NETWORK

EFFECTS

THE BUSINESS
MODEL ANALYST
1
SUPER GUIDE:
NETWORK
EFFECTS

BY DANIEL PEREIRA

2
 © THE BUSINESS MODEL ANALYST

The Business Model Analyst is a website dedicated to

analyzing business model types, patterns, and innovation
using the business model canvas as its primary tool. The site
o ers a wide variety of free and premium content, including
digital products such as PDF tools, presentations,
spreadsheets, ebooks & guides, and much more. Check it out
here.

Daniel Pereira
The Business Model Analyst
Ottawa, ON, Canada
businessmodelanalyst.com

3
ff
 TABLE OF CONTENTS
Introduction 1
What are Network E ects 2
What aren't Network E ects 4
Feedback Loop of Network E ects 6
Direct vs Indirect Network E ects 8
Properties of Networks 9
Terms of networks 11
Laws of networks 12
The Dynamics of Network E ects 13
Value proposition 13
Users and Inventory 14
Competition 15
Types of Network E ects 16
Direct Network E ects 16
2-Sided Network E ects 18
“Social” Network E ects 21
Hidden Network E ects 23
Slow networks 23
Un nished networks 24
Throttled networks 25
Latent networks 25
Hidden advantages 26
4
fi
ff
ff
ff
ff
ff
ff
ff
ff
ff
ff
Strategies for Building Network E ects Metrics 27
Network E ects Metrics 29
Acquisition 29
Competitor-Related Metrics 30
Engagement-related Metrics 31
Marketplace Metrics 32
Economics-Related Metrics 34
Case Studies of Companies with Networks E ects 35
Facebook 35
Airbnb 35
Medium 36
WhatsApp 37
When Network E ects are not enough 38
Conclusion 41
References 42
About the Author 44

5
ff
ff
ff
ff
 INTRODUCTION
Some of the greatest current brands - Facebook, Amazon,
Airbnb, WhatsApp, Uber ( just to mention a few) - have their
businesses built upon network e ects. Because these and
other companies have already learned that the value of their
products increases as the importance of the network does.
And the curious part of that is the network e ects don’t
require much maintenance, once they are built, they tend to
keep running.

However, not all businesses are able to develop network

e ects - and not all of them should, anyway. Their dynamics
can change a lot depending on the application. Besides,
there are many di erent approaches and strategies, and they
are directly dependent on the type of product, service,
market, and industry.

1
ff
ff
ff
ff
 WHAT ARE NETWORK EFFECTS
Also known as demand-side economies of scale, the network
e ects take place when the more people use the product or
service, the more valuable it gets to everybody. That’s
because as more users participate, they get an improved
experience. At the same time, this encourages new
participants to take part, as they see bene ts in the network.

It is fair to say that the phenomenon of the network e ect

began in the early 1900s, with the invention of the telephone.
The telephone became the strong means of communication
we have watched due to the increasing number of users. If
you didn’t have anyone to call on the other side, why would
you want a telephone?

Nowadays, the network e ects can be strongly felt in social

media, such as Facebook, Twitter, Instagram, Linked In,
YouTube, and WhatsApp. The more people create accounts
on these sites, the more valuable they get, not only for users
but for shareholders.

Because as more users join, more businesses get interested

in advertising to those audiences. And the cycle keeps going:
2
ff
ff
fi
ff
more advertisers lead to more revenue for the website. More
revenue means greater investment in the service to the
consumer, enhancing the user experience.

And that is also why the network e ects might become the
key to success for many software companies, and, thus, they
make use of the phenomenon in order to build more
attractive products and services.

Among the bene ts of network e ects, are:

- Creating barriers for users to exit and for new entrants to

entry;
-- Helping develop winner-take-all markets;
Protecting software companies from the competition;
-- Helping scale your business;
Increasing your customer base, market share and value
proposition;
- Fueling pro t.

3
fi
fi
ff
ff
 WHAT AREN'T NETWORK EFFECTS
A pretty common misconception is making some confusion
between the concepts of network e ects and virality. The fact
that a platform has scale does not mean it has network e ect,
maybe it is simply a viral growth.

Sometimes, both the phenomena happen at the same time,

but they are not necessarily related and they are surely not
the same thing. The main di erence is that the network
e ects increase value due to the adoption by more users.
Viral growth simply increases the speed of adoption.

The viral growth doesn't always lead to network e ects. But

the product virality does. For example, Twitter would not be
interesting if there weren't’ anybody else to check on what
you have posted.

People commonly misconcep mouth-to-mouth and network

e ects. Sometimes, a product or service just grow fast as a
result of referrals and recommendations. But the original
users don’t have their experience improved due to new
participants - so, the value does not change.

Another regular mistake is by confusing network e ects and

scale. Having scale is not the same either. The economy of
4
ff
ff
ff
ff
ff
ff
ff
scale grows bigger because the product becomes cheaper to
be produced, meaning a lower cost per unit.

