Slide 1 © Carliss Y. Baldwin 2007 Design Theory and Methods Carliss Y. Baldwin Harvard Business School MiniConference with the Professors and Students

Slide 1 © Carliss Y. Baldwin 2007

Design Theory and Methods

Carliss Y. BaldwinHarvard Business School

MiniConference with the Professors and Students from L’Ecole des MinesOctober 16, 2007


Overview of my presentation Why we study designs: Their impact on

industry structure– Computers vs. Autos

Design Structure + Option Value => Industry Structure and Evolution

Our concept of design theory– Our Lineage– Our Definitions

Current projects


In the economy, value acts like a force operating on and through designs

Value = money or the promise of money

Consider the computer industry…

Why we study designs…


The changing structure of the computer industry

Andy Grove described a vertical-to-horizontal transition in the computer industry:

1995-“Modular Cluster”

1980-“Vertical Silos”


Grove’s Layer Map with Data

Take a sector, and consider the basic SIC / NAISC codes– 4 to 6 digit codes to compose the entire “ecosystem” as it

evolves– Work with industry experts to construct the sectors’ list

Tabulate the results in terms of “verticals” and “horizontals”– Objective: see how profit shifts from vertical to horizontal

layers– …and how much “churn” there is within layers

Map “Top N” Companies each year


How the map works


What a map looks like… Computers, 1979


The End of the Verticals

Value forced the industry to a new shape/structure

Does this always happen?


Look at the auto industry

In the beginning (1984)


The industry turned over, but most value stayed in the OEM layer…


Telecomm is yet another story…


What Causes One Industry to Break Apart and Another to Integrate and Consolidate?

Design Structure + Option Value


Design Structure + Option Value

Verticals HorizontalsHigh Will Dominate; Will Dominate;

High Turnover High Turnover

Option Value

Verticals HorizontalsWill Dominate; Will Dominate;

Low Low Turnover Low Turnover

Integral ModularDesign Structure


In other words… Design structure/Modularity (of products and processes)

determines industry structure– Because module boundaries are “thin crossing points” in the task-

and-design network– Transaction costs are low at module boundaries– Every thin crossing point/module boundary is a potential place to

put a transaction, i.e., bring in a different firm

Option value of designs determines rate of change/industry evolution– Option value makes design experiments worthwhile– Experiments yield new designs (of products and processes)…– Better new designs replace or augment the older ones!


Thus design theory holds the key to understanding the structure and dynamics of the economy

But what is design theory?


Influential Design Theorists Bell and Newell, computer hardware: Computer Structures Hennessy and Patterson, computer hardware-software

interface: Computer Architecture Mead and Conway, semiconductors: Intro to VLSI Nevins and Whitney, manufacturing: Concurrent

Engineering Nam Suh, mechanical engineering: The Principles of Design German design theorists (Hubka, Pahl and Beitz) in

mechanical engineering: The Theory of Technical Systems; Engineering Design: A Systematic Approach

March, Thompson, Galbraith, organizations: Organizations; Organizations in Action; Organizational Design


Our Direct Predecessors

Herbert Simon Christopher Alexander Fred Brooks David Parnas John Holland

Our theory builds on theirs


Herbert Simon

Sciences of the Artificial Fundamental insight:

Design is a decision-making process (under constraints of physics, logic and cognition)

Rational and reductionist


Christopher Alexander Notes on the Synthesis of

Form; A Pattern Language; A City is not a Tree; The Nature of Order

Fundamental insights: user-centered adaptive design; non-hierarchical complexity; unfolding designs; patterns

Mystic and visionary (frustrating to scientists)


Frederick Brooks The Mythical Man Month; No

Silver Bullet; Computer Architecture

Fundamental insights: the complexity catastrophe lurking in large designs; limits on the division of knowledge and labor; group inter-communication formula

Architect of System/360


David Parnas On the Criteria to be Used in

Decomposing Systems into Modules; Software Fundamentals

Fundamental insights: Information-hiding modularity; Abstraction; Interface; Modules are task assignments

Software designer


John Holland Hidden Order; Adaptation in

Natural and Artificial Systems; Emergence

Fundamental insights: Formal dynamics of complex adaptive systems; unified theory of natural and artificial evolution; operators

Our best link to complexity sciences (better than Kauffman)


Baldwin and Clark Design Rules: The Power of

Modularity; Modularity, Transactions and the Boundaries of Firms

Most important ideas: – designs are lodged in the larger

economy; – financial value is a force driving

design evolution; – designs are options; – modules are units of optional

substitution; – uncertainty is valuable; – new firms attach at module boundaries


That is our view of Design Theory

We are eager to learn yours…

But, first, for the sake of understanding, our definitions


Our Definitions Design (noun)

– Design (verb)—Process of Designing– Partial vs. Complete Designs

Design Hierarchy Design Space Value Landscape of a Design Space Design Structure/Modularity (Alan)

– Mirroring Hypothesis Options and Option Value (Carliss)


Designs are the instructions, based on knowledge, that turn resources into things people use and value.

Designs (noun)


Implications of the definition

Designs are not “the thing itself”, they are the instructions for making it– In software: source code = design– Compiled, running code = the thing itself

Designs are information Designs are part of human knowledge

– But not all knowledge is design– Designs make knowledge useful


Is the process of filling in the set of instructions Designing occurs in time

When design is complete, the design process is over, “production” can begin

Along the way, you have partial designs– A source of huge amounts of confusion!

