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.

92

Morrison, Roberts, and von Hippel (2000) explored user modifi-

by the manufacturers of their systems.

cations made by Australian libraries to computerized information

Table 2.2 OPAC modifications created by users served a wide

94

search systems called Online Public Access systems (“OPACs”).

variety of functions.

Libraries might not seem the most likely spot for technological in-

95

novators to lurk. However, computer technologies and the Inter-

Improved library management

Improved information-search capabilities

net have had a major effect on how libraries are run, and many li-

braries now have in-house programming expertise. Computerized

Add library patron summary statistics

Integrate images in records (2)

search methods for libraries were initially developed by advanced

Add library identifiers

Combined menu/command searches

and technically sophisticated user institutions. Development be-

Add location records for physical audit

Add title sorting and short title listing

gan in the United States in the 1970s with work by major universi-

Add book retrieval instructions for staff and patrons

Add fast access key commands

ties and the Library of Congress, with support provided by grants

Add CD ROM System backup

Add multilingual search formats <:br>Add key word

from the federal government (Tedd 1994). Until roughly 1978, the

searches (2)

only such systems extant were those that had been developed by

Add book access control based on copyright

Add topic linking and subject access

libraries for their own use. In the late 1970s, the first commercial

Patrons can check their status via OPAC

Add prior search recall feature

providers of computerized search systems for libraries appeared in

Patrons can reserve books via OPAC (2)

Add search "navigation aids"

the United States, and by 1985 there were at least 48 OPAC ven-

Remote access to OPAC by different systems

Add different hierarchical searches

dors in the United States alone (Matthews 1985). In Australia (site

Add graduated system access via password

Access to other libraries' catalogs (2)

of the study sample), OPAC adoption began about 8 years later

Add interfaces to other in-house IT systems

Add or customize web interface (9)

than in the United States (Tedd 1994).

Word processing and correspondence (2)

Hot links for topics

Umbrella for local information collection (2)

Extended searches

93

Morrison, Roberts, and I obtained responses from 102 Australian

Local systems adaptation

Hot links for source material

libraries that were users of OPACs. We found that 26 percent of

these had in fact modified their OPAC hardware or software far

beyond the user-adjustment capabilities provided by the system

Source of data: Morrison et al. 2000, table 1. Number of users

96

manufacturers. The types of innovations that the libraries devel-

(if more than one) developing functionally similar innovations is

oped varied widely according to local needs. For example, the li-

shown in parentheses after description of innovation.

brary that modified its OPAC to “add book retrieval instructions for

staff and patrons” (table 2.2) did so because its collection of books

The libraries in the sample were asked to rank themselves on a

97

was distributed in a complex way across a number of buildings---

number of characteristics, including “leading edge status” (LES).

SiSU

web.mit.edu/evhippel/www/

18

Democratizing Innovation

(Leading edge status, a construct developed by Morrison, is re-

lated to and highly correlated with the lead user construct (in this

sample, ρ (LES, CLU) = 0.904, p = 0.000). 1 Self-evaluation

bias was checked for by asking respondents to name other li-

braries they regarded as having the characteristics of lead users.

Self-evaluations and evaluations by others did not differ signifi-

cantly.

98

Libraries that had modified their OPAC systems were found to have

significantly higher LES---that is, to be lead users. They were also

found to have significantly higher incentives to make modifications

than non-innovators, better in-house technical skills, and fewer “ex-

ternal resources” (for example, they found it difficult to find outside

vendors to develop the modifications they wanted for them). Ap-

plication of these four variables in a logit model classified libraries

into innovator and non-innovator categories with an accuracy of 88

percent (table 2.3).

99

Table 2.3

Factors associated with innovating in librararies

(logit model).

χ2/4 = 33.85; ρ2 = 0.40; classification rate =

87.78%.

100

Coefficient

Standard error

Leading-edge status

1.862

0.601

Lack of incentive to modify

--0.845

0.436

Lack of in-house technology skills

--1.069

0.412

Lack of external resources

0.695

0.456

1LES contains four types of measures. Three (“benefits recognized early,”

“high benefits expected,” and “direct elicitation of the construct”) contain the

core components of the lead user construct. The fourth (“applications

generation”) is a measure of a number of innovation-related activities in which

users might engage: they “suggest new applications,” they “pioneer those

applications,” and (because they have needs or problems earlier than their

peers) they may be “used as a test site” (Morrison, Midgely, and Roberts 2004).

