Transcript NFRS Seminar - Harvard Business School

Architectural Strategy and
Open/Distributed Innovation
Carliss Y. Baldwin
DRUID/Scanscor Conference on Distributed Innovation
Stanford, CA
March 27, 2008
Slide 1
© Carliss Y. Baldwin 2008
Formal Models of Open/Distributed
Innovation

Johnson 2002
– Private provision of public goods

Bessen 2002
– User innovation addresses complexity of needs

Harhoff-Henkel-von Hippel 2004
– Reasons for free revealing

Henkel 2004
– Why firms collaborate on open innovation projects

Baldwin-Clark 2006
– Conditions for collaborative user innovation to be a self-sustaining
equilibrium
– Why the institution is “open”
Slide 2
© Carliss Y. Baldwin 2008
Formal Models of Open/Distributed
Innovation (cont)

Baldwin-Hienerth-von Hippel 2006
– Collaboration dominates isolation for user-innovators
– Commercialization of user innovations

Baldwin 2008
– Open innovation as an “unencapsulated” task network in contrast
to corporations which are “encapsulated” by transactions

This paper
– Open innovation as a complement to a firm’s competitive strategy

Baldwin-von Hippel (on hold)
– Competition between collaborative open innovation and
proprietary innovation based on property rights
Slide 3
© Carliss Y. Baldwin 2008
Open innovation as a complement to
a firm’s competitive strategy
Task Network for a complex
system (Baldwin, 2008)
Boxes are activities/components;
Lines are dependencies/interfaces
Slide 4
© Carliss Y. Baldwin 2008
Two forms of organization
Task Network for a complex
system
F
All in One
Big Firm
Boxes are activities/components;
Lines are dependencies/interfaces
Slide 5
© Carliss Y. Baldwin 2008
Two forms of organization
All in one big Firm
Distributed Innovation
(and Production)
F
F
U
U
F
U
U
F
F
U
F = Firms;
U=Users in an open
innovation setting
Slide 6
© Carliss Y. Baldwin 2008
My questions
 Can
a firm gain advantage from a technical
system based on distributed innovation and
production?
 When and why would a firm set up its
technical system in this hybrid fashion?
Such firms will be collaborators, supporters
and perhaps predators of systems of
open/distributed innovation
Slide 7
© Carliss Y. Baldwin 2008
My Main Argument
(Remember this slide)
“B’s” Technical System
“A’s” Technical System
A
U
U
B
F
U
U
F
F
U
–With selective outsourcing,
A’s system can beat B’s
system, causing B to exit
–Open/distributed innovation
and open standards are a key
part of A’s system
Slide 8
© Carliss Y. Baldwin 2008
Potential Examples
 Google
– “sucking up” and subsidizing the creation of
open source code
 IBM
and other corporate sponsors of Linux
 Platforms in two-sided markets
– E-Bay; NTT-Docomo
 Sun
Microsystems
– Ancient history, but interesting
Slide 9
© Carliss Y. Baldwin 2008
The Sun story
 Two
parts
– What they knew
» Their capabilities
– What they did
» Deployment of capabilities
 Both
parts of the story are essential to my
argument
Slide 10
© Carliss Y. Baldwin 2008
Their capabilities—what they knew
 Third
generation of computer architecture
– Hennessy and Patterson: quantitative analysis
of bottlenecks in a complex system
– Hardware-software interactions
– “Make the common case fast…”
 Textbook
lags
– Stanford-Berkeley vs. Boston
Slide 11
© Carliss Y. Baldwin 2008
Example—Sun 2
Bottleneck
remedies
• Patented MMU
chip
• Fast bus
Slide 12
© Carliss Y. Baldwin 2008
Deployment of capabilities—
what they did
 To
have a superior product, you only need
to control the bottleneck(s) and make them
better
 All other parts of the system architecture
– Don’t affect performance or cost very much
Strategy: Keep control of bottlenecks and let go
(outsource) the rest
Slide 13
© Carliss Y. Baldwin 2008
Comparative “Footprints”
 Apollo
(main competitor)
Apollo Computer
Sun Microsystems
Aegis proprietary
Operating Sy stem
Inhouse
Design
OS
Hardware Design
DN series = 3-4 boards incl.
IO and Display controllers,
Power supply
Purchase Components
Hardware
x
Component Test x
Kits x
Board stuf f and Solder x
Test Boards x
Board Assembly x
Sy stem Assembly x
Sy stem Test x
Quality Assurance x
Consolidate and Ship x
O
DOMAIN proprietary
Network Architecture
Network
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
highly interdependent task blocks
Footprint (tasks perf ormed inhouse).
Inhouse
Manufacturing
x
Key:
x= transf er of material or inf ormation f rom column
task to row task;
T= transaction: sale of good by column owner to row
owner;
O= outsourced task blocks;
Slide 14
 Sun
x
x
x
x
x
x
BSD Unix
Operating Sy stem
O
Ethernet
Network Architecture
O
Standard IO and Display controllers
O
Power supply
x x
x x x
Customize Unix
x x
x x x
Inhouse
Proprietary MMU
x x
x x x
Design
Internal bus
x x
Single Board Lay out
T
T T
T x x x x
Purchase Components
Component Test x x x x x
O
Kits x x x x
T
ManuBoard stuf f and Solder x x x x
x O
facturing
Test Boards x x x x
T
Board Assembly x x x x
x
Sy stem Assembly x x x x
Sy stem Test x x x x
x
Quality Assurance x x x x
x
Consolidate and Ship x x x x
x
Opportunistic commoditization—
Notice selective use of open standards
(Ethernet) and open source code (BSD
Unix) to make non-bottleneck
components into commodities
© Carliss Y. Baldwin 2008
Result => Smaller Technical Core
=> Smaller “Footprint”
 Fewer
inhouse activities
– Relative to competitors that don’t have same
architectural knowledge…
– Remember textbook lags!
 No
penalty in performance and cost
– Because of the optionality of modular designs
– Architects can select the best treatment for each
module
Slide 15
© Carliss Y. Baldwin 2008
When outsourcing…