When it comes to network e ects, the product is not cheaper,

it is actually more valuable, leading to higher engagement
and higher margins.

5
ff
FEEDBACK LOOP OF NETWORK
EFFECTS
The term feedback loop is employed in many areas of
knowledge, since biology and physics to business. A
feedback is the information received as a reaction to a
product. This information can be used to modify the original
product, to create a better one. A feedback loop is, therefore,
the process by which the feedback and outputs of the
customers are circled back and used as inputs.

This interdependent cycle between producers and

consumers is known as a “positive feedback loop”. In a
positive feedback loop, two events are mutually reinforcing: a
change in one can cause a bigger nal increase in the other,
and so on. This is especially useful for multisided platforms.

6
fi
In spite of that, it is important to highlight that not every
growth on platforms leads to a higher quality service.
Because the growth will attract both high and low quality
users. That is why it is imperative that a high standard for
access is maintained.

Uber is a classic two-sided marketplace which is bene ted by

the feedback loop. More passengers attract more drivers and
vice-versa. More drivers mean higher availability and lower
waiting time, which will be perceived as value by the
passengers’ side. More passengers, on the other hand, mean
lower downtime and higher prices (due to Uber’s surge
pricing).

7
fi
 DIRECT VS INDIRECT NETWORK
EFFECTS
Direct network e ects

It is when the value of the product or service increases along

with the number of users. It is the case of brands such as
WhatsApp and Skype (when it comes to one-sided network
e ect), or Airbnb and Ebay (when it comes to two-sided
network e ect, in which more supply means more demand
and vice-versa).

Indirect network e ects

It is when the value proposition increases as a result of

complementary products and services that aggregate to the
business. Famous examples are Microsoft Windows and
Google Android.

8
ff
ff
ff
ff
 PROPERTIES OF NETWORKS
Types of nodes - Homogeneous or Heterogeneous

- Homogeneous network: is composed of similar types of

nodes. In other words, most of its participants are the
same kind of users and share the same interests. For
example: Skype and WhatsApp, with individuals who want
to communicate among each other.

- Heterogeneous network: on the contrary, is composed of

di erent types of nodes. The participants divide in two
categories, with two distinct roles in the business. For
example: Uber and Airbnb, with one side composed of
consumers like passengers and renters, and another side
composed of providers like drivers and real estate owners.

Type of clustering

- Hub-and-spoke or star network: the central hub is the

ttest node, and it grabs all the links.

- Clique or connected network: all the nodes are connected

among each other.

Directionality of connections

- Unidirectional (or one-way following): the connections

might not be mutual. One may follow another, and the
other one does not need to follow one back.

- Bidirectional (or two-way friending): the connections are

based on one mutual linking. For example, on Facebook, if
9
fi
ff
 you add a friend, they add you back.

Complementary networks

It occurs when the usage of one product increases the value

of a complementary but separate product. In turn, this
complementary product enhances the value of the original
one. That’s usually the case of operating systems, such as
Microsoft Windows, Google Android or Apple iOS.

10
TERMS OF NETWORKS

- Network: is a group of interconnected nodes, which may

be people, such as in social networks, or things, such as in
telephones or computers networks.

- Marketplace: is a network where transactions take place

between two or more sides with distinct customer
segments on each. It aims for supply and demand to get
attracted to the same place.

- Platform: is a multi-sided network where users,

developers and the platform itself create an environment
which increases value for every and each of the groups. It
may also be programmed, customized and extended by
outside users.

11
 LAWS OF NETWORKS

- Sarno ’s law: the value of a network is proportional to the

number of viewers. Examples: Yahoo.

- Metcalfe’s law: the value of a network is proportional to

the square of numbers of connected users. Example:
Facebook.

- Reed’s law: the value of a group-forming network is

proportional to the number and ease with which groups
form within it. Example: Slack.

12
ff
 THE DYNAMICS OF NETWORK
EFFECTS
Every product or service will have di erent kinds of network
e ects, and they will develop in their own way over time. The
challenge for business people is to try to project how they will
evolve. In order to that, it is necessary to understand how
three aspects of the company might change.

VALUE PROPOSITION
It is important to know what value proposition drives the
network e ects, so you can understand if they are strong or
not and how they evolve. Take a look at some examples:

- Rideshare: the network e ects here are based on the

reinforced relationship between driver supply and
passenger demand. The more drivers, the less waiting
time. The less waiting time, the more riders. The more
riders, the more drivers, and on. However, when the
rideshare supply hits a real low waiting time, such as 5
minutes, for instance, for any platform in the market, the
platforms will have to create an upgraded value
proposition to superpass competition, like reputation,
price, experience, etc.