SiSU

web.mit.edu/evhippel/www/

19

Democratizing Innovation

Coefficient

Standard error

pelling), and swimming in canyons. Members do things like rap-

Constant

--2.593

0.556

pel down the middle of an active waterfall into a canyon below.

Canyoning requires significant skill and involves physical risk. It is

also a sport in rapid evolution as participants try new challenges

101

Source: Morrison et al. 2000, table 6.

and explore the edges of what is both achievable and fun.

102

The commercial value of user-developed innovations in the library

The second community studied was devoted to sailplaning.

106

OPAC sample was assessed in a relatively informal way. Two de-

Sailplaning or gliding, a more mature sport than canyoning,

velopment mangers employed by the Australian branches of two

involves flying in a closed, engineless glider carrying one or two

large OPAC manufacturers were asked to evaluate the commer-

people. A powered plane tows the glider to a desired altitude by

cial value of each user innovation in the sample. They were asked

means of a rope; then the rope is dropped and the engineless

two questions about each: (1) “How important commercially to your

glider flies on its own, using thermal updrafts in the atmosphere

firm is the functionality added to OPACs by this user-developed

to gain altitude as possible. The sailplaning community studied by

modification?” (2) “How novel was the information contained in the

Franke and Shah consisted of students of technical universities

user innovation to your firm at the time that innovation was devel-

in Germany who shared an interest in sailplaning and in building

oped?” Responses from both managers indicated that about 70

their own sailplanes.

percent (25 out of 39) of the user modifications provided function-

ality improvements of at least “medium” commercial importance to

Boardercross was the focus of the third community. In this sport,

107

OPACs---and in fact many of the functions were eventually incor-

six snowboarders compete simultaneously in a downhill race.

porated in the OPACs the manufacturers sold. However, the man-

Racetracks vary, but each is likely to incorporate tunnels, steep

agers also felt that their firms generally already knew about the

curves, water holes, and jumps. The informal community studied

lead users' needs when the users developed their solutions, and

consisted of semi-professional athletes from all over the world

that the innovations the users developed provided novel informa-

who met in as many as ten competitions a year in Europe, in North

tion to their company only in 10--20 percent of the cases. (Even

America, and in Japan.

when manufacturers learn about lead users' needs early, they may

The fourth community studied was a group of semi-professional

108

not think it profitable to develop their own solution for an “emerging”

cyclists with various significant handicaps, such as cerebral palsy

need until years later. I will develop this point in chapter 4.)

or an amputated limb. Such individuals must often design or make

improvements to their equipment to accommodate their particular

103

“Consumer” Innovation in Sports Communities

disabilities. These athletes knew each other well from national and

104

Franke and Shah (2003) studied user innovation in four commu-

international competitions, training sessions, and seminars spon-

nities of sports enthusiasts. The communities, all located in Ger-

sored by the Deutscher Sportbund (German National Sports Coun-

many, were focused on four very different sports.

cil).

105

One community was devoted to canyoning, a new sport popular in

A total of 197 respondents (a response rate of 37.8 percent) an-

109

the Alps. Canyoning combines mountain climbing, abseiling (rap-

swered a questionnaire about innovation activities in their commu-

SiSU

web.mit.edu/evhippel/www/

20

Democratizing Innovation

nities. Thirty-two percent reported that they had developed or mod-

ified equipment they used for their sport. The rate of innovation

varied among the sports, the high being 41 percent of the sailplane

enthusiasts reporting innovating and the low being 18 percent of

the boardercross snowboarders reporting. (The complexity of the

equipment used in the various sports probably had something to do

with this variation: a sailplane has many more components than a

snowboard.)

110

The innovations developed varied a great deal. In the sailplane

community, users developed innovations ranging from a rocket-

assisted emergency ejection system to improvements in cockpit

ventilation. Snowboarders invented such things as improved boots

and bindings. Canyoners' inventions included very specialized so-

lutions, such as a way to cut loose a trapped rope by using a chem-

ical etchant. With respect to commercial potential,

111

Franke and Shah found that 23 percent of the user-developed in-

novations reported were or soon would be produced for sale by

a manufacturer. Franke and Shah found that users who inno-

vated were significantly higher on measures of the two lead user

characteristics than users who did not innovate (table 2.4). They

also found that the innovators spent more time in sporting and

community-related activities and felt they had a more central role

in the community.

