NFRS Seminar - Harvard Business School

Download Report

Transcript NFRS Seminar - Harvard Business School

Designs and Design Architecture: The Missing Link between “Knowledge” and “the Economy”

Carliss Y. Baldwin Harvard Business School Conference on Networks of Knowledge Brussels, Belgium, EU June 7, 2004 Slide 1 © Carliss Y. Baldwin and Kim B. Clark, 2004

First Example—

Vertical-to-Horizontal Transitions, Modular Clusters, Ecosystems Slide 2 © Carliss Y. Baldwin and Kim B. Clark, 2004

Modularity creates financial forces that can change the structure of an industry

 In 1995, Andy Grove described a vertical-to horizontal transition in the computer industry:

“Vertical Silos”

Slide 3

“Modular Cluster”

© Carliss Y. Baldwin and Kim B. Clark, 2004

The Computer Cluster, 1996

180 160 140 120 20 0 60 40 100 80 50 53 56 59 62 65 68 71 74 77 80 83 86 89 92 95 Slide 4 © Carliss Y. Baldwin and Kim B. Clark, 2004

The Computer Cluster, updated to 2002

800 700 600 500 400 300 200 100 0 50 54 58 62 66 70 74 78 82 86 90 94 98 02 Slide 5 © Carliss Y. Baldwin and Kim B. Clark, 2004

The Bright Side of Modularity

Slide 6 © Carliss Y. Baldwin and Kim B. Clark, 2004

The Dark Side …

4500 4000 3500 3000 2500 2000 1500 1000 500 0 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 Slide 7 © Carliss Y. Baldwin and Kim B. Clark, 2004

What Causes V-to-H Transition?

 Look to the underlying designs  Designs and their Architecture “lie between” Knowledge and the Economy Knowledge Design Architecture Designs Economy Slide 8 © Carliss Y. Baldwin and Kim B. Clark, 2004

What Causes V-to-H Transition?

 Look to the underlying designs  Designs and their Architecture “lie between” Knowledge and the Economy Knowledge Design Architecture Designs Economy  Modular, “option-rich” design architectures precede and enable modular clusters Slide 9 © Carliss Y. Baldwin and Kim B. Clark, 2004

Modularity

 Module = a set of tasks – separable from others; – with additive incremental value –

Unit of design substitution

– No. of modules = j

Global Design Rules

Slide 10

Module A Module B Module C Module D

© Carliss Y. Baldwin and Kim B. Clark, 2004

Virtues of Modularity

 1st Virtue: decentralizes knowledge and distributes action – Makes more complexity manageable – Enables parallel work  2nd Virtue: tolerates uncertainty – “Welcomes experimentation” –

—>

Creates Options – Property of “evolvability” Slide 11 © Carliss Y. Baldwin and Kim B. Clark, 2004

Option Value

 Design process is a search under uncertainty 

Design substitution is

optional

Versions

are evidence of option values being realized over time

Global Design Rules v.1

Version 1.2

Version 1.5

Version 1.0

Version 1.8

Slide 12 © Carliss Y. Baldwin and Kim B. Clark, 2004

Modularity and Option Values are “architectural properties” because

(1) They are observable in early and incomplete systems; and (2) They affect the way the system evolves , ie., changes over time Slide 13 © Carliss Y. Baldwin and Kim B. Clark, 2004

Modularizations with

s Significant Modular, Option-Rich Design Architectures IBM System/360 DEC PDP 11; VAX Apple 2 IBM PC Sun 2; 3; Java VM RISC to-end) 50 53 Internet Protocols (end 62 65 68 71 74 77 80 83 Unix and C; Linux 86 89 92 95 HTML; XML Slide 14 180 160 140 120 20 0 60 40 100 80 © Carliss Y. Baldwin and Kim B. Clark, 2004

Second Example—

Voluntary Collective Development (Open Source Development) Slide 15 © Carliss Y. Baldwin and Kim B. Clark, 2004

Open Source is—

Slide 16

Social Movement Free Software New System of Property Rights GNU GPL Many Software Development Processes— One Method?

Complementary Institutional Structure(s) A Bunch of Organizations + Governance Structures

© Carliss Y. Baldwin and Kim B. Clark, 2004

Software Development Processes Are Design Processes

Slide 17

Social Movement Free Software New System of Property Rights GNU GPL A Bunch of Organizations + Governance Structures Many Software Development Processes — One Method?

Design Processes

© Carliss Y. Baldwin and Kim B. Clark, 2004

What Causes Voluntary Collective Development?