The focal firm must maintain its power in the
supply chain
This means…

Competitive production=>lots of potential
suppliers
– Board stuffers

Competitive innovation=>Open standards, open
source, and open innovation
– Berkeley Unix
Slide 16
© Carliss Y. Baldwin 2008
Smaller “Footprint” means…
Fewer inhouse activities
 No penalty in performance and cost
Means…
 Equal machines with same unit cost
 But less investment in R&D, Net Working Capital and
Fixed Assets
Means…
 Higher Return on Invested Capital ROIC at any level of
pricing

– Structural advantage based on design of the technical system
Slide 17
© Carliss Y. Baldwin 2008
Competitive Dynamics
 Higher
ROIC is a loaded gun pointed at
competition
 Competitor may survive for a while, may
even grow
 But eventually must exit from the market
See paper for many gory details…
Slide 18
© Carliss Y. Baldwin 2008
Empirical Predictions
If A (Sun) pursues a smaller footprint strategy
against B (Apollo) —
1.
2.
3.
4.
ROICA> ROICB
gA > gB
Eventually ROICB< Cost of Capital
A can drive B out of the market
“B” can be a set of firms, not just one, but A
must have superior architectural
knowledge about the bottleneck
Slide 19
© Carliss Y. Baldwin 2008
ROIC
40%
30%
Sun
20%
10%
0%
Apollo
-10%
-20%
-30%
-40%
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
1985 1985 1985 1986 1986 1986 1986 1987 1987 1987 1987 1988 1988 1988 1988 1989
Slide 20
© Carliss Y. Baldwin 2008
Growth
500
400
Sun
$ millions
300
200
Apollo
100
0
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
1985 1985 1985 1986 1986 1986 1986 1987 1987 1987 1987 1988 1988 1988 1988 1989
Time
Slide 21
© Carliss Y. Baldwin 2008
Exit
 1985-1988 Apollo’s
ROIC was consistently
below their cost of capital
 Acquired by Hewlett Packard in April 1989
– Avoided bankruptcy
– HP’s rationale— Economies of scale
» “the largest engineering workstation company in the
world…”
 HP was
clueless about bottlenecks,
footprints and ROIC
Slide 22
© Carliss Y. Baldwin 2008
Strategic Implications
 If
you are a “B” type firm, don’t go to war
against an “A”
 If you are an “A” type firm
– Know your bottlenecks
» Must know more than you make (Brusoni and
Prencipe)
» Be prepared to shift footprint as bottlenecks move
around
– Maintain your power in the supply chain
– Make sure your “B” type competitors know that
you are an “A” (publish your ROICs!)
Slide 23
© Carliss Y. Baldwin 2008
Recap of Argument
A rc hitec tural
Knowledge
(bottlenec ks
& potential
modules )
P ower
in the
S upply
C hain
Slide 24
S maller
'F ootprint'
through
s elec tive
outs ourc ing
I nves ted
C apital
A dvantage
L ong- run
M arket
D ominanc e
© Carliss Y. Baldwin 2008
Recap of Argument
A rc hitec tural
Knowledge
(bottlenec ks
& potential
modules )
P ower
in the
S upply
C hain
Slide 25
S maller
'F ootprint'
through
s elec tive
outs ourc ing
I nves ted
C apital
A dvantage
L ong- run
M arket
D ominanc e
© Carliss Y. Baldwin 2008
Recap of Argument
A rc hitec tural
Knowledge
(bottlenec ks
& potential
modules )
P ower
in the
S upply
C hain
Slide 26
S maller
'F ootprint'
through
s elec tive
outs ourc ing
I nves ted
C apital
A dvantage
L ong- run
M arket
D ominanc e
© Carliss Y. Baldwin 2008
Recap of Argument
A rc hitec tural
Knowledge
(bottlenec ks
& potential
modules )
P ower
in the
S upply
C hain
S maller
'F ootprint'
through
s elec tive
outs ourc ing
I nves ted
C apital
A dvantage
L ong- run
M arket
D ominanc e
Power in the Supply Chain is a strategic complement to
Architectural Knowledge
Slide 27
© Carliss Y. Baldwin 2008
Significance for
Distributed/Open Innovation
 Certain
firms will be happy to build
strategies on top of open/distributed
innovation systems
 The economic benefits of open/distributed
innovation may be appropriated by strategic
complementors pursuing “small footprint”
strategies
Slide 28
© Carliss Y. Baldwin 2008
Distributed/Open Innovation
(cont)