112

Table 2.4 Factors associated with innovation in sports communi-

ties.

113

Innovatorsa

Non-

Significance

innovatorsb

of differencec

Time in community

Years as a community member

4.46

3.17

p < 0.01

Days per year spent with community members

43.07

32.73

p < 0.05

Days per year spent participating in the sport

72.48

68.71

not significant

SiSU

web.mit.edu/evhippel/www/

21

Democratizing Innovation

Innovatorsa

Non-

Significance

medical equipment for use in their own practices. Using a logit

innovatorsb

of differencec

model to determine the influence of user characteristics on inno-

Role in communityd

vation activity, Lüthje found that innovating surgeons tended to be

"I am a very active member of the community."

2.85

3.82

p < 0.01

lead users (p < 0.01). He also found that solutions to problems en-

"I get together with members of the community for ac-

3.39

4.14

p < 0.05

tivities that are not related to the sport (movies, dinner

countered in their own surgical practices were the primary benefit

parties, etc.)."

that the innovating surgeons expected to obtain from the solutions

"The community takes my opinion into account when

2.89

3.61

p < 0.05

making decisions"

they developed (p < 0.01). In addition, he found that the level of

technical knowledge the surgeon held was significantly correlated

Lead user characteristic 1: being ahead of the trendd

with innovation (p < 0.05). Also, perhaps as one might expect in

"I usually find out about new products and solutions ear-

2.71

4.03

p < 0.001

lier than others."

the field of medicine, the “contextual barrier” of concerns about le-

"I have benefited significantly by the early adoption and

3.58

4.34

p < 0.01

gal problems and liability risks was found to have a strongly sig-

use of new products."

nificant negative correlation with the likelihood of user invention by

"I have tested prototype versions of new products for

4.94

5.65

p < 0.05

manufacturers."

surgeons (p < 0.01).

"In my sport I am regarded as being on the "cutting edge."

4.56

5.38

p < 0.01

With respect to the commercial value of the innovations the lead

117

"I improved and developed new techniques in boarder-

4.29

5.84

p < 0.001

crossing."

user surgeons had developed, Lüthje reported that 48 percent

Lead user characteristic 2: high benefit from innovationd

of the innovations developed by his lead user respondents

"I have new needs which are not satisfied by existing

3.27

4.38

p < 0.001

were or soon would be marketed by manufacturers of medical

products."

equipment.

"I am dissatisfied with the existing equipment."

3.90

5.13

p < 0.001

Discussion

118

The studies reviewed in this chapter all found that user innovations

119

114

Source: Franke and Shah 2003, table 3.

a. All values are means; n = 60.

in general and commercially attractive ones in particular tended to

b. All values are means; n = 129.

be developed by lead users. These studies were set in a range

c. Two-tailed t-tests for independent samples.

of fields, but all were focused on hardware innovations or on in-

d. Rated on seven-point scale, with 1 = very accurate and 7 = not

formation innovations such as new software. It is therefore impor-

accurate at all. Two-tailed t-tests for independent samples.

tant to point out that, in many fields, innovation in techniques is at

least as important as equipment innovation. For example, many

115

Innovation among Hospital Surgeons

novel surgical operations are performed with standard equipment

116

Lüthje (2003) explored innovations developed by surgeons working

(such as scalpels), and many novel innovations in snowboarding

at university clinics in Germany. Ten such clinics were chosen ran-

are based on existing, unmodified equipment. Technique-only in-

domly, and 262 surgeons responded to Lüthje's questionnaire---a

novations are also likely to be the work of lead users, and indeed

response rate of 32.6 percent. Of the university surgeons respond-

many of the equipment innovations documented in the studies re-

ing, 22 percent reported developing or improving some item(s) of

viewed here involved innovations in technique as well as innova-

SiSU

web.mit.edu/evhippel/www/

22

Democratizing Innovation

tions in equipment.