 Look to the underlying designs  Designs and their Architecture “lie between” Knowledge and the Economy Knowledge Design Architecture Designs Economy  Modular, “option-rich” design architectures

precede and enable

voluntary collective development Slide 18 © Carliss Y. Baldwin and Kim B. Clark, 2004

Design Architecture and Voluntary Collective Development

 Cooking dinner (Rival good: lot size = 12 portions)  – One big stew = Not modular, no option value » A cook has no incentive to join with other cooks – Meat, salad, dessert = Modular, j=3 » Three cooks have incentives to get together – Two different stew recipes = Option value, s » Two cooks, pick the best recipe after the fact > 0 – Three courses, two recipes per course = Modules with option value » Six cooks will voluntarily

join up, cook, and feed each other

»

May feed an additional 6-18 people (free riders)

Collective Church recipe book (Non-rival good) – Contributions = #courses x #recipes per course Slide 19 © Carliss Y. Baldwin and Kim B. Clark, 2004

The Two Linked Games

 Work (write code, patch, etc.) /* bitmap.c contains the code that handles the inode and block bitmaps */ #include #include #include #define clear_block(addr) \ __asm__("cld\n\t" \ "rep\n\t" \ "stosl" \ ::"a" (0),"c" (BLOCK_SIZE/4),"D" ((long) (addr)):"cx","di")  Reveal (publish code, comments, etc.) Slide 20 © Carliss Y. Baldwin and Kim B. Clark, 2004

Conclusions: A Voluntary, Collective Design Process Requires—    For existence: – Designer-users – Non-rivalrous goods –

A design architecture with modules and/or options

– Communication speeds matching the design interval for one module – Methods of SYSTEM INTEGRATION AND TESTING (omitted here—see DR1 and Bessen) For efficiency: – Ways to know who’s working on what – Ways to know which module design is better or best (Module-level testing—see DR1, contrast to Bessen) For robustness (to solve the Prisoners’ Dilemma game): – Rewards for communication – Iteration with an indeterminate horizon (not strict repetition) Slide 21 © Carliss Y. Baldwin and Kim B. Clark, 2004

Modular Clusters

Slide 22 © Carliss Y. Baldwin and Kim B. Clark, 2004

When and if it arrives…

Modularity in design, coupled with high option value is

compelling, surprising and dangerous…

Dangerous for whom?

Slide 23 © Carliss Y. Baldwin and Kim B. Clark, 2004

IBM System/360

 The first modular computer design  IBM did not understand the option value had created it  Did not increase its inhouse product R&D  Result: Many engineers left – to join “plug-compatible peripheral” companies  San Jose labs

—> Silicon Valley

“Compelling, surprising, dangerous”

Slide 24 © Carliss Y. Baldwin and Kim B. Clark, 2004

IBM Personal Computer

 Highly modular architecture  IBM outsourced hardware and software  Controlled one high-level chip (BIOS) and the manufacturing process

Then

 Compaq reverse-engineered the BIOS chip  Taiwanese lowered manufacturing costs

By 1990 IBM was seeking to exit the unprofitable PC marketplace!

Slide 25 © Carliss Y. Baldwin and Kim B. Clark, 2004

Compaq vs. Dell

 Dell did to Compaq what Compaq did to IBM…  Dell created an equally good machine, and  Used

process modularity

to reduce its production, logistics and distribution costs and increase ROIC – Negative Net Working Capital – Direct sales, no dealers

By 1990 Compaq was seeking to exit the unprofitable PC marketplace!

Slide 26 © Carliss Y. Baldwin and Kim B. Clark, 2004

Modular Clusters are Turbulent

 The value created moves around with the high s options  In the Computers Cluster: value has moved a lot!

 Because s s were so high (Moore’s Law, Amdahl’s Law, User Learning, …) Slide 27 © Carliss Y. Baldwin and Kim B. Clark, 2004

Concentration of Value — 1980

T he Computer Cluster in 1980 Market Values in Constant $

80 70 60 50 40 30 20 10 0 IBM AD 35 R 70 s e x IBM 35 71 35 72 35 75 35 76 35 77 36 70 36 72 36 In te 74 l e x In te l 36 78 73 70 73 71 Mi cro so ft 73 72 e x Mi cro so ft 73 73 73 74 73 77 Slide 28 IBM ADRs 3570 ex IBM 3571 3572 3575 3576 3577 3670 3672 Intel 3674 ex Intel 3678 7370 7371 Microsoft 7372 ex Microsoft 7373 7374 7377 © Carliss Y. Baldwin and Kim B. Clark, 2004

