ownership of the community and to participate directly — both by contributing directly as well as by monitoring and regulating user behaviour to be consistent with the desired objectives.
Marketplaces with same-side network effects are the ones where users collaborate with each other to complete a task. When current users collaborate with new users to derive more value from the network, it deepens the emotional connect amongst the users. Likewise, in a business setting, collaboration networks amongst members helps to increase value amongst the network participants.
Network Effect Strengths
Indirect networks are the weakest form of network effects because the benefits accrue slowly and asymptotically. Viral networks, Exchange networks, and Connected networks are progressively stronger forms of network effects. Note that these three correspond to the three stages of customer journey — each stage with progressively deeper association between users and the product/company. User involvement also, as a result, progressively becomes deeper across these three stages — resulting in increasingly strong network effects.
The four types of network effects along with their relative strengths are shown below (with Level 4 being the weakest and Level 1 being the strongest):
Is it possible to quantify the relative strengths of the different types of networks? Based on our initial analysis, the strength of viral networks seems to be proportional to the number of users (who can each invite a constant number of more users from amongst their network of influence). The strength of exchange network seems to be proportional to the square of the number of users because each contribution made by a user can benefit all the existing users. And, finally, the strength of connected networks seems to be proportional to the exponential power of the number of users because each user can not only impact/influence all users but also create newer circles of influence for stronger connection amongst themselves.
In other words, viral networks, exchange networks, and connected networks correspond to the three well-known “laws” of networks: Sarnoff Law, Metcalfe Law and Reed Law, respectively. These three are shown below to illustrate the similarity with the mechanics of the three types of network effects (as explained above).
Indirect network effects grow at slower than the growth of users themselves — whether this is log (n) or s-curve (or something else) depends on how the company is able to leverage data/metadata to improve user experience. In general also, the above should be taken as a first-cut guidelines; the exact benefits of each of type of network effect has to measured explicitly based on the actual Invite/Engage/Connect dynamics based on direct user involvement.
There’s another popular network effect metric known as k-factor that is used to describe the growth rate of a product/service. It depends on the number of invites sent out by each user and the percentage conversion of each invite. As it can be seen, the k-factor (and its derivatives such as Reforge’s growth multiplier; link) only captures the benefits of the viral networks and, therefore, is an incomplete and insufficient network effect metric.
The relative strengths of different network effect types gives a clue towards why PinDuoDuo’s “connected network” was able to beat JD.com’s and VIP.com’s “exchange network” and how Uber’s “exchange network” can be strengthened by upgrading to “connected network”. It also explains the importance of Spotify’s Collaborative Playlists and the rumored “social listening” feature (link) that allows friends to listen to shared music in real-time — these might help build “connected network” and pull ahead of “exchange networks” supported by Apple Music, YouTube Music, etc.
However, this doesn’t explain Facebook’s ability to beat MySpace? It doesn’t explain why WhatsApp, Instagram, and Snapchat have been able to build better network than Facebook either. Or, the more interesting question: how are Bird, Bounce, Lime, and Vogo (with no network effects) able to rapidly carve out niche for themselves despite competing with incumbents with large networks effects?
To understand these, we have to look at two fundamental parameters that influence all human activities:
These two parameters are used to define an “Engagement Graph”, which help companies exploring solutions for a specific activity/use-case to figure out how they can create value more naturally (by establishing scale, inculcating habit, or building brand). Read more about these here.
Besides helping identify the strategy to create value, Engagement Graph also succinctly highlights two important observations:
Higher the activity frequency, higher the defensibility
Higher the activity importance, higher the ability to extract value
The first observation mentions that a product/brand offering solution for lower frequency activity will find it difficult to defend against product/brand being used for higher frequency activity.
The second observation mentions that a product/brand offering solution for higher importance activity (often in conjunction with frequency of usage) will find it easier to extract more value from its customers. Of course, monetization depends on several other factors; we explore these in more depth in a separate article.
The first observation is more relevant for our discussion here: the strength of different networks (and, for that matter, any product) depends on activity frequency as well. A startup with network-effect product with higher usage frequency will be able to compete against an established network-effect product.
MySpace’s primary use-case was enabling music & entertainment artists to communicate and connect their fans and, therefore, had built a form of Exchange network. Facebook’s primary use-case was communication amongst friends — esp. college students and young adults. Facebook, therefore, had same-side network effects and built a form of Connected network. Moreover, Facebook had higher frequency of usage as well as higher importance vis-à-vis MySpace. These two reasons helped Facebook build stronger network effects.
And, how do we compare Facebook with WhatsApp, Instagram, and Snapchat? All of these targeted similar groups of users — though Instagram and Snapchat went after slightly younger audiences initially. However, it turned out that the use-case and audiences that WhatsApp, et al went after that higher frequency of usage. WhatsApp also leveraged then-newly-launched push notifications functionality to drive higher usage. It is also not a surprise that WhatsApp grew the fastest amongst societies (such as Brazil, India, Mexico, and Russia) with the strongest familial bonds (link). These factored contributed towards unprecedented growth of these networks at the expense of the Facebook.
Higher frequency of usage is also the reason for Bird, Bounce, Lime, and Vogo (with no network effects) being able to grow rapidly despite competing with incumbents with large networks effects.
Some of the initial attempts to delve into network effects were focused on understanding and defining the network effects. More recent efforts have yielded a “typology” — for example, NfX provides a categorization schema and enumerates 13 different types of network effects (link). Niskanen Center has a better (though still incomplete) categorization schema wherein they provide a taxonomy of five types of network effects based on the strengths of the “direct” network effects and “indirect’ network effects (link). And, of course, there are researchers and investors who have their variant of “ideology” — a belief system that helps them define a high-level strategy for identifying and leveraging network effects.
However, there is no “theory” that hypothesizes how and why the network effects emerge. They don’t explain why different types of network effects have different strengths. This has lead to significant confusion about differences between network effects and “viral marketing’, “economies of scale”, “brand”, etc.
In this article, we have defined not only a theory underlying the network effects but have also presented an outline of a framework that provides a systematic and step-by-step process of building and strengthening network effects.
We have explained that different types of network effects provide different benefits: viral networks help increase scale; exchange networks help increase engagement and stickiness; and connected networks help strengthen the brand. Finely crafted and activated network effects become stronger on an ongoing basis and help the company to pull ahead from their competitors based on the following:
Viral networks: when current users help to invite and activate new users, it broadens the gap between the company and its competitors and makes it difficult to bridge the ever-growing scale gap.
Exchange networks: ongoing value exchange amongst the company’s users increases the switching costs for current users and, therefore, makes it tougher for competitors to lure customers away from the company.
Connected networks: deeper emotional connect with the brand makes it difficult for competitors to lure core customers away from the brand.
Historically, it has been assumed that network effects are dependent on the product’s category in the sense that a company can leverage network effects if and only if company’s category intrinsically is dependent on marketplace dynamics, community-based interactions, etc. By identifying different types of network effects and their dependence on different types of direct user involvement, we have highlighted how companies — even those that don’t have the natural marketplace mechanics or social networking dynamics — can thoughtfully and systematically craft various types of network effects into their products/services.