3 Why Many Users Want Custom Products

121

120

Despite the strength of the findings, many interesting puzzles re-

The high rates of user innovation documented in chapter 2 suggest

main that can be addressed by the further development of lead

122

that many users may want custom products. Why should this be

user theory. For example, empirical studies of innovation by lead

so? I will argue that it is because many users have needs that

users are unlikely to have sampled the world's foremost lead users.

differ in detail, and many also have both sufficient willingness to

Thus, in effect, the studies reviewed here determined lead users

pay and sufficient resources to obtain a custom product that is just

to be those highest on lead user characteristics that were within

right for their individual needs. In this chapter, I first present the

their samples. Perhaps other samples could have been obtained

case for heterogeneity of user needs. I then review a study that

in each of the fields studied containing users that were even more

explores users' heterogeneity of need and willingness to pay for

“leading edge” with respect to relevant market trends. If so, why

product customization.

were the samples of moderately leading-edge users showing user

innovation if user innovation is concentrated among “extreme” lead

Heterogeneity of User Needs

123

users? There are at least three possible explanations. First, most

of the studies of user innovation probably included users reason-

If many individual users or user firms want something different in

124

ably close to the global leading edge in their samples. Had the “top”

a product type, it is said that heterogeneity of user need for that

users been included, perhaps the result would have been that still

product type is high. If users' needs are highly heterogeneous, only

more attractive user innovations would have been found. Second,

small numbers of users will tend to want exactly the same thing. In

it may be that the needs of local user communities differ, and so

such a case it is unlikely that mass-produced products will precisely

local lead users really may be the world's lead users with respect to

suit the needs of many users. Mass manufacturers tend to want to

their particular needs. Third, even if a sample contains lead users

build products that will appeal to more users rather than fewer, so

that are not near the global top with respect to lead users' char-

as to spread their fixed costs of development and production. If

acteristics, local lead users might still have reasons to (re)develop

many users want something different, and if they have adequate

innovations locally. For example, it might be cheaper, faster, more

interest and resources to get exactly the product they need, they

interesting, or more enjoyable to innovate than to search for a sim-

will be driven either to develop it for themselves or to pay a custom

ilar innovation that a “global top” lead user might already have de-

manufacturer to develop it for them.

veloped.

Are users' needs for new products (and services) often highly het-

125

erogeneous? A test of reason suggests that they are. An indi-

vidual's or a firm's need for a many products depends on detailed

considerations regarding the user's initial state and resources, on

the pathway the user must traverse to get from the initial state to

the preferred state, and on detailed considerations regarding their

preferred end state as well. These are likely to be different for each

individual user and for each user firm at some level of detail. This,

SiSU

web.mit.edu/evhippel/www/

23

Democratizing Innovation

in turn, suggests that needs for many new products and services

Inspection of these descriptions shows a great deal of variation

that are precisely right for each user will differ: that needs for those

and few near-duplicates. Different functionality, of course, implies

products will be highly heterogeneous.

that the developers of the products had different needs. In the

2000 study of user modifications of library IT systems by Morri-

126

Suppose, for example, that you decide you need a new item of

son, Roberts, and von Hippel, discussed earlier, only 14 of 39 in-

household furnishing. Your house is already furnished with hun-

novations are functionally similar to any other innovations in the

dreds of items, big and small, and the new item must “fit in” prop-

sample. If one type of functionality that was repeatedly developed

erly. In addition, your precise needs for the new item are likely to

(“web interface”) is excluded, the overlap is even lower (see table

be affected by your living situation, your resources, and your pref-

2.2). Other responses by study participants add to this impression

erences. For example: “We need a new couch that Uncle Bill will

of high heterogeneity of need among users. Thirty percent of the

like, that the kids can jump on, that matches the wallpaper I adore,

respondents reported that their library IT system had been highly

that reflects my love of coral reefs and overall good taste, and that

customized by the manufacturer during installation to meet their

we can afford.” Many of these specific constraints are not results of

specific needs. In addition, 54 percent of study respondents agreed

current whim and are not easy to change. Perhaps you can change

with the statement “We would like to make additional improvements

the wallpaper, but you are less likely to change Uncle Bill, your kids,

to our IT system functionality that can't be made by simply adjust-

your established tastes with respect to a living environment, or your

ing the standard, customer-accessible parameters provided by the

resource constraints.

supplier.”

