Editor's Note: This article comes from36 kryptonEditor's Note: This article comes fromA16z36 krypton

, the scale of the translator God Translation Bureau; the original text comes from

, by D'Arcy Coolican, reprinted by Odaily with permission.

"Network effects" are the foundation of many important projects in the Internet age. As more and more people use the Internet, the Internet will become more valuable to users. But now, this law does not seem to be enough. There are many successful companies that seem to have no "network effect" during the development process, and then have network effects overnight. Behind this is the "hidden network" at work.

Many of the most important projects in the Internet age—from Wikipedia to Facebook and Bitcoin—are based on “network effects,” the idea that as more people use the web, the more become more valuable.

So we've become very good at analyzing and evaluating network effects. The metrics for identifying strong and long-lasting “network effects” in reducing customer acquisition costs, increasing liquidity, and improving retention appear to be fairly well established.

These are true and valid for many types of companies. Using these tools, it is easy to differentiate and analyze standard marketplaces, payment networks and many social platforms.

But for many companies, this traditional framework really doesn't work either.

There are many companies that have (or will have) strong network effects that don't apply to these analytics metrics, nor do their data help you measure, track, or even identify them.

But their network effects are real, just hidden.

So what does it matter?

Companies that have network effects but don't look like they have network effects are diamonds in the rough.

Because their networks are so hard to measure, they tend to go unnoticed in the short term, yet explode with power in the long term.

Likewise, the best startup ideas are the ones that sound bad at first—because the obvious good ones are picked and the competition is fierce—those companies with the strongest network effects end up being the best. So powerful precisely because they don't seem to have strong network effects initially.

Having "paradox" network effects creates a unique advantage that lays the foundation for a company's long-term success.

Most important is how this affects the competitive landscape.

If the value of early networks is not obvious in volume, then those teams and markets will get less attention and fewer copycats trying to develop similar products and similar networks.

This will give founders more time and space to develop products and build networks in an efficient, sustainable and defensible manner.

Think about what a ride-hailing company or food delivery company might look like if they had years to build their network (and competitive advantage) before competition arrived.

While hidden networks have their unique advantages, they also have various challenges: it is more difficult to raise funds using network effects theory rather than data, funding time horizons are often longer and more uncertain, and the ultimate strength of network effects can be very ambiguous .

So what is the hidden network? Here are three examples.

1. Slow networks

A slow network refers to the time delay between the creation of the network and the beginning of the value of the network.

Products on slow networks usually have long usage cycles or infrequent usage, which weakens the network effect. People tend to underestimate the value of a slow network compared to a fast network because the benefits are not immediately apparent.

Even if the company itself is growing rapidly, it may take years for a slow network to show its network effects and its value. In fact, some of today's fastest-growing companies also have slow networks.

Fast network vs slow network

Take Lambda School, a full-stack education platform and one of the fastest-growing startups today. It offers a programming learning program based on an Income Share Agreement (Income Share Agreement), and strives to help students find jobs after graduation.

In theory, it's easy to understand the network effects of Lambda School: as they recruit more (better) students, they should be able to:

(1) Find more employers looking to hire Lambda graduates

(2) Build a deeper Lambda alumni network, so that fresh graduates can rely on, learn from, and obtain job opportunities.

This flywheel brings more and better students to the top of the funnel and more and better employers.

So network effects win big, right?

However, Lambda is a 30-week program where graduates can find employment. Let's say employers need an extra few months to decide whether their first Lambda graduate is a good hire.

So far, this has taken almost a year to complete this single cycle.

Then, if a potential employer sees a student's resume and wants to hire him. But it still takes a few hours for him to complete his study before hiring him. At this time, it will be 7-10 months before the whole cycle ends and other employers see the value of Lambda.

Therefore, it will take several years for the value brought by the network effect of Lambda to start to appear. This is the logic of how slow networks come into play.

Lambda Founder Talks Network Effects' Role in Education

The advantage of slow networks is that, once established, they are often difficult to replace. Top universities have spent hundreds of years building network effects that don’t seem to show any signs of abating, despite every now and then predictions of their demise.

Another example of a slow network is social lending. When my co-founders and I started Frank, a social lending platform, we thought it would have a strong network effect.

