Demystifying Network Effects
This post is brought to you by the team at platformdesigntoolkit.com and is dispatched in our fortnightly newsletter.
Premise
This recap on network effects has been inspired and is heavily based on several important pieces of work that the reader should look up to go more in deep with the topic. Some of the illustrations in this piece are also taken from these pieces of work. In particular:
- “The Network Effects Bible” by James Currier & the NFX team.
Available at: https://www.nfx.com/post/network-effects-bible - “All about Network Effects “ by Anu Hariharan at Andreessen Horowitz. Available at: https://a16z.com/2016/03/07/all-about-network-effects/
- Max Olson’s “Advantage Flywheels” on FutureBlind. Available at: https://futureblind.com/2019/08/03/advantage-flywheels/
- Ben Thompson‘s “The Moat Map”. Available at: https://stratechery.com/2018/the-moat-map/
“Platforms are eating the world” some would say: we are transitioning into a new paradigm where value gets created and distributed among participants outside the traditional enterprise. Entrepreneurs, venture capitalists, incumbents, consultants, designers: everyone is trying to clarify concepts and extract learnings from successful platforms to apply them to their context.
A lot has been written in the last 10 years but, as often happens with information overload, things get fuzzy and a good cheat-sheet can be of great help to cut the clutter and get the essentials. In this blog post, we intend to bring in a little clarity around one of the most important aspects of platform thinking, achieving network effects; as well as to share a basic framework to help strategize and activate the mechanisms to achieve better network effects.
As pointed out by James Currier in the NFX Bible, data indicates that network effects (NFXs often in the rest of this post) are responsible for 70% of the value created in technology since 1994. This also shows how NFXs offer superior defensibility against incumbents and new entrants. Figuring out how to identify, distinguish and activate NFXs has become an important topic for investors and is one of the reasons entrepreneurs keep awake at night. Entrepreneurs concerned with social impact, or even platform-cooperatives cannot afford to ignore how NFXs work: for better or worse, achieving NFXs is vital for any platform-organization. In other words, if you have any stake in creating or running a platform you should be trying to understand the NFXs that you’re subject to. You should also be looking to distinguish network effects from counterfeits and how to activate them and make them more robust.
This post has been co-written by Esme González Pillado and Simone Cicero with a special thanks for contributions and reviews to Manfredi Sassoli de Bianchi, and Luca Ruggeri.
Network Effects
Technology, automation, access to information and connectivity are driving the shift of the revolution we are living in. The way we produce value and our expectations about growth have radically shifted. Successful companies of the industrial era were designed to produce a solution for a customer, reduce costs of manufacturing and distribution, and push marketing messages for sales: in other words, achieve economies of scale at any cost. In contrast, aggregators or platforms are connecting and empowering entities in their reference ecosystem in a space specifically designed for interaction, interconnectedness, and exchange of value. Aggregators, as they grow the size of the network, they become increasingly more attractive and valuable for participants. One key driver of this increase in value perceived depends on the increasing (with network size) possibility for new entrants to find the “right” niche answer to their needs: the more accommodations are available on booking.com, the easier for a traveler is to find the right accommodation, in terms of location in the city, standing, style, accessibility, etc.
Network Effects vs. Virality
What is the difference between let’s say the growth of OpenTable and the growth of the ubiquitous toy, Fidget Spinner, that sold about 19 million in 2017? They both experienced a tipping point and a noticeable exponential growth. How can I be sure that the growth of my strategy, startup or ecosystem weaving effort is indeed a result of network effects and it is not just a transient fad? How can I recognize if I NFXs from virality?
Essentially:
- network effects are about retention, monopoly, hacks, and defensibility while viral growth is about the speed of adoption;
- network effects become more valuable as more people use it while viral growth may produce a cost-benefit through economies of scale but not necessarily making the product or service more valuable to the people subscribing to it;
- as the network/platform grows we see more and better engagement, user rates and more customer lifetime value while viral growth there is no correlation that more use leads to more customer lifetime value;
- as the network/platform scales, network effects produce better margins due to a reduction in customer acquisition cost with a trend leading to zero and activating pull. On the other hand viral growth, there is still an expense for customer acquisition and as it occurs, often there is a limit for virality as certain products or services tend to lose its attractiveness after a period of success over a period of time.
The “Laws” of Network Effects and the Struggle between Abstractions and Reality
Several researchers have been studying this phenomenon within time. In particular Sarnoff, Metcalfe and Reed’s work gave us clarity on how network value grows in relationship with the network size and shape. The so-called Network Effects Laws, are not immutable and are not necessarily present as such in our complex reality, especially when talking about far-reaching digital networks, which are often multi-sided with nuggets of influence of asymmetric value where often participants act as a supply and other times as demand-side (think of an Airbnb host who sometimes is a guest).
These “Laws” are great at helping us picture how the different types of networks create value; but — in reality — networks are much messier, asymmetric and chaotic. It is hard to find networks that have such perfect forms and symmetric relations among participants: therefore the laws should not be taken as immutable frames but more as ways to identify how in a certain part of the network is behaving especially since very often more than one “law” is present in a network.