- Social lending: the network e ects here are basically two -

the liquidity (amount of money available) and the number
of participants interested in lending or borrowing. So, more
people would mean more demand and more liquidity. But,
the truth is that people don’t usually trust more than 7
friends. So, when the lending groups hit more than 7
13
ff
ff
ff
ff
ff
 participants, they would become less likely to lend or
borrow, turning out to be a negative e ect.

- Social network: taking the example from the biggest

social network in the world, Facebook, the network e ects
weakened when the social network turned into social
media. Because too many friends made the network less
personal, and people got uncomfortable to share personal
content, and started sharing more public content.

- Decentralized platforms: bitcoins is an example that

di erent value propositions for one same platform may
mean stronger or weaker network e ects. If bitcoin is
concepted as digital gold, that would attract more buyers
and sellers and that means more liquidity. However, when
it is seen as a payment method, then more people just
means network congestion.

USERS AND INVENTORY

The type of users and inventory your business has today will
certainly de ne the future network e ects.

- Commoditized supply: it occurs in platforms with

commoditized inventory, such as on-demand storage,
ridesharing or delivery companies. That’s because the
user/customer will perceive the supply as interchangeable.

- Di erentiated supply: platforms and marketplaces which

possess more di erentiated inventory will have stronger
network e ects, since their diversity will suit the particular
preferences of users/customers. They will reach liquidity
across di erent types of inventory and also with continued
new supply. Nevertheless, the more di erentiated the
inventory is, the better the curation and matching service
needs to be.

14
ff
ff
ff
ff
fi
ff
ff
ff
ff
ff
ff
 - Incremental users: not all members of a network are
equal, and some are more valuable than others for the
business. There are mainly three kinds of incremental
users: contaminants, neutrals and contributors.
Contributors are the ones who add value to the platform
(for example, a great content producer in Twitter or
Instagram). Contaminants are the ones who disengage
other users with their behavior. And neutrals are the ones
who won’t in uence the balance. The key is to focus on
attracting the good users for your business.

COMPETITION
The nature of the market your business is inserted in and its
competition will also be essential when forecasting network
e ects.

- Network overlap: when another business has a very close

network to yours, it is more likely it will get into your
market, no matter if it is not a direct competitor. That
happened, for example, when Instagram included their
Snapchat-like (and based) Stories.

- Switching costs: when the switching costs are very low,

they can make network e ects weaken. They may
encourage your customer to use products or services from
direct competitors at the same time, due to the low
barriers to entry. A good example here is the dating apps.

- Multi-tenanting: as seen right above, low switching costs

may foment multi-tenanting. But there can be other
reasons for users to not rely on a single platform to meet
their goals. That’s the case of jobs marketplaces for
example. Usually the companies will list their openings in
several hiring platforms, and that will put some pressure
on the operator, especially regarding prices and features.

15
ff
fl
ff
TYPES OF NETWORK EFFECTS
There are 13 types of network e ects, which will get into ve
broader categories:

DIRECT NETWORK EFFECTS

The strongest but simplest network e ects are direct: the
more users, the bigger the value. These e ects started back
in 1908, when AT&T was a telecom leader, and its value was
due to the number of customers, not to any more advanced
technology over other companies. Let’s take a look at some
kinds of direct network e ects:

1. Physical (Direct)

These network e ects are tied to physical nodes and links.

Each node is connected to every other node. So, every
additional node adds a new connection for all the previous
16
ff
ff
ff
ff
ff
fi
 ones. Since the value of the network is proportional to the
density, each new node adds value to it.

Competing with a these physicial network e ects demand a

large upfront investment. Physical networks are usually
utilities, such as telecommunications, transportation and
infrastructure (water, gas, electricity…), so they are able to
develop monopolies and even end up getting nationalized.

The proof of their strenght is exaclty this: many of these

companies have poor services, but keep leading the market -
simply because no one can compete with these winner-take-
all businesses.

2. Protocol (Direct)

A protocol network e ect comes up when some standard is

declared and all nodes (and their creators) can plug into it by
using that protocol. Great examples are Bitcoin and
Ethereum. Whenever a protocol is adopted, it is extremely
di cult to get replaced.

However, the protocol creator does not normally capture

most of the value on having the protocol adopted. It is quite
defensible, since it tends to get embedded in all products
that use the protocol.

3. Personal Utility (Direct)

These network e ects are base in two qualities: it is

indispensable for users’ lives and their personal identities are
tied to the network. They are typically used for people to
interact with their personal networks, so it is important for
them to be part of the network. Some examples include
WhatsApp, Slack and Skype.

17
ffi
ff
ff
ff
 4. Personal (Direct)

Personal network e ects occur when a good parte of people

one knows is using the same product, so they feel like their
personal identity and reputation are related to it, getting
in uenced to join the network.

It is not the same as Personal Utility because this one does

not involve an actual practical utility as the one before and
personal networks are less vital. Some examples are
Facebook and Twitter. As you may agree, they are not
essential, but still very strong, since they are a way to build
connections with others.

