Internal network effects
One can potentially distinguish many different types of network effects (recall network effects arise when the value to a user increases in the number of other users that use the same product or service). For instance, one might distinguish network effects that work through data from those that do not, or network effects arising between people that use the same communication platform from those arising between buyers and sellers on a marketplace (i.e. one-sided vs. two-sided network effects). The network effects manual, produced by the venture firm NFX, has a classification of 15 different types of network effects. Here, we propose adding one more to their list – internal network effects.
The benefits of classifying network effects
Before doing so, however, it is useful to start by reflecting on why classifying network effects might be useful, and briefly commenting on the NFX list.
We see two main benefits from doing the classification exercise. First, it helps reveal that network effects are more ubiquitous than one may realize. Some of the network effects listed by NFX (e.g. bandwagon, expertise, and tribal network effects) will not be obvious to most people. And by being aware of these, founders will see more opportunities to leverage network effects to build defensibility into their products. Second, network effects can work in different ways across the various categories, and understanding this can help uncover important differences in terms of the strength and defensibility afforded by the various types of network effects.
With this in mind, and in order to set the stage for network effect #16, here we briefly review and comment on the last three types of network effects identified by NFX, two of which were added after the initial list with 13 types was published:
Bandwagon network effects (nfx #13): people are willing to pay more for some products that others buy because they want to be part of a new trend, or because they fear being left out. Tesla certainly benefits from this effect, as did Apple’s iPhone in its earlier days, as illustrated in this well-known satire and this one. Compared to most network effects, this one can be quite fickle though: what is cool today may no longer be cool tomorrow, and when too many people join, the coolness factor evaporates or can even work against the initial product – early-adopters may want to leave and chase the next novelty. Put differently, bandwagon network effects are unlikely to work at scale. Oftentimes, they look more similar to virality: creating a “trendiness” buzz around a product that makes everyone talk about it may help generate awareness, but does not constitute a sustainable source of defensibility. Especially when the features that made the initial product cool are copied by competitors with strong enough brands of their own (e.g. Blackberry).
Expertise network effects (nfx #14): people get more value from a product that is more widely used, because this attracts more expertise to develop around the product, which benefits users who want to tap into that expertise. An example is Salesforce’s CRM software, where part of the value to a firm adopting it is knowing that its current and future employees are more likely to know how to use it, and that there will be many experts that can help (either via online communities or via direct hires). This is why once a product becomes an industry standard, it becomes difficult to unseat. As NFX points out, in the face of an incumbent with expertise network effects, competitors will need to try to minimize switching costs, so making their product as familiar as possible (e.g. Google Sheets’ approach to take on Excel), whereas an incumbent can focus on adding more features and making their product even more powerful.
Tribal network effects (nfx #15): belonging to a tribe (e.g. the Harvard alumni club, the YC alumni network, or a BMW user group) is more valuable when there are more people in it. This is of course a type of social network, except the network effects between members of the same tribe tend to be stronger than with regular social networks. Indeed, members of the same alumni club are much more likely to do favors for each other than say random Facebook or Twitter users. While it is not easy for a firm to develop tribal instincts among its customers, some brands do seem to have produced a cult-like following (e.g. Comic-Con and Harley Davidson). An interesting implication of tribal network effects is that the existence of a competitor (a rival tribe) can actually strengthen them.
NFX’s list of 15 different types of network effects is not supposed to be definitive and there is plenty of overlap between many of the different types. In that spirit, we now propose network effect #16.
Internal network effects
One typically thinks of network effects as operating between clearly distinct entities that make independent adoption decisions – different consumers, different app developers, different firms (as either buyers or sellers on B2B marketplaces like Alibaba or Faire or AWS marketplace).
However, there are instances in which the network effect is between different employees within the same firm: each employee can individually decide whether or not to use a product or service, but as more employees use it, it becomes more valuable for everyone within the enterprise. This is what we call an “internal network effect”. And internal network effects have become increasingly important as a source of defensibility for software providers in the last decade or so due to the adoption of the bottom-up go-to-market approach to selling into enterprises, in which software is targeted at individual employees rather than at the company HQ. This approach may take longer (though selling to enterprises can also have very long sales cycles), but when it succeeds (i.e. when a large share of employees adopt a product), it makes it harder for an enterprise to switch to a competing product.
Slack, in its original form, is a classic example of internal network effects: the more people within an enterprise used Slack, the more valuable it became for everyone to join. Over time, people used Slack channels to communicate in all kinds of communities, and so what was originally an internal network effect grew to become external.
Many business productivity software products enjoy internal network effects, since they are designed for employees to use collectively, even when they are not focused on communications, like Slack. Examples include Asana (allows teams to manage projects and tasks), Monday.com (allows teams to plan, organize and track work), and many others. In these examples, the network effects are mostly internal; the only way they can include an external component is when employees change firms and carry their software preferences with them (we discuss this below).
Internal network effects are generally less strong and less defensible when compared to regular (external) network effects for several reasons:
They are usually limited to each enterprise customer, so their scale is relatively small, especially if the target market is small and medium-sized enterprises.
As in the case of regular network effects that are local (e.g. Uber, DoorDash), one has to start from zero and repeat the same playbook for each new “locality”; here each locality corresponds to each new enterprise customer, and requires “seeding” them and growing internal adoption each time.
It is easier for users within the same organization to coordinate and switch to a rival product if it is superior, relative to the case when these individuals are not connected in any way. Indeed, a company’s HQ can mandate such a change. This means products built on internal network effects are less defensible than those built on regular (external) network effects, all other things equal.
Having said this, some enterprises have many thousands of employees, and when they have all learnt to use a particular software product, the switching costs of getting them all to adopt a new software can be high enough to create a powerful lock-in effect. While the company HQ could in principle still dictate matters, in practice the increased autonomy of individual employees in their software choices means this possibility is a lot less likely today. Winner-take-all at the level of the entire market is still less likely than with internal network effects, but winner-take-all is more likely within any given enterprise.
Moreover, as employees shift across employers, the advantage of a common software that is used by many large enterprises (in terms of reduced learning costs) comes into play. Each new enterprise may be more likely to adopt the software used by other enterprises because it becomes easier to attract new employees (an expertise network effect). This is particularly important when the software becomes a key part of the employees’ tasks in the enterprise and involves a significant learning curve (e.g. the choice of a programming language or technical software). And, as happened with Slack, once a product becomes industry standard, it may be possible to extend its use cases beyond each individual enterprise boundaries.
Concluding thoughts
In some sense, there is almost no limit to the number of ways one could characterize network effects. The real question is whether the characterizations lead to useful insights about the strength and defensibility of the relevant network effect, and more generally about firm strategy. Above we have argued that our proposal for nfx #16 meets that standard. What do you think? What would be your choice for nfx #16 or beyond?
If you enjoyed reading this, sign up now so you will receive our posts directly in your inbox as they are published.
In the meantime, tell your friends!
Interesting. I'm working on a similar construct but a bit more specific ie developer platforms. Dame dynamics but with a few Interesting nuances.