Furthermore, besides the key driver of the network effect, an important factor to consider is the growing depth and the available choice of the offering and the way the entities are connected. How they influence, contribute and reinforce the value growth with size and participation.
We could call these further value drivers “reinforcing” mechanisms. As an example, the more data is available on a network, the more the platform can train an AI system to be able to become prescriptive and help the participants navigate the complexity of the network or execute particular strategic actions. As another example, the more a solution becomes embedded into the business process of many players, the harder to switch to another solution for participants. Also, network growth can as well produce traditional economies of scale for the owner of the platform and therefore generate savings for participants as the network grows.
This improved perception of value is key as it makes the cost of customer acquisition for platform organizations (that are network effects based), decrease over time, in comparison to traditional products.
Here above we present some typical reinforcing effects — taken from the excellent blog post from Max Olson cited in the opening — to explain some of these reinforcing effects through the illustration of flywheels, and interesting and a truly useful way to picture network effects as a virtual cycle that reinforces itself.
The original post has very interesting illustrations of Amazon’s and Google’s network effects and reinforcing mechanisms, and we share here the example of Airbnb.
As one can see from figure 1, all starts with a 2-sided network effect. Then suddenly the brand value proposition perception increases as a reinforcing mechanism (figure 2). In figure 3 we highlight how the technology becomes better thanks to increasing access to data, and finally, in figure 4 we explain how the existing demand allows Airbnb to introduce an integrative business model, experience hosting.
Critical Mass & the Chicken-and-Egg Problem
Whoever has been dealing with a network effect dependent venture must know what does it mean to aim for critical mass to be reached. Often it means lots of cash in investments until reaching it or painfully giving up. Therefore understanding what it is and what is involved can be of use.
Critical Mass is often described as the number of participants or size of the network needed to allow the platform itself to auto-generate its own growth — essentially by having a value perception that grows faster than the growth of the network — . Critical Mass can be understood as a tipping point or threshold where a notorious change in the trajectory of a growth curve occurs and a significant increment in the value of the network happens. We’ve been covering the concept of achieving liquidity (a concept overlapped with critical mass) on two previous posts, and we definitely suggest the reader give them a read in parallel with this recap we’re offering today.
For early-stage networks, reaching this point becomes existential, especially for two-sided marketplaces where is important to attack both sides of the network. Often referred to as the Chicken-and-Egg problem is the endless chase to find out how to kick-start the flywheel when you need multiple sides of the ecosystem to join and interact through the platform. To give a quick example, OpenTable found out that they needed to have about 25 restaurants in any given city to be attractive enough for guests and use the booking service, while for Airbnb, the quasi-magic number was 300 homes per city.
Network Properties
To understand which mechanisms will impact positively the growth of our network-platform and will facilitate reaching critical mass, we need to know deeply how the nature and behavior of the participants interrelate with the Value Proposition perception of the Platform. In particular, there are five essential aspects we’d like to focus on.
One thing that we need to keep in mind is that often when we refer to these properties, it’s easier to explain them — and most likely to understand them — in relation to a two-sided marketplace-platform while, in reality, opportunities for development of platform strategies increasingly relate with more complex multi-sided systems where multiple two-sided relationships overlap. Let’s keep this in mind when analyzing our system.
Commoditized or Differentiated Supply
How easy is it to replicate the offering among the suppliers? This is a key question to ask. As an example while Airbnb hosts offer places to stay that are partially unique (at least in terms of position, if not vibe or host reputation) — and this is increasing with experience hosting — Uber drivers are easily replicable since the value one gets is that of getting from A to B. One could reflect on the infamous quote from Travis Kalanick thinking about replacing “the dude in the car” with an AI. Having a commoditized supply normally creates a so-called “asymptotic” curve in network effects: in a commoditized supply marketplace, the value reaches a plateau at some point even when adding more depth to the offering. If the offering is commoditized how much you can improve while still adding value? Once you have a car coming up in 3 minutes, having it in 2 doesn’t really change your life.
Symmetry or Asymmetry of the Supply / Demand
Most of the networks have asymmetric weights of their supply and demand Asymmetry can offer a good hint in choosing what side one should attract first (we’ll talk about this later in more details): as an example, given that one expert content provider can attract many thousands of learners, online learning platform Masterclass is choosing to manually curate the list of first-class teachers (“learn from the best”). Similarly, when Apple launched the iPhone, the company self-developed a number of key applications — knowing that there was no real economic potential for mobile app development — to later open access to developers when the number of users was substantially grown. Essentially Apple was subsidizing the supply slide.
Normally the asymmetry of the capability of producers to serve also impacts the behavior of network effects in a way to anticipate the plateauing and, on the other hand, anticipates the achievement of the liquidity: it’s easy to explain why on marketplaces asymmetric on the supply side (with the supply side able to serve a disproportionate amount of demand) shapers tend to attract supply first. A powerful way to attract supply-side — or in general the most important of the sides — may be that of providing them with so-called single-user-value: a value proposition that doesn’t necessarily need the network effect to exist. As an example, despite in the very early days, OpenTable traction was small, the free tools to manage the bookings were convincing enough for the restaurants to join the platform.
