Networks and Positive Feedback - Faculty Directory | Berkeley-Haas

advertisement
Networks and Positive
Feedback
John Morgan
Important Ideas



Positive feedback
Network effects
Returns to scale


Demand side
Supply side
Positive Feedback

How system adjusts to perturbations



Negative feedback: stabilizing
Positive feedback: destabilizing
Positive feedback makes a market
“tippy” – winner take all


Examples: VHS v. Beta, Wintel v. Apple,
56k modems, DVD v DivX
“Winner-take-all markets”
Sources of Positive Feedback

Supply side economies of scale




Declining unit cost of supply
Marginal cost less than average cost
Example: information goods are all fixed cost
Demand side economies of scale



Increasing value to users as market share increases
A single standard: fax, email, Web
Competing standards: VHS v. Beta, Wintel v. Apple, 56k
modems, Codec “standards” in Digital Audio
Color fax machine




You have a new technology for color fax
machines
License it to many sellers
Technology can only be used by sender
if receiver has color fax machine
So there is a strong network effect--value to user of technology depends on
number of adopters
Setup





Indicate your adoption of the technology by standing
up
The number on your card shows that value v (to you)
of adopting if you are the only adopter
In n people (including you) adopt, your value is nv
I will write the price of the fax machine on the board
Adopt when your value (strictly) exceeds the price.
Analysis

Standalone values were 10,9,8,…,1
Analysis



vn = p : marginal user wants to adopt
n=10-v :everyone with greater value than
marginal user adopts
Combine to give demand=supply condition





(10-n)n=p
n=1, (10-1)1=9
n=2, (10-2)2=16
n=3, (10-3)3=21… n=9, (10-9)9=9
Takeoff occurs when p=9
Adoption Dynamics
Saturation
Number of
Users
Takeoff
Launch
Time
Network Effects

Demand side economies of scale often
referred to as network effects



Real networks – telecom
Virtual networks – network of users
Metcalfe’s Law: Value of network of
size n proportional to n2

Experiment is an example of this
Expectations Matter


What is the difference between success and
failure in tippy markets?
Expectations Matter:


I want to join network that I expect to succeed
Central part of successful marketing is setting
appropriate expectations



Commitment can help
Old economy example: Banks
New economy example: Microsoft version 1.0
products
Virtuous Cycles and Death
Spirals: 1-2-3 v Visicalc
Value to User
Virtuous cycle
“Death spiral”
Number of Users
Direct and indirect Network
Effects

Value to me depends directly on
number of adopters


Fax machine, telephone, email
Value to me depends on adoption of
some complementary product



DVD player/ DVD disks
eBook reader + content
Palms and software
Are All Information Markets
“Tippy”?




AOL
Dell
Google
What are the key drivers of network
effects?
Network Effects–Likelihood of
Tipping
Low Scale High Scale
Economies Economies
Low Demand
For Variety
Unlikely
High
High Demand
For Variety
Low
Depends
Chicken & Eggs




Fax and fax machines
VCRs and tapes
DVDs: A more complicated problem
Online dating
Igniting Positive Feedback

Evolution


Give up some performance to ensure
compatibility, thus easing consumer
adoption
Revolution

Wipe the slate clean and come up with the
best product possible
Compatibility versus
Performance
Compatibility
Evolution
Revolution
Performance
Evolution


Offer a migration path
Examples




Microsoft Windows
Intel
Borland v Lotus
Build new network by links to old one
Important Considerations


Use creative design for migration
Think in terms of whole system
Compatibility versus
Performance
Compatibility
Migration “sweet spot”
Evolution
Revolution
Performance
Technical Obstacles

Converters and bridge technologies



One-way compatibility or two way?
Windows for Wordperfect users
Importance of UI
Legal Obstacles


May need IP licensing
Example: Sony and Philips had
advantage since they held the patents
on CDs

DVD players usually play CDs as well
Revolution



Grove’s law: “10X rule”
But depends on switching costs
Examples: Nintendo, Iomega Zip, DVD,
CD
Openness v. Control


Your reward = Total added to industry x
your share
Value added to industry



Depends on product and
Size of network
Your share

Depends on how open technology is
Openness

Full openness




Anybody can make the product
Problem: no champion
Unix v BSD v Linux
Alliance



Only members of alliance can use
Problem: holding alliance together
Conflicting incentives: China and DVDs
Control



Control the standard and go it alone
If several try this strategy, may lead to
standards wars
iPod/iTunes v MP3 players
Generic Strategies
Compatible
Control
Open
Controlled
Migration
Open
Migration
Incompatible Performance Discontinuity
Play
Lessons





Positive feedback means strong get
stronger and weak get weaker
Consumers value size of network
Works for large networks, against small
ones
Consumer expectations are critical
Fundamental tradeoff: performance and
compatibility
Lessons, continued


Fundamental tradeoff: openness and
control
Generic strategies





Performance play
Controlled Migration
Open Migration
Discontinuity
Lessons of history
Download