Democratizing Innovation by Eric von Hippel - HTML preview

PLEASE NOTE: This is an HTML preview only and some elements such as links or page numbers may be incorrect.
Download the book in PDF, ePub, Kindle for a complete version.

as how to achieve a certain look by means of layer cutting, or how to

manufacturer-based designers, and so they will require much less

achieve a certain streaked color pattern by selectively dying some

training to use the toolkit effectively.

strands of hair. However, an expert user is often very well practiced

at the skill of examining the shape of his or her face and hairstyle

589

For example, in the case of custom integrated circuit design, the

as reflected in a mirror, and visualizing specific improvements that

users of toolkits are typically electrical engineers who are design-

might be desirable in matters such as curls, shape, or color. In

ing electronic systems that will incorporate custom semiconduc-

addition, the user will be very familiar with the nature and func-

tor chips. The digital design language normally used by electrical

tioning of everyday tools used to shape hair, such as scissors and

engineers is Boolean algebra. Therefore, user-friendly toolkits for

combs.

custom semiconductor design are provided that allow toolkit users

to design in this language. That is, users can create a design,

A user-friendly toolkit for hairstyling innovation can be built upon

592

test how it works, and make improvements using only their own,

these familiar skills and tools. For example, a user can be invited

customary design language. At the conclusion of the design pro-

to sit in front of a computer monitor, and study an image of her face

cess, the toolkit then translates the user's logical design into the

and hairstyle as captured by a video camera. Then, she can select

design inputs required by the semiconductor manufacturer's pro-

from a palette of colors and color patterns offered on the screen,

duction system.

can superimpose the effect on her existing hairstyle, can exam-

590

A design toolkit based on a language and skills and tools familiar

ine it, and can repeatedly modify it in a process of trial-and-error

to the user is only possible to the extent that the user has familiar-

learning. Similarly, the user can select and manipulate images of

ity with some appropriate and reasonably complete language and

familiar tools, such as combs and scissors, to alter the image of the

SiSU

web.mit.edu/evhippel/www/

100

Democratizing Innovation

length and shape of her own hairstyle as projected on the computer

by adding to and subtracting from that starting point.

screen, can study and further modify the result achieved, and so

Translating Users' Designs for Production

596

forth. Note that the user's new design can be as radically new as is

desired, because the toolkit gives the user access to the most ba-

The “language” of a toolkit for user innovation must be convertible

597

sic hairstyling variables and tools such as hair color and scissors.

without error into the language of the intended production system

When the user is satisfied, the completed design can be translated

at the conclusion of the user's design work. If it is not, the entire

into technical hairstyling instructions in the language of a hairstyling

purpose of the toolkit will be lost---because a manufacturer receiv-

specialist---the intended production system in this instance.

ing a user design will essentially have to do the design work over

again. Error-free translation need not emerge as a major problem--

593

In general, steady improvements in computer hardware and soft-

-for example, it was never a major problem during the development

ware are enabling toolkit designers to provide information to users

of toolkits for integrated circuit design, because both chip design-

in increasingly friendly ways. In earlier days, information was often

ers and chip producers already used a language based on digi-

provided to users in the form of specification sheets or books. The

tal logic. In contrast, in some fields, translating from the design

user was then required to know when a particular bit of information

language preferred by users to the language required by intended

was relevant to a development project, find the book, and look it

production systems can be the central problem in toolkit design.

up. Today, a large range of potentially needed information can be

As an illustration, consider a recent toolkit test project managed by

embedded in a computerized toolkit, which is programmed to offer

Ernie Gum, the Director of Food Product Development for the USA

the user items of information only if and as a development being

FoodServices Division of Nestlé.

worked on makes them relevant.

