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.

product, communities devoted to physical products can range

had lost interest in working further on the program, and so his re-

from simple information exchange sites to sites well furnished

sponse to Raymond's suggestions was to offer his role to Raymond

with tools and infrastructure. Within sports, Franke and Shah's

so that he could evolve the popclient further as he chose.

study illustrates relatively simple community infrastructure. Thus,

380

Raymond accepted the role of popclient's maintainer, and over the

the boardercross community they studied consisted of semi-

next months he improved the program significantly in conjunction

professional athletes from all over the world who meet in up to 10

with advice and suggestions from other users. He carefully culti-

competitions a year in Europe, North America, and Japan. Franke

vated his more active beta list of popclient users by regularly com-

and Shah report that community members knew one another well,

municating with them via messages posted on an public electronic

and spent a considerable amount of time together.

They also

SiSU

web.mit.edu/evhippel/www/

66

Democratizing Innovation

assisted one another in developing and modifying equipment for

aerodynamicist employed by an aerospace firm.

their sport. However, the community had no specialized sets of

tools to support joint innovation development.

Note that physical products are information products during the

386

design stage. In earlier days, information about an evolving de-

384

More complex communities devoted to the development of physical

sign was encoded on large sheets of paper, called blueprints, that

products often look similar to open source software development

could be copied and shared. The information on blueprints could

communities in terms of tools and infrastructure. As an example,

be understood and assessed by fellow designers, and could also

consider the recent formation of a community dedicated to the de-

be used by machinists to create the actual physical products repre-

velopment and diffusion of information regarding novel kitesurfing

sented. Today, designs for new products are commonly encoded

equipment. Kitesurfing is a water sport in which the user stands

in computer-aided design (CAD) files. These files can be created

on a special board, somewhat like a surfboard, and is pulled along

and seen as two-dimensional and three-dimensional renderings by

by holding onto a large, steerable kite. Equipment and technique

designers. The designs they contain can also be subjected to au-

have evolved to the point that kites can be guided both with and

tomated analysis by various engineering tools to determine, for ex-

against the wind by a skilled kitesurfer, and can lift rider and board

ample, whether they can stand up to stresses to which they will

many meters into the air for tens of seconds at a time.

be subjected. CAD files can then be downloaded to computer-

controlled fabrication machinery that will actually build the compo-

385

Designing kites for kitesurfing is a sophisticated undertaking,

nent parts of the design.

involving low-speed aerodynamical considerations that are not

yet well understood. Early kites for kitesurfing were developed

The example of the kitesurfing group's methods of sharing design

387

and built by user-enthusiasts who were inventing both kitesurfing

information illustrates the close relationship between information

techniques and kitesurfing equipment interdependently. In about

and physical products. Initially, users in the group exchanged de-

2001, Saul Griffith, an MIT PhD student with a long-time interest in

sign ideas by means of simple sketches transferred over the Inter-

kitesurfing and kite development, decided that kite-surfing would

net. Then group members learned that computerized cutters used

benefit from better online community interaction. Accordingly, he

by sail lofts to cut sails from large pieces of cloth are suited to cut-

created a site for the worldwide community of user-innovators

ting cloth for surfing kites. They also learned that sail lofts were

in kitesurfing (www.zeroprestige.com). Griffith began by posting

interested in their business. Accordingly, innovation group mem-

patterns for kites he had designed on the site and added helpful

bers began to exchange designs in the form of CAD files compat-

hints and tools for kite construction and use. Others were invited to

ible with sail lofts' cutting equipment. When a user was satisfied

download this information for free and to contribute their own if they

with a design, he would transmit the CAD file to a local sail loft for

wished. Soon other innovators started to post their own kite de-

cutting. The pieces were then sewn together by the user or sent to

signs, improved construction advice for novices, and sophisticated

a sewing facility for assembly. The total time required to convert an

design tools such as aerodynamics modeling software and rapid

information product into a physical one was less than a week, and

prototyping software. Some kitesurfers contributing innovations

the total cost of a finished kite made in this way was a few hundred

to the site had top-level technical skills; at least one was a skilled

dollars---much less than the price of a commercial kite.