5. Market Networks (Direct)

They are a combination of the transactions from a

marketplace and the communication from a personal
network. Usually, market networks are result of a previous
o ine network, so the relationship among the nodes is
direct.

Therefore, the service prodvider here is a di erentiated

individual, not a commodity, enhancing long-term
relationships. Some examples are HoneyBook and Houzz.

2-SIDED NETWORK EFFECTS

Also called indirect network e ects, the 2-sided network is
composed of supply-side and demand-side users, who come
to the network for di erent reasons, building value for the
other side. For example, when there are buyers and sellers,
each new seller increases the o ers, and every new buyer is
a potential customer. Let’s go deeper:

18
ffl
fl
ff
ff
ff
ff
ff
 6. Marketplace (2-Sided)

A marketplace depends on buyers and sellers. To compete

with success marketplaces, there is necessary to have a
better value for both sides. The network is the greatest value,
by connecting them two.

It is highly defensible, but the weakness can arise from multi-

tenanting, when buyers and sellers use more than one
marketplace at the same time. That requires a to add some
“lock-in”, to avoid the phenomenon.
Some famous examples are: eBay, Amazon, Facebook,
Google, Tinder, Visa, and many others, especially
eCommerces and Media.

7. Platform (2-Sided)

The platform has a supply side and a demand side that create
value to each other but only through an intermediary
platform, which will, therefore, provide value for them both.
Microsoft OS, Android and PlatStation are great examples.

Unlike online marketplaces, the platform itself can play an

important role in the utility of it. But, just like marketplaces,
the challenge is to deal with multi-tenanting, for example
qhen developers create versions of the same app for both
iOS and Android. The great advantage here is the pricing,
di erent from marketplaces that are usually free.

8. Asymptotic Marketplace (2-Sided)

No two 2-sided marketplaces are the same. Sometimes, both

demand and supplu side grow together, sometimes (like
OpenTable case) the supply side has to get strong in order to
attract demand-side, and there are other times when the
supply side adds value quickly, but soon starts losing it.

19
ff
It is the case of ridesahring apps mentioned before, where
the biggest value for demand side was the reduced waiting
time. The problem is that, when it reaches a certain point,
there are virtually no di erence in waiting time, and that is not
a value anymore, requiring the business to reinvent and
create a new value proposition, to stand out.

The Asymptotic Marketplaces are, therefore, more vulnerable

to competition than other marketplaces. Besides, they can
also be susceptible to multi-tenanting. Many people use both
Lyft and Uber, for example, depending on the price.
9. Data Network E ects

Data networks e ects arise, of course, when the value

increases with more data added. In this network each user
(node) feeds useful data to the central database (central
node).

If this data is really bene cial to the users, the product

becomes powerful. That’s not true when data is marginal/
complementary to the product. And if the usage has no
relationship with useful data production, then there isn’t
network e ect, just scale (Experian, for example).

Waze is a great example of data network e ect: nearly every

one who consumes data on Waze contributes with useful
data at the same time, in real time. The larger the network,
the more accurate and up-to-date the data is..

10. Tech Performance Network E ects

For these network e ects, the more users or devices

connected to the network, the better the technology
becomes, making the product faster, cheaper or easier to
use. So, the more nodes, the better the whole network
performance. It is the case of BitTorrent, for instance. The

20
ff
ff
ff
ff
ff
fi
ff
ff
 more people downloading a le, the more seeding le to the
network.

“SOCIAL” NETWORK EFFECTS

Networks are nodes and links. Witn network among people,
our physical bodies are the nodes, and our words and
behaviors are the links. These social network e ects are able
to create more value in your product, the more people use it.

There are three main types of social network e ects:

language, belief, and bandwagon e ects. But That number
could easily expand, since human psychology, behaviors and
interactions are very complex and changeable. Let’s
understand each one:

12. Language

In all history, language has always been the main

intermediary between the nodes (humans) in the social
network. The concept can be extended to vernacular of
speci c groups and tribes.

Startups may bene t from these network e ects by creating

business category language or by naming a product or brand.
It is the case of terms like “portal” for Yahoo!, criptocurrency
for Bitcoin or for the actions of Googling or Xeroxing.

Making people verbally use your company ir a great

advantage. However, the language network e ects are kind
of weak, and the leader can easily lose the place for
competitors.

13. Belief

21
fi
fi
fi
ff
ff
ff
ff
ff
fi
Sharing common beliefs is essential to be part of a group and
to be accepted by the other members. That’s why beliefs
become more valuable as more people believe them. Taking
gold as an example: we can’t eat it or use ir for so many
things. But it is valuable, because we have believe so.

The same occurs for products like Bitcoins. The more people
believe it is valuable, the more valuable it actually gets. It is
self-reinforcing, indeed. Other good examples are religions
and ideologies.