Dispersion of Value — 1996

T he Computer Cluster in 1996 M arket Values in Constant $

80 60 40 20 180 160 140 120 100 0 IBM AD 35 R 70 s e x IBM 35 71 35 72 35 75 35 76 35 77 36 70 36 72 36 In te 74 l e x In te l 36 78 73 70 73 71 Mi cro so ft 73 72 e x Mi cro so ft 73 73 73 74 73 77 Slide 29 IBM ADRs 3570 ex IBM 3571 3572 3575 3576 3577 3670 3672 Intel 3674 ex Intel 3678 7370 7371 Microsoft 7372 ex Microsoft 7373 7374 7377 © Carliss Y. Baldwin and Kim B. Clark, 2004

Reconcentration of Value — 2002

T he Computer Cluster in 2002 Market Values in Constant $

300 250 200 150 100 50 0 IBM AD 35 R 70 s e x IBM 35 71 35 72 35 75 35 76 35 77 36 70 36 72 36 In te 74 l e x In te l 36 78 73 70 73 71 Mi cro so ft 73 72 e x Mi cro so ft 73 73 73 74 73 77 IBM ADRs 3570 ex IBM 3571 3572 3575 3576 3577 3670 3672 Intel 3674 ex Intel 3678 7370 7371 Microsoft 7372 ex Microsoft 7373 7374 7377 Slide 30 © Carliss Y. Baldwin and Kim B. Clark, 2004

Two Questions:

Is the Modular Cluster form of industrial organization a healthy and sustainable form?

– Pricing holds the key – If firms kill each other in product markets, consolidation will occur

Are some types of clusters healthier than others?

Slide 31 © Carliss Y. Baldwin and Kim B. Clark, 2004

What is a healthy cluster?

 In terms of system prices?

– High, medium, low  In terms of aggregate profits?

– High, medium, low  In terms of the distribution of value, concentration of power in layers?

– Is the Wintel platform cluster a healthy form of cluster?

 In terms of accessibility through open standards?

Slide 32 © Carliss Y. Baldwin and Kim B. Clark, 2004

Applications/Extensions

 Science of Design – Mapping DSMs of Large Systems » Calculating Coordination Cost and Change Cost – Formalizing Design Structure using Predicate Logic – Valuing Functions and Functionalities – Locating Functionality in the Design Structure of Large Systems – Estimating the Technical Potential of Modules » Tracking the evolution of versions and functionalities in large systems Slide 33 © Carliss Y. Baldwin and Kim B. Clark, 2004

Applications/Extensions

 Empirical Studies of Modular Clusters – Pricing and profitability of large clusters – Tracking Evolution, Design Dependencies, and Economic Value in Supply Networks—Video games and Software – Modular Operators as Correlates/Predictors of Mergers & Acquisitions – Mortgage Banking – Electronics Slide 34 © Carliss Y. Baldwin and Kim B. Clark, 2004

Applications/Extensions

 Computational Experiments – Impact of local, voluntary mechanisms, eg, bargaining on network formation and efficiency – Innovation and imitation in modular vs. “near modular” design and task structures – Pricing and profitability of firms in “symmetric” and “asymmetric” modular clusters Slide 35 © Carliss Y. Baldwin and Kim B. Clark, 2004

Applications/Extensions

 Management studies/Strategy – Comparing Knowledge Spans to Task Structures – The architecture of self-organizing projects » Game theoretic structures of open source projects » Tools and methods of coordination in open source projects – Competitive strategies for firms in clusters » Managing footprints for maximum ROIC » Design of a design monopoly Slide 36 © Carliss Y. Baldwin and Kim B. Clark, 2004

Applications/Extensions

 Public policy/Design of institutions – Impact of intellectual property rights and M&A on entry into design competitions – Impact of design tournaments on design effort – Epistemic problems in estimating value » What are sufficient beliefs and how are they supported?

» Where do we get valid estimates of “technical potential”?

» Bubbles and crashes Slide 37 © Carliss Y. Baldwin and Kim B. Clark, 2004

Slide 38

Thank you!

© Carliss Y. Baldwin and Kim B. Clark, 2004

Primer on Modularity and Design Architecture

Slide 39 © Carliss Y. Baldwin and Kim B. Clark, 2004

Mapping the Modular Structure of a Complex Engineering System

We can “see” a system’s modular structure via a Design Structure Matrix (DSM) Map Slide 40 © Carliss Y. Baldwin and Kim B. Clark, 2004

Design Structure Matrix Map of a Laptop Computer

Drive System Main Board LCD Screen Packaging

.

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 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 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 .

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 .

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 .

Graphics controller on Main Board or not?

If yes, screen specifications change; If no, CPU must process more; adopt different interrupt protocols Slide 41 © Carliss Y. Baldwin and Kim B. Clark, 2004

Design Structure Matrix Map of a Modular System

Design Rules Drive System Main Board LCD Screen Pack aging System Testing & Integ ration

. x x x x x .

x x x .

x x x x x .

x x x .

x x x x x x x Design Rules Task Group .