SiSU

web.mit.edu/evhippel/www/

67

Democratizing Innovation

388

User-to-User Assistance

Table 7.3 Innovators tended to be the ones assisting others with

394

their innovations (p < 0.0001).

389

Clearly, user innovation communities can offer sophisticated sup-

395

port to individual innovators in the form of tools. Users in these

innovation communities also tend to behave in a collaborative man-

Innovators

Non-innovators

Total

ner. That is, users not only distribute and evaluate completed inno-

Gave assistance

28

13

41

vations; they also volunteer other important services, such as as-

Did not give assistance

32

115

147

sisting one another in developing and applying innovations.

Total

60

128

390

Franke and Shah (2003) studied the frequency with which users in

four sporting communities assisted one another with innovations,

and found that such assistance was very common (table 7.2). They

Source: Franke and Shah 2003, table 7.

396

also found that those who assisted were significantly more likely

Such helping activity is clearly important to the value contributed

397

to be innovators themselves (table 7.3). The level of satisfaction

by innovation communities to community participants. Why peo-

reported by those assisted was very high. Seventy-nine percent

ple might voluntarily offer assistance is a subject of analysis. The

agreed strongly with the statement “If I had a similar problem I

answers are not fully in, but the mysteries lessen as the research

would ask the same people again.” Jeppesen (2005) similarly found

progresses. An answer that appears to be emerging is that there

extensive user-to-user help being volunteered in the field of com-

are private benefits to assistance providers, just as there are for

puter gaming.

those who freely reveal innovations (Lakhani and von Hippel 2003).

391

Table 7.2 Number of people from whom innovators received as-

In other words, provision of free assistance may be explicable in

sistance.

terms of the private-collective model of innovation-related incen-

392

tives discussed earlier.

Number of people

Number of cases

Percentage

0

0

0

1

3

6

2

14

26

3--5

25

47

6--10

8

15

> 10

3

6

Total

53

100

393

Source: Franke and Shah 2003, table 4.

SiSU

web.mit.edu/evhippel/www/

68

Democratizing Innovation

398

8 Adapting Policy to User Innovation

and Wagner (1975) found the overall probability of success for

new industrial products to be only 27 percent. Elrod and Kelman

399

Government policy makers generally wish to encourage activities

(1987) found an overall probability of success of 26 percent for con-

that increase social welfare, and to discourage activities that re-

sumer products. Balachandra and Friar (1997), Poolton and Bar-

duce it. Therefore, it is important to ask about the social welfare

clay (1998), and Redmond (1995) found similarly high failure rates

effects of innovation by users. Henkel and von Hippel (2005) ex-

in new products commercialized. Although there clearly is some re-

plored this matter and concluded that social welfare is likely to be

cycling of knowledge from failed projects to successful ones, much

higher in a world in which both users and manufacturers innovate

of the investment in product development is highly specific. This

than in a world in which only manufacturers innovate.

high failure rate therefore represents a huge inefficiency in the con-

version of R&D investment to useful output, and a corresponding

400

In this chapter, I first explain that innovation by users complements

reduction in social welfare.

manufacturer innovation and can also be a source of success-

enhancing new product ideas for manufacturers. Next, I note that

Research indicates that the major reason for the commercial fail-

405

innovation by users does not exhibit several welfare-reducing ef-

ure of manufacturer-developed products is poor understanding of

fects associated with innovation by manufacturers. Finally, I eval-

users' needs by manufacturer-innovators. The landmark SAPPHO

uate the effects of public policies on user innovation, and suggest

study showed this in a very clear and convincing way. This study

modifications to those that---typically unintentionally---discriminate

was based on a sample of 31 product pairs. Members of each

against innovation by users.

pair were selected to address the same function and market. (For

example, one pair consisted of two “roundness meters,” each de-

401

Social Welfare Effects of User Innovation

veloped by a separate company.) One member of each pair was

402

Social welfare functions are used in welfare economics to provide

a commercial success (which showed that there was a market for

a measure of the material welfare of society, using economic vari-

the product type); the other was a commercial failure. The devel-

ables as inputs. A social welfare function can be designed to ex-

opment process for each successful and failing product was then

press many social goals, ranging from population life expectancies

studied in detail. The primary factor found to distinguish success

to income distributions. Much of the literature on product diversity,

from failure was that a deeper understanding of the market and

innovation, and social welfare evaluates the impact of economic

the need was associated with successful projects (Achilladelis et

phenomena and policy on social welfare from the perspective of

al. 1971; Rothwell et al. 1974). A study by Mansfield and Wagner

total income of a society without regard to how that income is dis-

(1975) came to the same conclusion. More recent studies of in-

tributed. We will take that viewpoint here.