Flexibility of Location: Locally or Globally Bound
The supply and demand of a network are also more or less attached to a certain location. Sometimes networks are highly local, other times demand (or supply) can consume or serve a more larger network. Let’s take the example of the French platform “The Food Assembly / La Rouche qui dit Oui”: in this case, the network of food producers, food buyers and community organizers is heavily local (basically city-region bound) making this a disconnected local network. In the case of Airbnb instead, the traveler can travel and “spillover” the value from one city to another: a traveler can discover the possibility to be a host and effectively trigger internal growth. In other cases, let’s think for example about gig worker platform Fiverr, most of the work can be performed online, making the network effectively global.
Quite intuitively the more a network is locally bounded, the more it will be hard to achieve liquidity — if one looks on a global scale. More precisely, the concept of liquidity in a heavily local network must probably be reimagined as a sum of different liquidity/critical mass events at a local level. The question will be that of understanding what and when the value from a (local) sub-network can spill over the larger network of networks.
When you have a local network — especially with interconnected local networks — a particular strategy called “bowling pin” may come handy in achieving liquidity. This strategy consists of launching different cohort-networks gradually and extending to the others “like a bowling strike that gets all the pins”. It’s common for geographically bounded strategies: Airbnb launched city by city, Facebook started from Harvard. A bowling pin strategy can also work for tribal differences (e.g.: from rock music to all music).
Single Tenancy or Multi-tenancy
Multi-tenancy is when participants from the supply and/or the demand side may tend to juggle on multiple platforms to get out most of the value. Uber drivers are free to use two or more (Uber/Lyft/Juno) platforms that offer a similar service to extract as much value as possible. Airbnb hosts can — as well — list their properties on Booking and other platforms. This happens more often if the offering is commoditized. On the other hand, sometimes artificial constraints exist to prevent such behavior. This often happens when a public, regulated service is “platformized” by institutional players. As an example, Taxi drivers that are part of Rome’s biggest Taxi cooperative can only use an official platform and can’t, for example, join the FreeNow platform. Supporting Multi-tenancy often helps achieving liquidity faster: as suppliers are not “bounded” to one single platform they’re, on one hand, less concerned about joining yours as the opportunity-cost ratio may be higher and, on the other hand, if multi-tenancy is the industry standard suppliers do not disappear from availability once they join one network.
Transaction Frequency and Lifetime
Another key property of networks is the frequency of the transactions. In multi-sided systems, transaction frequency might vary — depending on what relationship we’re focusing on and to what sides of the system. As an example there’s a relevant difference in frequency of interactions between a real estate agency and a buyer in a real estate network: while from the point of view of the buyer the transaction is — often — very sparse in a lifetime, the real estate agency aims at increasing the transaction frequency as this represent its key value generator. When we look into a multi-sided network or — evermore — in market-network as James Currier once defined them, we may want to look at liquidity and critical mass as achievable separately — or relatively separately — looking at the different relationships. If one looks into the supply side, a supply-side that works on high-frequency transactions remain available in the network after a transaction is performed, in case our network instead is based on low-frequency transactions — on both sides — it may be challenging to achieve liquidity since a continuous outflow of entities would be hard to contrast with pull. The reader should vary of trying to organized ultra low-frequency markets — where both sides are transacting rarely — as this may be a signal of such a high transaction value that may make the case for a platform organized marketplace (that is promised on reducing transaction cost) not so applicable.
In the following picture, we give the reader a quick and — rather generic — highlight of how the properties we mentioned potentially impact the network effect “function”. As anticipated, a heavy local network and single tenancy may slow down acquiring liquidity while asymmetry (towards supply) and high frequency (o the supply side) normally help anticipate liquidity but also contribute to asymptotically plateauing perceived value.
Commoditized supply is also normally related to asymptotic network effects.
Conclusions and Further Considerations
In a world of plummeting coordination costs, mastering the basics of network effects is an increasingly important skill, independently from your intention, goals or priorities. It is though a subject full of mystique since it gets to us as a result of great narratives and success stories.
Network effects are essential to the contemporary theory of the organization especially the interconnected society we live in. Understanding the basics of Network effects is functional to play with the advancements we’re seeing in technologies such as AI or the blockchain — as highlighted years ago by Chris Dixon in his famous reflection on blockchain-based tokens — new tools to help us overcome the challenges of launching and getting to liquidity.
We hope that this piece will help you grasp better the key concepts and support you in identifying and activating the mechanisms driving the NFX in your platform/network. New content and examples will come upon this blog to offer you further guidance and inspiration.
Click to Connect with us and explore what the Platform-Ecosystem landscape means for your organization, or explore the learning opportunities.
Before You Go!
Everything we do at Boundaryless, the company behind Platform Design Toolkit, is released in Creative Commons. In case you’re getting value out of these reads and tools, click the 👏button and hold down to 20–50 claps so we keep developing content to serve you and our community.
You can also connect with us on Linkedin here and here for continuing the conversation.