One major business of Nestlé FoodServices is producing custom

598

594

Module Libraries

food products, such as custom Mexican sauces, for major restau-

rant chains. Custom foods of this type have traditionally been

595

Custom designs seldom are novel in all their parts. Therefore, a li-

developed by or modified by the chains' executive chefs, using

brary of standard modules will be a valuable part of a toolkit for user

what are in effect design and production toolkits taught by culinary

innovation. Provision of such standard modules enables users to

schools: recipe development procedures based on food ingredi-

focus their creative work on those aspects of their product or ser-

ents available to individuals and restaurants, and processed with

vice designs that cannot be implemented via pre-designed options.

restaurant-style equipment. After using their traditional toolkits to

For example, architects will find it very useful to have access to a

develop or modify a recipe for a new menu item, executive chefs

library of standard components, such as a range of standard struc-

call in Nestlé Foodservices or another custom food producer and

tural support columns with pre-analyzed structural characteristics,

ask that firm to manufacture the product they have designed---and

that they can incorporate into their novel building designs. Simi-

this is where the language translation problem rears its head.

larly, users who want to design custom hairstyles will often find it

helpful to begin by selecting a hairstyle from a toolkit library. The

There is no error-free way to translate a recipe expressed in the

599

goal is to select a style that has some elements of the desired

language of a traditional restaurant-style culinary toolkit into the

look. Users can then proceed to develop their own desired style

language required by a food-manufacturing facility. Food factories

SiSU

web.mit.edu/evhippel/www/

101

Democratizing Innovation

must use ingredients that can be obtained in quantity at consis-

developing Mexican sauces would contain a chili puree ingredient

tent quality. These are not the same as, and may not taste quite

processed on industrial equipment identical to that used to produce

the same as, the ingredients used by the executive chef during

food in commercial-size lots. (Each ingredient in such a toolkit also

recipe development. Also, food factories use volume production

contains traces of materials that will interact during production---for

equipment, such as huge-steam-heated retorts. Such equipment

example, traces of tomato are included in the chili puree---so that

is very different from restaurant-style stoves and pots and pans,

the taste effects of such interactions will also be apparent to toolkit

and it often cannot reproduce the cooking conditions created by

users.)

the executive chef on a stove-top---for example, very rapid heat-

Chefs interested in using the Nestlé toolkit to prototype a novel

ing. Therefore, food-production factories cannot simply produce a

602

Mexican sauce would receive a set of 20--30 ingredients, each in a

recipe developed by or modified by an executive chef “as is” under

separate plastic pouch. They would also be given instructions for

factory conditions---it will not taste the same.

the proper use of these ingredients. Toolkit users would then find

that each component differs slightly from the fresh components he

600

As a consequence, even though an executive chef creates a pro-

totype product using a traditional chef's toolkit, food manufactur-

or she is used to. But such differences are discovered immediately

ers find most of that information---the information about ingredients

through direct experience. The chef can then adjust ingredients

and processing conditions---useless because it cannot be straight-

and proportions to move to the desired final taste and texture that

forwardly translated into factory-relevant terms. The only informa-

is desired. When a recipe based on toolkit components is finished,

tion that can be salvaged is the information about taste and texture

it can be immediately and precisely reproduced by Nestlé factories-

contained in the prototype. And so, production chefs carefully ex-

-- because now the executive chef is using the same language as

amine and taste the customer's custom food prototype, then try to

the factory. In the Nestlé case, field testing by Food Product De-

make something that tastes the same using factory ingredients and

velopment Department researchers showed that adding the error-

methods. But an executive chef's taste buds are not necessarily

free translation feature to toolkit-based design by users reduced

the same as production chef taste buds, and so the initial factory

the time of custom food development from 26 weeks to 3 weeks by

version---and the second and the third---is typically not what the

eliminating repeated redesign and refinement interactions between

customer wants. So the producer must create variation after vari-

Nestlé and purchasers of its custom food products.

ation until the customer is finally satisfied.

Discussion

603

601

To solve the translation problem, Gum created a novel toolkit of

A toolkit's success in the market is significantly correlated with

604

pre-processed food ingredients to be used by executive chefs dur-

that toolkit's quality and with industry conditions. Thus, Prügl and

ing food development. Each ingredient in the toolkit was the Nestlé

Franke (2005) studied the success of 100 toolkits offered in a sin-

factory version of an ingredient traditionally used by chefs during

gle industry: computer gaming. They found that success, evalu-

recipe development: That is, it was an ingredient commercially

ated by independent experts, was significantly correlated with the

available to Nestlé that had been processed as an independent in-

quality of execution of the attributes of toolkits that have been dis-

gredient on Nestlé factory equipment. Thus, a toolkit designed for

cussed in this chapter. That is, success was found to be signif-

SiSU

web.mit.edu/evhippel/www/

102

Democratizing Innovation

icantly affected by the quality of trial-and-error learning enabled

can. For example, a design for a gate array generated with a

by a toolkit, by the quality of fit of the solution space offered to

toolkit will typically take up more physical space on a silicon chip

users' design problems, by the user friendliness of the tools pro-

than would a fully custom-developed design of similar complexity.

vided, and by the quality of module libraries offered with the toolkit.