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 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 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 x x .

x x x x x x Hidden Modules many Task groups .

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 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 Sys tem x x x .

x x x Inte gration x x x x x x and Tes ting x x x x x x x .

.

x Tas k Group

Slide 42 © Carliss Y. Baldwin and Kim B. Clark, 2004

A “modularization” (splitting) of a complex design goes from Map A to Map B

 Via Design Rules , which specify  Architecture, Interfaces and Tests , that provide Module  Encapsulation and Information Hiding .

Slide 43 © Carliss Y. Baldwin and Kim B. Clark, 2004

Mapping Modular Structures over Time

 “Modular Design Evolution” (MDE)  John Holland’s theory of

CAS

Slide 44 © Carliss Y. Baldwin and Kim B. Clark, 2004

Design Hierarchy View

Global Design Rules Module A Module B Module C Module D System Integration & Testing

The system comprises four "hidden" modules and a "System Integration and Testing" stage. From the standpoint of the designers of A B, C, and D (the hidden modules), system integration and testing is simply another module: they do not need to know the details of what goes on in that phase as long as they adhere to the global design rules. Obviously, the converse is not true: the integrators and testers have to know something about the hidden modules in order to do their work. How much they need to know depends on the completeness of the design rules. Over time, as knowledge accumulates, more testing will occur within the modules, and the amount of information passed to the integrators will decrease. The special, time-dependent role of integration and testing is noted by the heavy black border around and gray shading within the "module." Slide 45 © Carliss Y. Baldwin and Kim B. Clark, 2004

Six Modular Operators

System I Global Design Rules Module F Design Rules System II Design Rules System Integration & Testing Module A Design Rules Module C D-E Design Rules Module G Module D Module E Architectural Module System I Module F Translator F-1 F-2 F-3 System II Module F Translator A-1 A-2 A-3 Splitting Substitution Exclusion Augmenting Inversion Porting

We started with a generic two-level modular design struc ture, as shown in Figure 5-3, but with six modules (A, B, C, D, E, F) instead of four. (To display the porting operator, we moved the "System Integration & Testing Module" to the left-hand side of the figure.) We then applied each operator to a different set of modules .

Module A was Three different Module C was

Split

into three sub-modules.

Substitutes Excluded.

A new Module G was c reated to were developed for Module B.

Augment

the system.

Common elements of Modules D and E were Module F was

Ported.

Inverted.

Subs ystem design rules and an architectural module were developed to allow the inversion.

Firs t it was split; then its "interior" modules were grouped within a shell; then translator modules were developed.

The ending s ystem is a three-level sys tem, with two

modular subsystems

performing the func tions of Modules A, D and E in the old system.

In addition to the standard hidden modules, there are three kinds of spec ial modules, whic h are indic ated by heavy blac k borders and shaded interiors: System Integration & Testing Module Architectural Module Translator Module(s )

System II Modules ...

Slide 46 © Carliss Y. Baldwin and Kim B. Clark, 2004

The Costs of Modularity

Slide 47 © Carliss Y. Baldwin and Kim B. Clark, 2004

Every important cross-module interdependency must be addressed via a design rule.

Drive System Main Board LCD Screen Packaging

.

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 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 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 .

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 .

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 .

Graphics controller on Main Board or not?

If yes, screen specifications change; If no, CPU must process more; adopt different interrupt protocols

This is costly Costs eat up the option value Modularization may not pay

Slide 48 © Carliss Y. Baldwin and Kim B. Clark, 2004

Costs of Experiments vary by module

Size

large large small small

Visibility

hidden visible hidden visible

Examples

disk drive; large application program microprocessor; operating system cache memory; small application program instruction set; internal bus Slide 49 © Carliss Y. Baldwin and Kim B. Clark, 2004

Thus each module has its own “value profile”

Large; Hidden

50% 40% 30% 20% 10% 0% -10% 0 -20% -30% -40% -50% 5

Large; Visible

10 15

Small; Hidden Small; Visible

20

No. of Experiments

25 Slide 50 © Carliss Y. Baldwin and Kim B. Clark, 2004

Value Profiles for a Workstation System

Sun Microsystems Workstation circa 1992

Slide 51 © Carliss Y. Baldwin and Kim B. Clark, 2004

“Modularity-in-design is not good or bad. It is important and it is costly. And dangerous to ignore.”

Slide 52 © Carliss Y. Baldwin and Kim B. Clark, 2004

Just remember—

“Compelling, surprising, dangerous…”

Slide 53 © Carliss Y. Baldwin and Kim B. Clark, 2004