A network effect is an attribute, demonstrating how much value is created within a network as a factor of its size (i.e. the Metcalfe’s Law). So the short answer is yes, size does matter, a lot. But, value creation extends way beyond it. That is, if you measure the value just by the network’s sheer size, you won’t get too far and your network or marketplace won’t build substantial value.
Size Is Not Everything
I'm sorry to break it to you, but size is not everything. Network size is a necessary condition for value formation and we already know the the value starts kicking in only above the minimal critical mass point. But, the network size is not a sufficient condition in order to create network effect. There are additional core attributes that need to be accounted for. These are:
Homogeneous Network
Liquidity
Repeatability
Trust
The Mathematical Formulation Of A Network
Before we dive into each of the above attributes and learn about how they affect value creation, it’s important we have a way to formulate and visualize a network. Let’s take this discussion pseudo mathematical and use an analogy from the Graph Theory:
A network can be depicted by a Graph (G) including a set of Nodes (N) and Edges (E). putting it all together: G = [N, E].
The number of nodes constitutes the network size. In the online world we’ll usually find networks or marketplaces that can be depicted as Incomplete Graphs. It means that not each and every node is connected to all other nodes, e.g.: not all people (i.e. nodes) connected (i.e. edges) to all the people on Whatsapp.
We can further describe the relationship between the nodes by attaching attributes to the edges. If a node is a supplier-only (e.g. lessor on AirBnB), it will have only outgoing arrows. A receiver-only (e.g. a lessee on AirBnB) will have only incoming arrows. Finally, a node that serves as both supplier and a receiver (e.g. both a lessee and a lessor on AirBnB), will be depicted as a bi-directional arrow.
Now that we have a terminology straight, let’s dive in to the additional attributes of the network effect. For sake of simplicity, I'm using examples of a two-sided marketplace. The ideas, concepts and formulas presented here can be extended and developed to other types of networks and marketplaces.
Network Homogeneity
A network homogeneity [H] can be defined as how many users exists in the network under the same Class. We will define a Class to be a core attribute (or a set of attributes) that defines the user's reason to conduct a core activity in the network.
In online marketplace and networks, there will be at least two classes, one for each side of the fence (supply and demand), where the supply and demand classes are complementary. For example, if the supply class, denoted S(c), is fine wine sellers, then the demand class, denoted D(c), will be fine wine buyers. A marketplace where supply and demand embeds the same amount of complementary classes shall be defined as Homogeneous Complete or H = [S(c), D(c))], where S=D.
It's worth noting that sometimes, users fit more than one class (e.g. a user can be a seller and a buyer). Thus a single user can be a trigger to the formation of two classes. This will be important when we discuss liquidity.
Why all this matter?
The process of identifying, classifying and matching users on each side allows detecting several important ingredients in the process of creating value and therefore a network effect:
Identify and classify: mapping all the existing classes allows you to detect the various drivers of the users existing on each side of a network. Continuing on the fine wine example, classifying will allow you to learn that some sellers sell only very high priced wines and you need to match them with buyers who are willing to pay for it. One the other hand, some buyers want to by only specific type of wines (flavors, origin, years, etc.) and you will need to make sure there's a supply for it. During the classification process you may recognize that you have some classes that you can’t serve (either at all, or in the current phase of the network evolution), or ones you don't want to serve and that are not relevant for you. Classification is a critical input for steering the marketing towards the right target-users who can actually help generating value in the network.
Matching classes on both sides: after you classified each and every side in the network, you can now easily search for matching classes on each side. If a class on one side doesn’t have at least one class matching it on the other side, there will be no value creation or exchange at all! In such cases, such users/nodes should not count towards network size and therefore will not increase value creation and network effect.
Finally, It is a the preliminary tool in checking and making sure you have liquidity
Every marketplace should strive to be Homogeneous Complete.
Liquidity
In simple terms, liquidity is defined by how many of the listed users on the network are currently active and performing a core action in it.
Going back to Metcalfe’s law, In online marketplaces, liquidity at any given moment determines
the effective, real-time network size and therefore, the value creation. Looking at the rhombus class in Graph 3, there are no active users on the demand side, thus no value creation can be formed. So now we know we actually have two network size values, the theoretical one N(t), totaling the number of nodes registered in the network, and the number of active nodes, N(a). In Graph 3 N(t)=11 and N(t)=6.