Even when toolkits are on offer, therefore, manufacturers may

Schreier and Franke (2004) also obtained information on the im-

continue to design certain products (those with difficult technical

portance of toolkit quality in a study of the value that users placed

demands) while customers take over the design of others (those

on consumer products (scarves, T shirts, cell phone covers) cus-

involving complex or rapidly evolving user needs).

tomized with a simple, manufacturer-supplied toolkit. They found

Toolkits can be designed to offer a range of capabilities to users.

user willingness to pay for custom designs, as measured by Vick-

607

At the high end, with toolkits such as those used to design custom

rey auctions, was significantly negatively affected by the difficulty

integrated circuits, users can truly innovate, creating anything im-

of creating custom designs with a toolkit. In contrast, willingness to

plementable in digital electronics, from a dishwasher controller to

pay was significantly positively affected by enjoyment experienced

a novel supercomputer or form of artificial life. At the low end, the

in using a toolkit.

product configurators commonly offered by manufacturers of mass-

customized products enable, for example, a watch purchaser to

605

With respect to industry and market conditions, the toolkit-for-user

innovation approach to product design is likely to be most appealing

create a custom watch by selecting from lists of pre-designed faces,

to toolkit suppliers when the heterogeneous needs of many users

hands, cases, and straps. (Mass-customized production systems

can be addressed by a standard solution approach encoded in a

can manufacture a range of product variations in single-unit quan-

toolkit. This is because it can be costly to encode all the solution

tities at near mass-production costs (Pine 1993). In the United

and production information relevant to users' design decisions. For

States, production systems used by these manufacturers are gen-

example, a toolkit for custom semiconductor design must contain

erally based on computerized production equipment.)

information about the semi-conductor production process needed

The design freedom provided by toolkits for user innovation may

608

to ensure that product designs created by users are in fact pro-

not be of interest to all or even to most users in a market character-

ducible. Encoding such information is a one-time cost, so it makes

ized by heterogeneous needs. A user must have a great enough

the best economic sense for solution approaches that many will

need for something different to offset the costs of putting a toolkit

want to use.

to use for that approach to be of interest. Toolkits may therefore be

offered only to a subset of users. In the case of software, toolkits

606

Toolkits for user innovation are not an appropriate solution for

may be provided to all users along with a standard, default version

all product needs, even when heterogeneous needs can be ad-

of the product or service, because the cost of delivering the extra

dressed by a common solution approach. Specifically, toolkits will

software is essentially zero. In such a case, the toolkit's capability

not be the preferred approach when the product being designed

will simply lie unused in the background unless and until a user has

requires the highest achievable performance. Toolkits incorporate

sufficient incentive to evoke and employ it.

automated design rules that cannot, at least at present, translate

designs into products or software as skillfully as a human designer

Provision of toolkits to customers can be a complement to lead user

609

SiSU

web.mit.edu/evhippel/www/

103

Democratizing Innovation

idea-generation methods for manufacturers. Some users choosing

Manufacturers that think long-term disadvantages may accrue from

611

to employ a toolkit to design a product precisely right for their own

a switch to toolkits for user innovation and design will not necessar-

needs will be lead users, whose present strong need foreshadows

ily have the luxury of declining to introduce toolkits. If any manu-

a general need in the market. Manufacturers can find it valuable

facturer introduces a high-quality toolkit into a field favoring its use,

to identify and acquire the generally useful improvements made by

customers will tend to migrate to it, forcing competitors to follow.

lead users of toolkits, and then supply these to the general mar-

Therefore, a firm's only real choice in a field where conditions are

ket. For this reason, manufacturers may find it valuable implement

favorable to the introduction of toolkits may be whether to lead or

toolkits for innovation even if the portion of the target market that

to follow.

can directly use them is relatively small.

610

Toolkits can affect existing business models in a field in ways that

may or may not be to manufacturers' competitive advantage in the

longer run. For example, consider that many manufacturers of

products and services profit from both their design capabilities and

their production capabilities. A switch to user-based customiza-

tion via toolkits can affect their ability to do this over the long term.