formation stickiness and the resulting asymmetries of information

held by users and manufacturers, discussed in chapter 3, support

403

User

Innovation

Improves

Manufacturers'

Success

the reasonableness of this general finding. Users are the genera-

Rates

tors of information regarding their needs. The decline in accuracy

404

It is striking that most new products developed and introduced to

and completeness of need information after transfer from user to

the market by manufacturers are commercial failures. Mansfield

manufacturer is likely to be substantial because important elements

SiSU

web.mit.edu/evhippel/www/

69

Democratizing Innovation

of this information are likely to be sticky (von Hippel 1994; Ogawa

of surgical equipment and coming upon prototype equipment de-

1998).

veloped by surgeons need not understand precisely why the in-

novators want this product or even precisely how it is used; the

406

Innovations developed by users can improve manufacturers' infor-

manufacturer need only understand that many surgeons appear

mation on users' needs and so improve their new product introduc-

willing to pay for it and then reproduce the important features of

tion success rates. Recall from previous chapters that innovation

the user-developed prototypes in a commercial product.

by users is concentrated among lead users. These lead users tend,

as we have seen, to develop functionally novel products and prod-

Observation of innovation by lead users and adoption by follow-

408

uct modifications addressing their own needs at the leading edge of

on users also can give manufacturers a better understanding of

markets where potential sales are both small and uncertain. Manu-

the size of the potential market. Projections of product sales have

facturers, in contrast, have poorer information on users' needs and

been shown to be much more accurate when they are based on

use contexts, and will prefer to manufacture innovations for larger,

actual customer behavior than when they are based on potential

more certain markets. In the short term, therefore, user innova-

buyers' pre-use expectations. Monitoring of field use of user-built

tions will tend to complement rather than substitute for products

prototypes and of their adoption by other users can give manufac-

developed by manufacturers. In the longer term, the market as a

turers rich data on precisely these matters and so should improve

whole catches up to the needs that motivated the lead user devel-

manufacturer's commercial success. In net, user innovation helps

opments, and manufacturers will begin to find production of similar

to reduce information asymmetries between users and manufactur-

innovations to be commercially attractive. At that point, innovations

ers and so increases the efficiency of the innovation process.

by lead users can provide very useful information to manufacturers

User Innovation and Provisioning Biases

409

that they would not otherwise have.

The economic literature on the impact of innovation on social wel-

410

407

As lead users develop and test their solutions in their own use en-

fare generally seeks to understand effects that might induce society

vironments, they learn more about the real nature of their needs.

to create too many product variations (overprovisioning) or too few

They then often freely reveal information about their innovations.

(underprovisioning) from the viewpoint of net social economic in-

Other users then may adopt the innovations, comment on them,

come (Chamberlin 1950). Greater variety of products available for

modify and improve them, and freely reveal what they have done

purchase is assumed to be desirable, in that it enables consumers

in turn. All of this freely revealed activity by lead users offers man-

to get more precisely what they want and/or to own a more diverse

ufacturers a great deal of useful information about both needs em-

array of products. However, increased product diversity comes at

bodied in solutions and about markets. Given access to a user-

a cost: smaller quantities of each product will be produced on aver-

developed prototype, manufacturers no longer need to understand

age. This in turn means that development-related and production-

users' needs very accurately and richly. Instead they have the

related economies of scale are likely to be less. The basic tradeoff

much easier task of replicating the function of user prototypes that

between variety and cost is what creates the possibility of overpro-

users have already demonstrated are responsive to their needs.

visioning or underprovisioning product variety. Innovations such

For example, a manufacturer seeking to commercialize a new type

as flexible manufacturing may reduce fixed costs associated with

SiSU

web.mit.edu/evhippel/www/

70

Democratizing Innovation

increased diversity and so shift the optimal degree of diversity up-

innovations. One of these was the effect of reduced pricing power

ward. Nonetheless, the conflict still persists.