127

The net result is that the most desired product characteristics might

be specific to each individual or firm. Of course, many will be will-

Similar moderate overlap in the characteristics of user innovations

130

ing to satisfice---make compromises---on many items because of

can be seen in innovation descriptions provided in the study of

limits on the money or time they have available to get exactly what

mountain biking by Lüthje, Herstatt, and von Hippel (2002). In

they want. Thus, a serious mountain biker may be willing to simply

that study sample, I estimate that at most 10 of 43 innovations had

buy almost any couch on sale even if he or she is not fully happy

functionality similar to that of another sample member. This diver-

with it. On the other hand, that same biker may be totally unwill-

sity makes sense: mountain biking, which outsiders might assume

ing to compromise about getting mountain biking equipment that is

is a single type of athletic activity, in fact has many subspecial-

precisely right for his or her specific needs. In terms of industrial

ties.

products, NASA may insist on getting precisely right components

for the Space Shuttle if they affect mission safety, but may be will-

As can be seen in table 3.1, the specializations of mountain bik-

131

ing to satisfice on other items.

ers in the our study sample involved very different mountain bik-

ing terrains, and important variations in riding conditions and riding

128

Evidence from Studies of User Innovation

specializations. The innovations users developed were appropri-

129

Two studies of innovation by users provide indirect information on

ate to their own heterogeneous riding activities and so were quite

the heterogeneity of user need. They provide descriptions of the

heterogeneous in function. Consider three examples drawn from

functions of the innovations developed by users in their samples.

our study:

SiSU

web.mit.edu/evhippel/www/

24

Democratizing Innovation

132

• I ride on elevated, skinny planks and ladders, do jumps, steep

technical downhills, obstacles and big drops. Solution devised: I

needed sophisticated cycling armor and protective clothing. So I

designed arm and leg armor, chest protection, shorts, pants and

a jacket that enable me to try harder things with less fear of injury.

133

• I do back-country touring and needed a way to easily lift and carry

a fully loaded mountain bike on the sides of steep hills and moun-

tains and dangle it over cliffs as I climbed. Solution devised: I

modified the top tube and the top of my seat post to provide se-

cure attachment points for a carrying strap, then I modified a very

plush and durable mountaineering sling to serve as the over-

shoulder strap. Because the strap sits up high, I only need to

bend my knees a little bit to lift the bike onto my shoulders, yet it

is just high enough to keep the front wheel from hitting when I am

climbing a steep hill. Eventually, I came up with a quick-release

lateral strap to keep the main strap from sliding off my shoulder,

but it will easily break away if I fall or land in a fast river and need

to ditch my bike.

134

• When riding on ice, my bike has no traction and I slip and fall.

Solution devised: I increased the traction of my tires by getting

some metal studs used by the auto industry for winter tires. Then

I selected some mountain biking tires with large blocks of rubber

in the tread pattern, drilled a hole in the center of each block and

inserted a stud in each hole.

135

Table 3.1

Activity specializations of innovating mountain bik-

ers.

136

Preferred terrain

Number

of

Outside conditions

Number

of

Focus on particu-

Number

of

bikers

bikers

lar riding abilities

bikers

Fast downhill tracks

44 (39.6%)

Darkness, night rid-

45 (40.5%)

Jumps,

drops,

34 (30.6%)

(steep, drops, fast)

ing

stunts, obstacles

Technical

single

68 (61.3%)

Snow, ice, cold

60 (54.1%)

Technical

abil-

22 (19.8%)

tracks

(up

and

ity/balance

down, rocky, jumps)

SiSU

web.mit.edu/evhippel/www/

25

Democratizing Innovation

Preferred terrain

Number

of

Outside conditions

Number

of

Focus on particu-

Number

of

bikers

bikers

lar riding abilities

bikers

Since the 1970s, nearly all market-segmentation studies have been

140

carried out by means of cluster analysis (Green 1971; Green and

Smooth

single

13 (11.7%)

Rain, mud

53 (47.7%)

Fast

descents

/

34 (30.6%)

tracks (hilly, rolling,

downhill

Schaffer 1998). After cluster analysis places each participant in the

speed, sand, hard-

pack)

segment of the market most closely matching his needs, a mea-

Urban and streets

9 (8.1%)

Heat

15 (13.5%)

Endurance

9 (8.1%)

sure of within-segment need variation is determined. This is the

No special terrain

5 (4.5%)

High altitude

10 (9.0%)

Climbing

17 (13%)

proportion of total variation that is within each cluster, and it shows

preferred

how much users' needs deviate from the averages in “their” re-

No extreme outside

29 (26.1%)

Sprint

3 (2.7%)

conditions

spective segments. If within-segment variation is low, users within

No focus on specific

36 (32.4%)

the segment will have fairly homogeneous needs, and so may be

riding ability

reasonably satisfied with a standard product designed to serve all

customers in their segment. If it high, many users are likely to be

dissatisfied---some seriously so.