The Process of Designing

Design Process

Start CompleteTime


Completing a Design—for a Mug

Mug= Cap Handle Shape Material Decoration100110 xxxxx xxxxx xxxx xxxxxx xxxxxx

Complete design of a Mug100110 100010001 1100100100 1001110100 1010001010 100010001 …

All parameters have been selected and encoded

Partial design of a Mug100110 100010001 xxxxxx xxxxxxx xxxxxxx 100010001 …


There are Many Different Names for Partial Designs

Design Process

Start CompleteTime


Decision-Information Hierarchies

Process of design is a decision-making process

Some design decisions create the need for subsequent decisions

MUG

Cap?

XXXX XXXX

Handle

XXXX

Logo

XXXX

Material


Decision-Information Hierarchies

If no cap—Decisions contingent on “cap” disappear

List of instructions becomes shorter

MUG

No Cap

XXXX

Handle

XXXX

Logo

XXXX

Material


Decision-making design hierarchy (Marple, ducts and valves, 1961)


Design Space

A design space comprises the set of all possible variants of a set of designs

Design spaces are bounded by prior design decisions– Mug … w/ cap– Pentium chip … w/ out-of-order, superscalar

microarchitecture Can be mapped, unmapped, partially

mapped


All these—and many more— are in the design space of “bicycles”

The newer the artifact, the less we know, the more exploration of the design space is valuable…


Value Landscape Maps points in design space to value In biology, value=fitness Fitness landscape for designs of eyes

Created by Mike Land. Height represents optical quality and the ground plane evolutionary distance. From Dawkins R: Climbing Mount Improbable. New York, Norton, 1996.


Value Landscape and Search History of a computer program

Scored results of submissions to the Mathworks “Sudoku” programming contest. Red path shows trajectory of best design over time.

http://www.mathworks.com/contest/sudoku/evolution.html


Design Structure/Architecture The degree and pattern of interdependence among

the elements of a design Establishes dependencies between design spaces,

hence their scope/complexity Some key architectural types

– Integral (all interdependent)

– Layered (X depends on Y; Y does not depend on X)

– Modular (independent blocks, all depending on design rules—combination of integral and layered)

Architectures are somewhat under the control of designers (subject to constraints of knowledge)

Mozilla Before Redesign (RAD) Mozilla After Redesign (Targeted)

Two Architectures, same functions, different states of knowledge

© Alan MacCormack, Johh Rusnak and Carliss Baldwin, 2006

Mirroring Hypothesis: Integral architectures require integral design teams

Mozilla BEFORE refactoring Linux of similar size

Coord. Cost = 30,537,703Change Cost = 17.35%

Coord. Cost = 15,814,993Change Cost = 6.65%

One Firm, Tight-knit Team, Rich communication

Distributed Open Source Development

© Alan MacCormack, Johh Rusnak and Carliss Baldwin, 2006


Alan and John will tell you more about our findings…

But we have one more set of critical concepts—

Options

Optional substitution

Option potential


Options, Optional Substitution

The right but not the obligation to take an action– Action = Use a new design

– If new is better than old, use new;

– Otherwise, keep the old (“optional substitution”)

Designs have the property of optional substitution Unit of optional substitution is a module (that

which can be changed without changing something else)

Thus option value resides in modules


Global Design Rules v.1

Version 1.0Version 1.2

Version 1.5Version 1.8

Low Medium Zero High

Option Potential () Determines value of optional substitution Successive, improving versions are evidence of

option potential () being realized over time—after the fact


Optional substitution at work—Evolution of a computer program


/Option potential is like dark matter in the universe

We can measure its effects but we can’t measure “it”

“Architects” can perceive /option potential– But architects often don’t talk to scientists!

Thus we lack ways to measure /option valuescientifically– It is a “research frontier”


Sources of /option potential Physics—

– Moore’s Law—dynamics of miniaturization in electronic circuits (Mead and Conway)

– Power and heat systems vs. logic systems (Dan Whitney) Users—

– Users experiment to discover their own needs/tastes– New features, techniques, and applications=> new

willingness to pay– Exaptations (designed for one thing, does another)

Architecture —– Isolate sources of uncertainty, variability, and bottlenecks– Support new compositions & combinations (e.g.

YouTube=movie+PC+Internet+Library)


When options are present, variability in outcomes is valuable…

When is this true?

Which industries, which NPD processes? Why?

How does one obtain variability in outcomes?


Current Projects & Collaborations

Dividing up the economic system—modularity and transactions

Design structure in software (Alan and John) Price competition in modular clusters (J. Woodard) Design theory and user innovation (E. von Hippel) Strategic variability (M. Szigety) The value of modular production systems (V.

Kuppaswamy) Transparency vs. modularity (L. Colfer) Modularity and intellectual property rights (J. Henkel) Layer Maps and Industry Evolution (M. Jacobides) ‘Selfish’ Designs—the institutional structure of innovation


Thank you!

Documents

Slide 1 © Carliss Y. Baldwin 2007 Design Theory and Methods Carliss Y. Baldwin Harvard Business School MiniConference with the Professors and Students