As open innovation takes over different layers of a
technical system, the overall system price may not
decline
– Utility to users may increase, however
» if user information is “sticky”
» hence user designs more functional

Firms will seek bottlenecks
– Whether they find them depends on
» architectural knowledge and
» textbook lags
Slide 29
© Carliss Y. Baldwin 2008
Distributed/Open Innovation
(cont)
 Sources
of bottlenecks
– Technical determinants of performance (discussed in the
paper)
» Absolute
» Fractional
– Transactional efficiencies—Indirect network
externalities—(two-sided markets literature)
» Standards of compatibility (design rules)


Interoperability (artifact level)
Design of transactions (agent/organization level)
» Search funnels and filters
 Systems
can have multiple bottlenecks, causing firms
to bet on different footprints
– Mobile telephony today
Slide 30
© Carliss Y. Baldwin 2008
Can one break a bottleneck?

As architectural knowledge diffuses, technical
bottlenecks disappear
– Memory management for CPU

In contrast, standards of compatibility tend to be
very stable—a chokepoint, not a bottleneck
– Windows APIs (owned by Microsoft)
– New York Stock Exchange (owned by members/users)

Don’t know about search funnels/filters
– Google
Slide 31
© Carliss Y. Baldwin 2008
In closing…
Slide 32
© Carliss Y. Baldwin 2008
Remember this slide?
“B’s” Technical System
“A’s” Technical System
A
U
U
B
Slide 33
F
U
U
F
F
U
© Carliss Y. Baldwin 2008
My Main Argument
“B’s” Technical System
“A’s” Technical System
A
U
U
B
F
U
U
F
F
U
–With selective outsourcing,
A’s system can beat B’s
causing B to exit
–Open/distributed innovation
and open standards are a key
part of A’s system
Slide 34
© Carliss Y. Baldwin 2008
I hope you are convinced.
Slide 35
© Carliss Y. Baldwin 2008
Thank you!
Slide 36
© Carliss Y. Baldwin 2008
Architectural Knowledge
 Architecture
= entities and relationships
– Function-to-component mapping
– Interfaces between components
– Linkages and interactions (“dependencies”)
 Architectural
knowledge means knowledge
about all these things
Slide 37
© Carliss Y. Baldwin 2008
Dynamics —Textbook lags
 1962—IBM
task group figures out how to
build a modular computer system
 1974—Bell and Newell publish textbook
 1980—Hennessy and Patterson begin to
teach graduate students about quantitative
approaches to computer architecture
 1990 H&P publish first text
 1994 H&P publish second text
Slide 38
© Carliss Y. Baldwin 2008
Architectural Innovation
 Generation
1—Integral systems
– Build a whole new system
 Generation
2—Modular systems
– Modular operators: split, substitute, augment,
exclude, invert, port
– Recombine, link, compose
 Generation
systems
3—Quantitatively measured
– Find a bottleneck and remedy it
– “Make the common case fast” (Amdahl’s Law)
Slide 39
© Carliss Y. Baldwin 2008
Third-generation architectural
knowledge tells you
Where bottlenecks are
 How to remedy a bottleneck
 How much the remedy is worth in terms of system
performance

– “Speedup formula”

How to change modular structure
– 2nd generation knowledge

Can have multiple objectives (cost and speed)
– Multiple bottlenecks
Slide 40
© Carliss Y. Baldwin 2008