14. Bandwagon

Bandwagoning os about the social pressure to join a network

to avoid being left out. In the late 1990s, you would be cool if
you change from Alta Vista to Google, for example.
Nowadays, it is the case of strong brands, as Apple, which
triggers the psychological need to be part of the cool kids.

These e ects can be frustrating for competitos who may

even have better products, but simply can’t beat bandwagon.
And some rare times, the bandwagon e ects can go too far: if
too many people join the network, then the early adopters
might leave it to avoid mainstreaming.

22
ff
ff
HIDDEN NETWORK EFFECTS

There are some hidden network e ects that can actually

create some unique advantages for a long term success. Let’s
take a look at some examples:

SLOW NETWORKS
Slow networks have a delay between the time they are
created and the moment they began to drive value indeed.
They typically have long product usage loops that slow down
the network e ects, and their value is usually underrated as
the advantages are not tangible right away. The slow
network’s e ects can take long to start coming up, even for
fast-growing businesses.

23
ff
ff
ff
 On the other hand, when these slow networks take their
market share, they are very di cult to displace. The
worldwide top universities, for instance, took years to build
their network e ects, and they are never weakening.

The secret here is to have patience and invest in necessary

resources once the slow network is identi ed, in order to
make it grow bigger. In general, when slow networks fail, it is
because the stakeholders don’t wait as long as necessary to
let them take place.

UNFINISHED NETWORKS
It occurs when the network keeps incomplete for some time,
to be able to build up the product or strategy. If you think
about food delivery apps, for example, you may notice that,
during the rst years, they just get restaurants together in the
app.

The goal is to have a good restaurant portfolio actually

launching it to the public. The problem is that these
un nished networks often keep un nished.

24
fi
fi
ff
ffi
fi
fi
 THROTTLED NETWORKS
They are similar to the un nished networks, but, while in the
un nished one, there is an important part missing before it
gets known, the throttled networks are actually nished, but
limited by a product feature or some strategic decision.

A famous example was Facebook in the early days, when the

users had to have a Harvard e-mail (and, later, an .edu e-mail)
to join the network. That ended up limiting the engagement.

Throttled networks are not exclusive, they are temporarily

small, because they can support a larger network, just not
immediately. This decision is made due to some technical or
operational constraint, for example. Or it can be the result of
poor management, execution or resources.

For knowing if the network is a throttled one, just waiting to

grow, it is necessary to check what would happen if the
current constraints can be relaxed. If the answer is positive or
neutral, then you may be sure about it.

LATENT NETWORKS
This occurs when the network comes before the actual
product. Therefore, no one is actually aware of their power
until it is too late to replace or compete with it. Usually, it
starts with a community which looks like a network, driving
value for the members.

Then, at some time, the product is introduced in the

community and it changes or ampli es the way the network
engages. Because, before that, there was no way of
monetizing the network. But then, some initial robust
communities can eventually become networks.

25
fi
fi
fi
fi
Nevertheless, the latent network are the hardest to predict,
because sometimes these communities are just an audience,
i.e., they perceive value simply from the central node, not the
whole network. Thus, the product will scale basically as a
linear regular business. In simple words: a network engages,
an audience consumes.

HIDDEN ADVANTAGES
Hidden networks should be considered for starting, investing
and working at. Because building network e ects is actually a
race, which you must conquer before your competition. A
hidden network may be an advantage, as you can get there
before your competitors even notice. You are already running
and the others don’t even know the race has started!

26
ff
STRATEGIES FOR BUILDING NETWORK
EFFECTS METRICS
1. Bowling pin strategy

This is a term coined by Geo rey Moore, and it means

entering the market by starting with one single pin (one
category or niche) and, then, move to the next pin, taking one
by one over.

2. Bootstrapping growth

The company creates and/or provides its rst contents or

members, in order to attract new users and make the network
grow.

3. Setting goals

It is crucial that the company knows the growth rate it needs

to meet, in order to maintain and then grow the network,
because when this rate is not met, the network can actually
die.

4. Building triggers

It is also imperative that the company constantly develop new

features to innovate and sustain the engagement to the
platform, for users not to lose interest.

5. Leveraging irregular network topologies

27
ff
fi
 It occurs when the brand or company nds clusters inside the
market and takes advantage of that by reaching critical mass
within these clusters and, then, expands them.

6. Attracting the harder side of a two-sided market

In every two-sided market, one of the sides is always more

di cult to reach than the other. A common way to solve the
problem is subsidizing the harder side, usually by reducing
prices. In order to do it, it is essential to also understand
which side the money and the value come from.

7. Showing long-term commitment

For the users and developers to feel comfortable about

investing their time (or even money) on the platform, it is
important that they feel safe about it. Therefore, you must
show it is a long-term relationship.