The N(a) / N(t) ratio is very telling and a very important indicator to monitor over time. It is not sufficient to grow the theoretical network size and expect value to increase exponentially. Value creation is generated by the active nodes, per class. It’s rare to see networks with theoretical-to-active ratio that equals one, but if this ratio is diminishing, you know you have a problem.
While liquidity is always important for asynchronous networks and marketplaces such as selling and buying cars (e.g. Beepi.com) or customers and beauty saloons (e.g. Styleseat.com), liquidity becomes a critical factor for synchronous networks and marketplaces such as Uber and real-time trading and auction platforms. In such cases, timing the supply and demand is key to liquidity management and sophisticated algorithms are required.
Repeatability & Frequency Expectancy
Repeatability is defined as the number of times a user returns back and perform an additional (one or more) core action, i.e. “transact” in the marketplace, in a given time frame.
Frequency Expectancy is defined as the number of times a user can be expected to return back and perform an additional (one or more) core action, i.e. “transact” in the marketplace, in a given time frame. Important to note, the frequency expectancy is a derivative of a customer behavior (to conduct such action/s online), rather than a derivative of the platform or the operating company’s execution capabilities which can impact repeatability. Therefore, factors such as the existence (or lack of) product features, current liquidity, marketing budget, etc. shouldn't be a factor.
While repeatability touches upon multiple (sometimes convoluted) issues, I’m highlighting here just a few of them:
(effectively) Zero Repeatability In networks and marketplaces, usually one side of the marketplace has higher frequency expectancy than the other. Whether due to a business model or an industry vertical behavior, sometimes, you have almost zero repeatability - meaning you essentially acquire one-timers. If you decide to go into a consumer marketplace for selling and buying 2nd hand real-estate, you shouldn’t expect the same seller, to come and sell a property twice a month. Though there are some ways to handle this, all in all this is a tight spot in hoping for a network effect and in such cases unit economics of the marketplace becomes a real critical element in creating a viable network effect. In such cases where there’s no inherent repetition, marketplaces and networks artificially inflating the network size, by acquiring one-timers. This will of course won’t help to drive value creation within the marketplace and effectively, network effect can not be created as the network grow. One of the indications for such a case will be by monitoring the liquidity ratio which will decrease over time, unless the network keep pumping one-timers to compensate.
The Repeatability Expectancy Coefficient We discussed the fact the one side of the marketplace is prone to have higher repeatability. By setting the repeatability expectancy for each side and monitoring their ratio we can understand the strength of a marketplace and the probability to form a network effect. Again, for simplicity, let's assume a two-sided marketplace. We denote repeatability expectancy of the supply side as R(s) and the demand side as R(d). The closer their ratio is to one, the higher the chances to obtain a network effect. This way, we make sure the demand and supply are aligned, as well as liquidity and that the exchange of value from already acquired users in the network is not being based entirely on newly acquired users. But that’s not enough, since the repeatability expectancy ratio can equal 1, even if we have zero repeatability on both sides (in that case the users come once and both R(s) and R(s) equals 1. Therefore, we will add another condition: both R(s) and R(d) values to strive to be as high as possible. A more detailed analysis with this ratio can be performed on a Class level rather the a network level, and so on and so forth.
Trust
Network effect is about creating value or the exchange of it, as a factor of the network size. By now we understand that size is not everything. So what else? Just gathering a big group of people in one place with no clear set of rules which everyone can appreciate and trust is not a marketplace, it’s well...more of a jungle really.
People need to trust the other side to a certain minimal level as well as the vehicle through which they transact in order to make sure that the swap of goods (i.e. value) will actually take place. The engagement rules that allow the creation and exchange of value is defined by the network. Beyond formulating it, the network needs to strictly impose them, as well being able to detect and analyze the motivation of the people joining it. Any deviation from the rules must be addressed with transparency and bring remedy to the damaged side, sometimes even if the network, is not directly or legally liable. Users who feel they have been deceived, or that their expectations were significantly unmet, will develop negative sentiment towards the network, even though it is merely the facilitator and not one of the transacting sides.
Commentaires