We build a network of relationships where people can borrow and lend to each other, and the more people each user connects to on the platform, the more valuable the platform becomes.

Network effects are obvious, right?

But because borrowing is relatively infrequent (approximately every three years for our users), there is a lag of many years between when a node is added to the network and when it brings value to other users. Three years may be a lifetime for a start-up company.

While we were "sowing seeds", our slow network building meant our biggest challenge was surviving a few years in order to see them "bloom".

Networks in the lending and education industries, usually slow networks, but it can easily be applied to other fields, such as recruitment, medical or real estate, where the feedback cycle is long and the frequency of users is very irregular .

Testing for slow network effects is more of a craft than a science.

A company with a slow network looks more like a linear business early on than a network effect business.

But, while difficult to measure with normal network effect metrics, it is likely to have slow network effects. Can be tested in two dimensions:

(1) Whether it has all the characteristics of network effects (for example, as the number of nodes increases, the product is more valuable)

(2) Whether the product cycle is very long or the number of users is not fully saturated.

If the answer is yes to both, it's worth digging a little deeper.

Once you've identified a slow network effect, it's important that everyone—founders, employees, investors, users—has the patience and resources necessary to make the network grow and function.

Companies with slow networks often fail not because network effects don't exist, but because companies can't hang around long enough for them to be worthwhile.

2.1, Unfinished networks (Unfinished networks)

An incomplete network is one that is temporarily incomplete because of a product feature or strategic decision. When the network is finally completed, though, network effects are immediately apparent.

an unfinished transport network

Much like slow networks, unfinished networks also dampen network effects, but they won't show up in any analysis or metrics.

OpenTable is a successful example of an unfinished network. In its early days, OpenTable looked more like a normal SaaS business than a network effect business.

Restaurants pay OpenTable $200 a month for online seat reservation support and embed OpenTable plug-ins on their websites.

Seeing this, you must feel that this is a very direct network effect, and there is no network effect, right?

As OpenTable includes more restaurants, it has created an opportunity for itself to become the easiest place for diners to discover restaurants.

Once they have enough restaurants on their radar, they can invest in consumer-facing products like a website and an app to help shoppers find restaurants.

In this way, they complete the network. More consumers lead to more restaurants, and more restaurants lead to more consumers, which means stronger network effects.

From another perspective, in the first five years after OpenTable was established, it mainly focused on signing restaurants. It takes 10% of the restaurants in any community to become a product that consumers find easy to use, and then the network can be completed.

In the early days of OpenTable, if you only looked at its superficial network effects, you would miss the forest for the trees.

The challenge, of course, is that unfinished networks often don't finish. The startup graveyard is full of companies that thought they could finish the web but didn't.

The situation is especially dangerous in supply and demand, when the part of the network that needs to be done is the supply part. Because most businesses and employees sign up for anything that brings them business.

Therefore, it is crucial to understand what parts of the network you have built and how difficult it is to complete the entire network.

The key variable is whether there is an "urgent" need at the end of the network that has not yet been built? Are users willing to use your product to fulfill this need? Are they doing their best to help you with networking?

If the answer is yes, then you most likely have an "unfinished network".

2.2. Throttled networks

A throttling network is one in which product features or strategic decisions materially limit the size of the network or user engagement, thereby obscuring the strength of network effects.

This is very similar to that of an unfinished network, since the signal emitted by both is very weak. However, the unfinished network is missing a key part, and the current-limited network is complete, but with limitations.

Like an unfinished network, a rate-limited network appears to have limited network effects—until suddenly, it's no longer limited.

A social network for executives is an example. Chief, is a social networking platform for women in management. You can think of it like the Young Presidents Organization (YPO), but with a focus on women in senior management.

It is still in the early stages of development and its members mainly:

(1) Participate in a tutoring session or group discussion with peers once a month;

(2) Participate in a series of salon activities and conversations.

Clearly, they are trying to build a strong and valuable network. As more qualified women join, the community becomes more valuable. But if you measure it by traditional standards, it's hard to see its value.