8. Providing standalone value of the base

Sometimes the network e ects can be improved and

increased with complimentary value. Then, the base is
strengthened by the complements and they both keep
running for long. It was the case of VCR and its
videocassettes.

9. Integrating vertically into complements

It is when the same company integrates vertically the base

product and the complement products, such as video gaming
businesses usually do. This way, they ensure proper supply of
both goods, and are not dependent on outside developers.

28
ffi
ff
fi
NETWORK EFFECTS METRICS
To take advantage of your company’s network e ects, it is
necessary to understand what’s working or not. The best way
to do it is by de ning a set of metrics able to measure the
performance. Take a look:

ACQUISITION
Organic vs. paid users

For successful network e ects, the amount of organic users

must keep increasing over time, since the business becomes
so attractive that it is able to make people willing to join on
their own. That doesn’t mean you can’t invest money to
acquire new users, but the growth has to be sustainable.

Sources of tra c

Just as the organic users, successful networks have to

become a destination, a space where users want to spend
time on. So, it’s important to measure how much tra c is
direct vs. the amount which comes from external sources.
More direct tra c usually means more value.

Time series of paid CAC

Once the network e ects arise, paid CAC should decline over
time. However, when it comes to the supply side, that
depends on several factors, such as the competitiveness,
marketing, channels, availability of substitutes, etc.
29
ffi
ffi
fi
ff
ff
ff
ffi
 For example, regarding ridesharing, there are many
possibilities to replace the supply - the drivers -, so it became
more expensive to acquire the supply. But, for platforms like
OpenTable, the supply became cheaper over time.

COMPETITOR-RELATED METRICS
Prevalence of multi-tenanting

Multi-tenanting, as mentioned above, occurs when your users

are using other services just like yours, at the same time. This
phenomenon can reduce usage and, consequently, compress
margins for all the competitors in the market.

To measure multi-tenanting is not simple. You might need to

ask your users if they use another similar service, research
churn and declines in usage, or even look for your users’
pro les on other platforms. But, more than that, you have to
get to know whether they just keep a pro le or use the other
platforms actively.

Switching or multi-homing costs

Other metrics that you may need to understand is how easy it

is for your users to sign up at a competing network. For
example, some products require upfront investment.
Although that may make it a little harder to acquire new
users, once they are active it is less likely that they change
platforms (because they would have to pay again).

On the other hand, products that demand less energy or

investment may su er with switch-over. For example, Uber
already had millions of users’ information, when they
launched UberEats. So, it was not di cult at all for Uber riders

30
fi
ff
ffi
fi
to begin ordering food through the app, even if they signed
up to other delivery platforms.

Potential metrics here could be the time required to complete

registers and pro les on a competing platform or how simple
it is to reach the minimum threshold or a product to be useful
(e.g. 10 friends for Facebook), for example.

ENGAGEMENT-RELATED METRICS
User retention cohorts

As the value of a network increases with the number of users,

this value should be re ected in retention cohorts. The newer
cohorts use the network when it is already large and,
consequently, more useful. So, the logic is that they should
have a better retention tha older cohorts, who joined the
network in the early days, when it was smaller.

However, it is common to verify a greater retention of older

cohorts, since they are usually early adopters, more
motivated and, in general, the ideal customers for the
product. In cases like that, there is a declining cohort
retention over time.

The metrics can also be in uenced by the presence of

competitors, hyperlocal network e ects, and even negative
network e ects.

Core action retention cohorts

The core action of your product is the one that drives value to
the customer. The more users the network gets, it is likely to
have retention improved, anchored on the core action. This
retention is one of the most important metrics you must
check, much more than longins or opens.
31
ff
fi
fl
fl
ff
 Dollar retention & paid user retention cohorts

Paying for a product indicates how much the users value it.
And a product with network e ects becomes more valuable
over time. Consequently, it is supposed to have increasing
paid user retention among newer cohorts.

Retention by location/geography

When it comes to location, it is fair to say that the network

e ects exist on a per-market basis. That means every new
geography is a new reality regarding network e ects. Each
location will build network density, thus, the older or more
established the market is, the better retention it must have.

Power user curves

Power user curves (aka L30 charts for 30 days of use, or L7

charts for 7 days of use) register users’ engagement, by
showing the number of days the users have been doing a
particular action in a given period.

These metrics allow you to see when a product is in fact

gaining utility with more users (the network e ects). Because
when a product drives more value with more users, then
there should be a more right leaning power user curve, over
time.

MARKETPLACE METRICS
Match rate

Marketplaces and two-sided platforms have to provide the

matching of supply and demand. There lies the importance of
measuring the match rate, which shows how buyers nd
sellers and vice versa.
32
ff
ff
ff
ff
fi
 Another way to measure it is actually nding the opposite, by
measuring the zeros. For example: how many users open the
app but don’t conclude the transaction/order? Then, it is
necessary to understand the reason why the match didn’t
take place, in order to remove the impediment.

