How Network Structure Shapes Defensibility and Scalability

Dense network connections and a consistently high creator-to-consumer (CTC) ratio can mitigate some of the challenges faced by weaker networks

Sameer Singh
Jul 27, 2020 · 10 min read
Image for post
Image for post
Image credit: Shutterstock

Most attempts at understanding network businesses revolve around studying user engagement. This is measured by a range of metrics including frequency of use, time spent, number of payments, etc. However, this tends to gloss over a very important fact — engagement is an effect, not a cause. If we want to truly understand what makes network startups work, we need to begin with the underlying cause of their engagement pattern — their network structure.

For the purpose of this discussion, let’s restrict ourselves to interaction networks, i.e. networks that connect users to enable interactions and information flow. This category can include social networks like Facebook, collaboration networks like Slack, payment networks like Paypal, etc. The network structure for these types of businesses is characterized by four different factors:

  1. User identity
  2. Network bridges
  3. Nature of connections
  4. Network (or cluster) density

I have already explained how user identity and network bridges influence the defensibility and scalability of interaction networks. So let’s take a deeper look at the remaining two — the nature of connections and network density.

Nature of Connections

Connections between any two users on a network can either be 1:1 or 1: many. In networks with 1:1 connections, users have symmetric relationships with other users. The most basic example of this is a telephone (or any other direct communication network) — any user can call any other user directly. Networks with 1:1 connections have a higher content creator to consumer (CTC) ratio — interactions rely on both participants creating and consuming information at the same time. As a result, every user brings the same incremental value to the network.

Networks with 1:many connections have asymmetric relationships between users, i.e. users can follow someone without that participant following them back. TV broadcasts (or any other broadcast network) are a great example of this connection type — all communication is directed from the content creator to viewers (or consumers). 1:many networks have a lower content creator to consumer (CTC) ratio, and so creators are significantly more valuable to the network than other users. For this reason, 1:many networks exhibit many of the same dynamics as 2-sided marketplaces (with extensive side switching).

Modern social networks often lie on a continuum between these two extremes.

Image for post
Image for post

Interaction networks like WhatsApp are much closer to the 1:1 end of the spectrum, while Twitter, YouTube, and TikTok are better described as 1:many networks. Others like Facebook and LinkedIn started out as 1:1 networks but now lie somewhere between 1:1 and 1:many. This concept can be extended to any type of interaction network — both Paypal and Venmo have 1:1 relationships where initiating an interaction (payment) is the equivalent of creating content or information. Similarly, in the domain of SaaS-enabled networks, Slack can be considered to have 1:1 connections while Carta and Figma have 1:many connections.

Impact on Defensibility: Nature of Connections x User Identity

I have previously explained how the defensibility of networks depends on the importance of user identity. To summarize, the strength of connections between any two users is much stronger when it is based on their unique identities. When people interact with specific individuals in a network, their loyalty depends on the ongoing presence of those specific individuals. These users are less likely to adopt competing networks if they do not have those individuals. This is the entire premise behind the “social graph” — a key component of Facebook, Snapchat, and even Twitter’s defensibility. This minimizes multi-tenanting with competing networks that offer the same value proposition, irrespective of the nature of connections (outside of small niches, e.g. Twitter vs. Parler or Gab).

However, in cases when user identity is not a core part of the network, the nature of connections comes into sharp focus. Typically, identity agnostic networks are more likely to be 1:many because there is no meaningful basis to connect users directly. At this point, the content creator to consumer (CTC) ratio becomes exceptionally important. Without a social graph tying users to a network, both user retention and defensibility become purely dependent on the content created by users. If the CTC ratio is low, competitors simply need to acquire a small base of creators to build a viable alternative. This means that these types of networks need a higher proportion of users to be creators — the content flywheel simply has to work harder to keep users interested.

Take TikTok as an example. TikTok’s feed is not based on any social or follower graph. Instead, videos pop up based on an algorithm that prioritizes engagement, irrespective of who users follow. This dilutes the identity of creators, resulting in very weak network effects. However, it has remained viable because of its high CTC ratio — about 34% of daily active users in the US also create videos every day. One reason for this is that its algorithm levels the playing field for creators — they don’t need a huge following to go viral. This incentivizes creation and the resulting content firehose gives it more of a runway than short-lived identity agnostic networks like Meerkat and Vine. However, this approach does not create lasting defensibility because it can also be replicated by competitors. As a result, “multi-tenanting” is a major risk. We can see this playing out in the short-video space already — Triller recently matched TikTok’s userbase in the US, Dubsmash is seeing a revival and even Byte appears to be gaining traction. This is, of course, in addition to creators reposting videos on larger networks like Instagram.

Network Density

Network density is a measure of the number of connections on a network relative to the number of participants. However, this is a very simplistic view because the density of connections varies across clusters (or hotspots). For example, Facebook users in London may have numerous connections in the city, but fewer connections in San Francisco and vice versa. This is precisely why network bridges (connections that “bridge” clusters) are so critical to scaling. Therefore, it is more meaningful to gauge the density of network clusters, as opposed to the network as a whole. In dense network clusters, every user in a cluster interacts with every other or most other users. On the other hand, users only interact with a handful of other participants in sparse network clusters.

Image for post
Image for post

Communication networks like Snapchat are the prototypical example of dense network clusters. Every user is connected to multiple users within the cluster leading to a large number of “mutual friends”. This not only gives users more participants to interact with but also unlocks group interactions (group chats). In contrast, payment networks like Paypal have sparse clusters, with each user sending payments to a handful of participants — “mutual friends” and group payments are rare use cases. Even Venmo, which has increased density on its payment network by adding social elements, has ended the ability to add group accounts.