In businesses with network effects, you will see customer acquisition costs (CAC) drop. But because Chief now deliberately limits its membership, they review and confirm the qualifications of each member, the waiting list is very long. Therefore, customer acquisition cost does not reflect network effects very well.

You might also expect that network effects would increase user engagement, but their engagement model, currently fixed at monthly group meetings, so no opportunity for increased engagement.

You may want to look for some important indicators from the quality of applicants or members, or refer to the company's Net Promoter Score (NPS), but these indicators can be vague and imperfect.

In the short term, Chief appears to have no network effects. But in the long run, they may unlock the value of the network by increasing engagement opportunities, increasing pricing as the value of the network increases, and even building a deeper membership base.

The ultimate manifestation of network effects will depend on what the founders think will work best for their product, and it doesn’t take a lot of effort on the part of the company to do so.

To some extent, Facebook was also a throttling network in its early days. Initial users must have a Harvard email address to join.

Then to have an email address with a .edu suffix, and eventually everyone else. This fits the definition of a current-limited network.

While Facebook doesn't limit people's engagement on the site (which makes the network effect more pronounced), it does intentionally limit the reach of the network.

This is an important difference between rate-limited networks and "private networks". In a sense, the current-limited network is only temporarily small, and the value proposition can support a larger network, but not yet.

By contrast, social networks that rely on exclusivity — dating apps like Raya, or membership clubs like Soho House are a few examples — tend to have an upper limit to their network effects.

Sometimes a throttling network is a deliberate decision by the founders, a temporary technical or operational restriction on the business, or a short-term regulatory control to keep the network small.

Sometimes the network is not small by design, but poor implementation or weak technology limits the network.

It is usually a good sign if the factors limiting the network can be addressed. Showing that the web is far more valuable than it appears.

The test for determining whether a network is a rate-limiting network with strong network effects is fairly straightforward: what happens if one of the constraints (eg, price, network growth, participation, etc.) is relaxed.

If the answer is positive or neutral, then it may be a network waiting to be unleashed.

3. Latent networks

This network, also known as the "come for the network, stay for the tool" network.

There are many companies that have developed tools or products before building the network. Delicious or Instagram are good examples of "come for the tools, stay for the network" companies.

But there are also companies that build networks before developing actual products or tools. Think of this as "come for the web, stay for the tools".

These companies can be especially powerful because no one knows what they are doing, often too late.

The concept of this network is that you start by building a community like a network, where users communicate with each other, engage with each other, and generally create value for each other. Ultimately, introduce a product that catalyzes or amplifies how the network works.

Before this product launch, there was nothing to measure network effects or monetize things, so it was difficult to really see the power and potential behind these networks.

This is a strategy that savvy game developers have followed for years. Even before playtesting, they set up a Discord server to help gamers build a community and network.

Most importantly, this ensures a vibrant ecosystem for the game's launch. This is important in social games because having other players can greatly improve the gaming experience.

Hypixel and Phoenix Point are two good examples of starting out with building communities that eventually become (or will become) in-game networks.

These latent networks are the most difficult to predict and most difficult to achieve "hidden network effects". Usually, these community participants are actually just the audience of the product without forming an underlying network, which means that users get value from the central node rather than the network.

When community participants are just the audience, the process of developing a tool or product is more of a linear business — such as a direct-to-consumer product — than a network.

It is very difficult to distinguish whether a business is a network without a product, or a player with an audience type. Many well-known entrepreneurs believe they have a network of people who want to engage with each other, only to realize that all they really have is an audience who wants to know their heroes.

So, how do you tell if a network has potential for activation? There are two dimensions: one is the participation characteristics of the network, and the other is the consumption of the audience. See if in the network, users are participating with each other or only with the central node.

Try asking yourself who gets extra value when someone new enters the community. If it's all (or at least some) community members, then it has network effects. If it is just a central node, then this newcomer may be an audience.

4. Hidden Networks Are Hidden Advantages

While hidden networks have their own unique challenges—often requiring a lot more patience, conviction, and capital—in my opinion, hidden networks are an overlooked type of company that everyone should consider creating, investing in, and investing in. companies, or work in them.

At the end of the day, building a business with network effects is essentially a race, and you need to always work hard to build network effects and reach the "magic turning point" before your competitors.