This metric is also related to multi-tenanting: when the match

rate is low, it is more likely to make users migrate elsewhere.

Market depth

Market depth or o er depth is the ability the market has to

sustain large orders without price movements. The higher the
numbers of transactions at each price, the greater the depth
of the market.

Market depth impacts directly the user’s experience. It is

essential to make it easy for participants to nd and match
the other side. Failing that can result in negative network
e ects, and the conversion rates may fall.

Time to nd a match

The metrics now is to verify how long it takes for supply to

match demand. Marketplaces usually have a curve for match
rate - the inventory turnover. For example, how many days for
the buyer to purchase from the seller, or for an employer to
nd an employee, etc.

Concentration or fragmentation of supply and demand

The greater the fragmentation on both supply and demand

sides in a marketplace, the more valuable and defensible it
gets. Because no participant accounts for a much higher
share of transactions, making the business sustainable and
diversi ed. When there is a big concentration, it is more likely
33
fi
ff
fi
fi
ff
fi
fi
for a large buyer or seller to take a great share of transactions
if they leave the platform.

ECONOMICS-RELATED METRICS
Pricing power

This metric refers to how much your customers are willing to

pay to be part of the network. If they receive greater value,
they are likely to pay more, in general. That’s why, over time,
the businesses can evolve from not being monetized to
turning on monetization and to increasing the prices with
minimal churn.

Unit economics

Sustainable and successful network e ects often appear in

improved unit economics over time, due to declining
incentives that businesses need to o er to di erent sides of
the market, lower share of paid users, and general
improvement in pricing power.

For local network e ects, the impact should show up over

time, on a market-by-market basis, because CAC should
decrease and the organic share of users should grow over
time.

34
ff
ff
ff
ff
CASE STUDIES OF COMPANIES WITH
NETWORKS EFFECTS

FACEBOOK
- Product value: started as an online student directory with
useful information for a single player. Then, it became a
way for college students to connect with each other.

- Growth tactics: accessed the whole Harvard directory,

thus driving early adoption. Due to its viral character, it
spread organically from one user to another.

- Engagement triggers: connecting a new user to 10 friends

within two weeks was key for retention. Engagement was
driven through e-mail contact imports, suggested friends
and embedded widgets.

- Network e ects: it kept improving the product, adding

new features, in order to get new users to join and
maintain retention. This way, not only the number of users
would increase, but also the usage rate.

AIRBNB
- Product value: spaces 30% to 80% cheaper than hotels,
and a more social type of inventory.

35
ff
 - Growth tactics: traditional and alternative marketing to
advertise in cities with sold-out events and constrained
hotel supply.

- Critical mass: photography services to o er more

appealing options to guests. Mutual connections to check
who else have stayed in the place, thus building trust in
the platform.

- Network e ects: more guests stayed in more places

(building demand), which attracted new hosts to o er new
places (building supply). In short words, Airbnb focused on
building the demand side, to attract suppliers.

MEDIUM
- Product value: the best web editor for both experienced
and inexperienced writers, with an elegante user-friendly
publishing tool.

- Growth tactics: curated special content contributors to

create perceived exclusivity and, thus, attract other
in uencers. Used the 1-9-90 internet rule - 1% active
writers, 9% participants who edit and 90% users who read.

- Engagement triggers: the platform is optimized for

engagement, thorough an in-content interactor that
recommends, responds and mentions. It also used
taxonomy to cluster highly engaged communities around
topics of interest.

- Network e ects: a network of people and their ideas, with

more people writing and more people reading on
Medium.

36
fl
ff
ff
ff
ff
WHATSAPP
- Product value: instant messaging, better and simpler than
SMS, for free.

- Growth tactics: started small with one little community

before spreading out. Each user invited users from their
phone contacts.

- Engagement triggers: Group Chats, which permitted

going beyond pairwise interactions. Multimedia, which
made people share photos and other media, especially in
places where people didn’t use many web-based apps.

- Network e ects: WhatsApp had both growth and

engagement, since more users added more value for
other users.

37
ff
WHEN NETWORK EFFECTS ARE NOT
ENOUGH
Before scaling, the objective for marketplaces is to create a
value proposition for both demand and supply. Besides, they
need to build trust and interest to maintain them on the
platform.

The problem is: to attract buyers, you need a critical mass of

suppliers, and to attract suppliers, you also need a good
amount of buyers. The classical chicken-and-egg problem.
But the challenge is bigger than that. There are several other
aspects that can derail these platforms.

You don’t have to grow quickly

When you rush to be the rst and chase early growth, you
may not have time to prove your value to both buyers and
sellers, leaving the market open to competition. However,
when demand has access to a nice o er of supply, and
supply earns attractive pro ts, more buyers will bring more
sellers and vice versa, building up strong network e ects.