Thus, a manufacturer that is early in introducing a toolkit approach

to custom product or service design may initially gain an advan-

tage by tying that toolkit to its particular production facility. How-

ever, when toolkits are made available to customer designers, this

tie often weakens over time. Customers and independent tool de-

velopers can eventually learn to design toolkits applicable to the

processes of several manufacturers. Indeed, this is precisely what

has happened in the custom integrated circuit industry. The toolk-

its revealed to users by the initial innovator, LSI, and later by rival

producers were producer-specific. Over time, however, Cadance

and other specialist toolkit supply firms emerged and developed

toolkits that could be used to make designs producible by a num-

ber of vendors. The end result is that manufacturers that previously

benefited from selling their product-design skills and their produc-

tion skills can be eventually forced by the shifting of design tasks

to customers via toolkits to a position of benefiting from their pro-

duction skills only.

SiSU

web.mit.edu/evhippel/www/

104

Democratizing Innovation

612

12 Linking User Innovation to Other Phenomena and

source have uses for what is revealed.

On an intuitive basis,

Fields

one can immediately see that these conditions are often met. Of

course, people and firms know different things. Of course there are

613

This final chapter is devoted to describing links between user-

many things that one would not be averse to freely revealing; and of

centered innovation and other phenomena and literatures.

Of

course others would often be interested in what is freely revealed.

course, innovation writ large is related to anything and everything,

After all, as individuals we all regularly freely reveal information not

so the phenomena and the literatures I will discuss here are only

generally known to people who ask, and presumably these people

those hanging closest on the intellectual tree. My goal is to enable

value at least some of the information we provide.

interested readers to migrate to further branches as they wish,

assisted by the provision of a few important references.

With

The economics of information communities can be much simpler

617

respect to phenomena, I will first point out the relationship of user

than that of the user innovation communities discussed earlier, be-

innovation to information communities---of which user innovation

cause valuable proprietary information is often not at center stage.

communities are a subset. With respect to related fields, I begin

When the service provided by information communities is to of-

by linking user-centric innovation phenomena explored in this

fer non-proprietary “content” in a more convenient and accessible

book to the literature on the economics of knowledge, and to the

form, one need consider only the costs and benefits associated

competitive advantage of nations. Next I link it to research on the

with information diffusion. One need not also consider potential

sociology of technology. Finally, I point out how findings regarding

losses associated with the free revealing of proprietary innovation-

user innovation could---but do not yet---link to and complement

related information.

the way that product development is taught to managers.

It is likely that information communities are getting steadily more

618

614

Information Communities

pervasive for the same reasons that user innovation communities

are: the costs of diffusing information are getting steadily lower as

615

Many of the considerations I have discussed with respect to user

computing and communication technologies improve. As a result,

innovation communities apply to information communities as well-

information communities may have a rapidly increasing impact on

--a much more general category of which user innovation commu-

the economy and on the landscape of industry. They are and will

nities are a subset. I define information communities as communi-

be especially empowering to fragmented groups, whose members

ties or networks of individuals and/or organizations that rendezvous

may for the first time gain low-cost access to a great deal of rich

around an information commons, a collection of information that is

and fresh information of mutual interest. As is the case for user

open to all on equal terms.

innovation networks, information networks can actually store con-

616

In close analogy to our discussions of innovation communities,

tent that participants freely reveal and make it available for free

I propose that commons-based information communities or net-

downloading. (Wikipedia is an example of this.) And/or, informa-

works will form when the following conditions hold: (1) Some have

tion networks can function to link information seekers and informa-

information that is not generally known. (2) Some are willing to

tion holders rather than actually storing information. In the latter

freely reveal what they know. (3) Some beyond the information

case, participants post to the network, hoping that someone with

SiSU

web.mit.edu/evhippel/www/

105

Democratizing Innovation

the requested information will spot their request and provide an

buyers. Similarly, product and service purchasers benefit from

answer (Lakhani and von Hippel 2003). Prominent examples can

good information on the characteristics of the various offerings

be found in the medical field in the form of specialized websites

in the market. Traditionally, firms have collected information on

where patients with relatively rare conditions can for the first time

users' needs and on products' characteristics by means of face-

find each other and also find specialists in those conditions. Pa-

to-face interviewing and (in the case of mass markets) question-

tients and specialists who participate in these groups can both pro-