on manufacturers that create “platform” products. Often, a manu-

facturer of such a product will want to sell the platform---a razor, an

411

Henkel and I studied the welfare impact of adding users as a source

ink-jet printer, a video-game player---at a low margin or a loss, and

of innovation to existing analyses of product diversity, innovation,

then price necessary add-ons (razor blades, ink cartridges, video

and social welfare. Existing models uniformly contained the as-

games) at a much higher margin. If the possibility of freely revealed

sumption that new products and services were supplied to the

add-ons developed by users makes development of a platform un-

economy by manufacturers only. We found that the addition of

profitable for a manufacturer, social welfare can thereby be re-

innovation by users to these analyses largely avoids the welfare-

duced. However, it is only the razor-vs.-blade pricing scheme that

reducing biases that had been identified. For example, consider

may become unprofitable. Indeed, if the manufacturer makes posi-

“business stealing” (Spence 1976). This term refers to the fact that

tive margins on the platform, then the availability of user-developed

commercial manufacturers benefit by diverting business from their

add-ons can have a positive effect: it can increase the value of

competitors. Since they do not take this negative externality into

the platform to users, and so allow manufacturers to charge higher

account, their private gain from introducing new products exceeds

margins on it and/or sell more units. Jeppesen (2004) finds that

society's total gain, tilting the balance toward overprovision of vari-

this is in fact the outcome when users introduce free game modifi-

ety. In contrast, a freely revealed user innovation may also reduce

cations (called mods) operating on proprietary game software plat-

incumbents' business, but not to the innovator's benefit. Hence,

form products (called engines) sold by game manufacturers. Even

innovation incentives are not socially excessive.

though the game manufacturers also sell mods commercially that

412

Freely revealed innovations by users are also likely to reduce dead-

compete with free user mods, many provide active support for the

weight loss caused by pricing of products above their marginal

development and diffusion of user mods built on their proprietary

costs. (Deadweight loss is a reduction in social welfare that oc-

game engines, because they find that the net result is increased

curs when goods are sold at a price above their marginal cost of

sales and profits.

production.) When users make information about their innovations

available for free, and if the marginal cost of revealing that informa-

Public Policy Choices

415

tion is zero, an imitator only has to bear the cost of adoption. This

is statically efficient. The availability of free user innovations can

If innovation by users is welfare enhancing and is also significant

416

also induce sellers of competing commercial offerings to reduce

in amount and value, then it makes sense to consider the effects

their prices, thus indirectly leading to another reduction in dead-

of public policy on user innovation. An important first step would

weight loss.

be to collect better data. Currently, much innovation by users--

-which may in aggregate turn out to be a very large fraction of

413

Reducing prices toward marginal costs can also reduce incentives

total economic investment in innovation--- goes uncounted or un-

to over-provision variety (Tirole 1988).

dercounted. Thus, innovation effort that is volunteered by users,

414

Henkel and I also explored a few special situations where social

as is the case with many contributions to open source software,

welfare might be reduced by the availability of freely revealed user

is currently not recorded by governmental statistical offices. This

SiSU

web.mit.edu/evhippel/www/

71

Democratizing Innovation

is also the case for user innovation that is integrated with prod-

from freely revealing what they know.

uct and service production. For example, much process innova-

These findings show that the characteristics of present-day intel-

tion by manufacturers occurs on the factory floor as they produce

420

lectual property regimes as actually experienced by innovators are

goods and simultaneously learn how to improve their production

far from the expectations of theorists and policy makers. The fun-

processes. Similarly, many important innovations developed by

damental reason that societies elect to grant intellectual property

surgeons are woven into learning by doing as they deliver services

rights to innovators is to increase private investment in innovation.

to patients.

At the same time, economists have long known that there will be so-

417

Next, it will be important to review innovation-related public poli-

cial welfare losses associated with these grants: owners of intellec-

cies to identify and correct biases with respect to sources of in-

tual property will generally restrict the use of their legally protected

novation. On a level playing field, users will become a steadily

information in order to increase private profits. In other words, in-

more important source of innovation, and will increasingly substi-