Startups will inevitably have aws, which will take some time
to be identi ed and corrected. But when growth is too fast,
the impact of these aws is ampli ed, turning them even
harder to be xed. And, if the business model needs some
kind of change, that may ruin the business.

38
fi
fi
fl
fi
fi
fl
fi
ff
ff
You must care about safety

Theoretically, the marketplace is not responsible for the

quality of products and services that are transacted via its
platform. Nevertheless, for having a nice reputation and for
building con dence, it is necessary to o er the users (both
supply and demand sides) some security, by minimizing
improper behavior, misquali ed products or even frauds.

That’s why the systems for rating and reviewing are used so
frequently. However, these systems are not enough to avoid
all the problemas or misservice. And, when something goes
wrong, it is common for the users to blame the marketplace
too, never returning to buy through the platform, harming all
the other suppliers.

Moreover, when the transactions involve more expensive

goods, such as automobiles or housing, the marketplaces are
used to providing insurance. Airbnb, for example, insures
hosts of up to $1 mi against property damage.

You don’t have to overestimate disintermediation

Marketplaces’ owners use to fear buyers and sellers will have

new transactions outside the platform. This is especially risky
for high-value and/or recurring transactions. But greate
marketplaces have not been severely hindered by this
threat.

Truth is most participants prefer to conduct business in a

safer space, which will reduce search timing and allow
transactions to be more secure and comfortable. As long as
the marketplace o ers value, there is no reason for both
sides to conduct their business outside. Besides, penalties
are not good for your reputation and algorithms for detecting
transactions outside are complicated and expensive.

39
fi
ff
fi
ff
You must pay attention to regulations

Marketplaces face serious regulatory issues due to their

unconventional businesses. But, when it comes to
regulations, no extreme is a good idea. Ignoring the
regulations tends to create a bigger problem - maybe
incorrigible - in the future. But trying to deal with every frame
is too much for beginners.

The right approach is somewhere in the middle, attempting to

engage regulators, without slowing down. It is important to
consult an attorney before launching, to understand all
relevant laws, and initiate a dialogue with regulators to best
develop the service. Even if the regulations are simply
inconveniente for the company, they cannot be ignored.

Besides, the media can be one of your worst enemies, so

develop a clear way to de ne your business and show it to
the world, before your competition or media do. On the other
hand, the greatest support you may have is having your users
by your side: they are voters and taxpayers who will resent
government interference with a service they enjoy.

Nevertheless, when there is no way to escape, the best to do

is to face the scenario, even when that means higher costs.

40
fi
CONCLUSION
It is fair to a rm that online marketplaces are changing the
nature of commercial transactions, working relationships,
communication and more.

Over the last years, countless platforms have shown a huge

variety of products and services, probably never seen before.
And new technologies keep enhancing the experience for
both supply and demand.

The growing amount of online marketplaces may probably

cause traditional transactional structures to gradually
transform (or shrink), and we may have a much more exible
marketing environment, also more able to quickly understand
the needs of customers.

41
ffi
fl
 REFERENCES

The following references were consulted to create this Super

Guide:

- https://medium.com/evergreen-business-weekly/the-
power-of-network-e ects-why-they-make-such-valuable-
companies-and-how-to-harness-them-5d3 c3659f8

- https://pt.slideshare.net/a16z/network-e ects-59206938/2-
What_arenetwork_e ectsProperties_termsand_laws

- https://www.investopedia.com/terms/n/network-e ect.asp

- https://pt.slideshare.net/a16z/network-
e ects-59206938/65-
What_arenetwork_e ectsWhat_arentnetwork_e ectsProp
erties

- https://www.applicoinc.com/blog/network-e ects/

- https://www.forbes.com/sites/forbescoachescouncil/
2018/01/02/how-to-harness-the-power-of-network-e ects/?
sh=12c6bd5762e8

- https://a16z.com/2018/12/13/network-e ects-dynamics-in-
practice/

- https://www.nfx.com/post/network-e ects-manual/

- https://a16z.com/2019/07/29/hidden-networks-e ects/

42
ff
ff
ff
ff
ff
ff
ff
fb
ff
ff
ff
ff
ff
 - https://pt.slideshare.net/a16z/network-
e ects-59206938/49-
What_arenetwork_e ectsStrategies_for_buildingnetwork

- https://a16z.com/2018/12/13/16-metrics-network-e ects/

- https://hbr.org/2016/04/network-e ects-arent-enough

43
ff
ff
ff
ff
ABOUT THE AUTHOR

Daniel Pereira is a Brazilian-Canadian entrepreneur that has

been designing and analyzing business models for over 15
years. You can read more about his journey as a Business
Model Analyst here.

E-mail Daniel if you have any questions at:

daniel@businessmodelanalyst.com

You can connect with Daniel at Linkedin:

https://www.linkedin.com/in/dpereirabr/
44
THE BUSINESS
MODEL ANALYST
45