Density has a direct impact on the frequency of interactions. In Q1 2020, Snapchat users averaged 17.5 snaps per day. On the other hand, Paypal users averaged 3.3 transactions per month. Granted their value propositions are very different, but a part of this dramatic engagement gap can be explained by the number of connections their users have.

Similarly, in the B2B domain, Figma is more likely to have dense network clusters while Carta’s cap table management software (not its exchange) would be closer to a sparse network.

Impact on Scalability: Network Density x Network Bridging

Network density obviously shapes engagement on any type of network, but its impact on scalability varies by the type of network. For example, hyperlocal networks like Bumble or Nextdoor can only connect users in the same local area. Since the number of users in an area is naturally limited, they cannot use a Facebook-like, friend invitation model — that would make it very hard (and expensive) to gain critical mass. Instead, they try to maximize connections between local users by either recommending matches frequently (Bumble) or by connecting everyone in the area by default (Nextdoor). In other words, unbridged networks need to maximize the number of connections within each cluster to maintain some level of organic user acquisition.

On the other hand, true cross-border networks, like Paypal, don’t face this challenge. They can leverage users in one cluster to organically attract users in another — this means that cluster density is not critical to scaling. Similarly, cluster density is certainly desirable for bridged networks like Facebook or Snapchat, but not a requirement for scaling — this gives them more leeway to create a social graph. By definition, the social graph of a user is a map of other users on a network they are connected to (a subset of total users on that network). In order to create this social graph, social networks intentionally restrict interactions (and density) until users make a deliberate connection with another user (a friend request). Of course, social networks can recommend new friend connections, but those recommendations are not as central to the value proposition as they are on a hyperlocal dating app.

To summarize, hyperlocal or unbridged networks need dense clusters to scale sustainably, but cross-border or bridged networks are less sensitive to density.

Impact on Defensibility: Network Density x User Identity

Unique user identity remains the primary factor influencing network defensibility. However, among networks that have a high reliance on identity, density enhances defensibility even further. For example, it is more difficult to get users (and their entire social graph) to multi-tenant if they interact with 100 users on a network as compared to one where they interact with 10.

However, network density gets less relevant as the importance of identity fades. For example, TikTok pushes recommended content into a user’s feed, irrespective of whether that user follows the creator. As a result, the number of connections between users and creators is less relevant. This allows competitors like Triller, Dubsmash, and Byte to simply build up a critical mass of content, without replicating user-creator connections.

Network Matrix: Impact of Network Structure

The modified network matrix below shows how all the different facets of network structure interact with each other.

Image for post
Image for post
Note: Networks highlighted in orange are SaaS-enabled

Tier-1 networks are not especially sensitive to the CTC ratio, but density can enhance their defensibility even further. At the other end of the spectrum, Tier-3 networks require both dense network clusters and a consistently high CTC ratio to remain viable. Let’s take a look at a couple of examples to see how these factors work together to shape long-term engagement.

tbh was an anonymous polling app that was acquired by Facebook in October 2017. It is best described as a Tier-3 network because interactions were anonymous and were largely localized to users from a specific school. To make up for its hyperlocal nature, tbh borrowed a mechanic from dating apps and proactively prompted users with a series of 12 polls to increase density. Each poll gave users (content creators) a choice of four friends, chosen at random. Users who were chosen by a friend (content consumers) received a gem that could then be used to unlock more polls. In other words, it had 1:many connections. In the beginning, its users were very engaged, answering an average of 6–7 polls per day at the time of its acquisition (high CTC ratio). However, their CTC ratio then began to decline as users gradually lost interest in anonymous compliments — their content flywheel simply could not keep pace. As a result, it was shut down because of “low usage” just 8 months after it was acquired and less than a year after its initial launch.

Carta is a good counterexample. Carta is a Tier-1 network that connects startup equity administrators to shareholders. Since most information is directed from startups to all of their shareholders, it can be considered a 1:many network. It last reported that it was connected to 11,000 startups and 700,000 shareholders, so it is safe to say that its CTC ratio is on the lower end. In addition, its network density is fairly low as most employees only have holdings in a handful of startups. Large, high profile investors like a16z only make up a small minority of the overall network and still only have an active portfolio of about 200 startups. Despite these seeming disadvantages, it has become indispensable because network bridges helped it spread across the startup community and unique shareholder identity made it exceptionally sticky. These two features have helped it overcome its low density and CTC ratio.

To summarize, engagement is largely a consequence of your network structure, and measuring it does not provide predictive value. Instead, it is more useful to understand the structure of your network — whether it is B2C or B2B — to forecast the opportunities and challenges ahead. Specifically, network density and the nature of connections are important, but they are secondary to user identity and the presence of network bridges. In general, it is safe to say that identity agnostic networks struggle with user retention, and hyperlocal networks face scaling challenges. However, higher network density and CTC ratios can mitigate these challenges to some extent.

A network effects guide for founders

Sameer Singh

Written by

Network Effects Advisor, Author & Investor

Network effects are among the most powerful, but also the most misunderstood, forces that shape technology startups. is my attempt to lay a “trail of breadcrumbs” to help founders find their way.

Sameer Singh

Written by

Network Effects Advisor, Author & Investor

Network effects are among the most powerful, but also the most misunderstood, forces that shape technology startups. is my attempt to lay a “trail of breadcrumbs” to help founders find their way.

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store