137

Source: Lüthje,Herstatt, and vonHippel 2002. This table includes

the 111 users in the study sample who had ideas for improvements

Within-segment variation is seldom reported in published studies,

141

to mountain biking equipment. (Of these, 61 had actually gone

but a survey of market-segmentation studies published in top-tier

on to build the equipment they envisioned.) Many of these users

journals did find 15 studies reporting that statistic. These studies

reported experience in more than one category of activity, so the

specified 5.5 clusters on average, and had an average remaining

sum in each column is higher than 111.

within-cluster variance of 46 percent (Franke and Reisinger 2003).

Franke and von Hippel (2003b) found similar results in an inde-

138

Evidence from Studies of Market Segmentation

pendent sample. In that study, an average of 3.7 market segments

139

Empirical data on heterogeneity of demand for specific products

were specified and 54 percent of total variance was left as within-

and services are sparse. Those most interested in studying the

segment variation after the completion of cluster analysis. These

matter are generally mass manufacturers of products and services

data suggest that heterogeneity of need might be very substantial

for consumers---and they do not make a practice of prospecting for

among users in many product categories. 2

heterogeneity. Instead, they are interested in finding areas where

users' needs are similar enough to represent profitable markets

2Cluster analysis does not specify the “right” number of clusters---it simply

for standard products produced in large volumes. Manufacturers

segments a sample into smaller and smaller clusters until the analyst calls a

halt. Determining an appropriate number of clusters within a sample can be

customarily seek such areas via market-segmentation studies that

done in different ways. Of course, it always possible to say that “I only want to

partition markets into a very few segments---perhaps only three,

deal with three market segments, so I will stop my analysis when my sample

four, or five. Each segment identified consists of customers with

has been segmented into three clusters.” More commonly, analysts will examine

relatively similar needs for a particular product (Punj and Stewart

the increase of squared error sums of each step, and generally will view the

optimal number of clusters as having been reached when the plot shows a

1983; Wind 1978). For example, toothpaste manufacturers may

sudden “elbow” (Myers 1996). Since this technique does not incorporate

divide their markets into segments such as boys and girls, adults

information on remaining within-cluster heterogeneity, it can lead to solutions

interested in tooth whitening, and so on.

with a large amount of within-cluster variance. The “cubic clustering criterion”

SiSU

web.mit.edu/evhippel/www/

26

Democratizing Innovation

142

A Study of Heterogeneity and Willingness To Pay

and other web server software programs have evolved into the

complicated front end for many of the technically demanding appli-

143

A need for a novel product not on the market must be accompanied

by adequate willingness to pay (and resources) if it is to be asso-

cations that now run on the Internet. For example, web server soft-

ciated with the actual development or purchase of a custom prod-

ware is now used to handle security and authentication of users, to

uct. What is needed to reliably establish the relationship among

provide e-commerce shopping carts, and gateways to databases.

heterogeneity of demand, willingness to pay, and custom product

In the face of strong competition from commercial competitors (in-

development or purchase is studies that address all three factors in

cluding Microsoft and Sun/Netscape), the Apache web server has

the same sample. My colleague Nikolaus Franke and I conducted

become the most popular web server software on the Internet, used

one such study in a population of users of web server software, a

by 67 percent of the many millions of World Wide Web sites extant

product used primarily by industrial firms (Franke and von Hippel

in early 2004. It has also received many industry awards for excel-

2003b).

lence.

144

Franke and I looked in detail at users' needs for security features

Franke and I created a preliminary list of server security functions

146

in Apache web server software, and at users' willingness to pay for

from published and web-based sources. The preliminary list was

solutions that precisely fit their needs. Apache web server software

evaluated and corrected by experts in web server security and

is open source software that is explicitly designed to allow modifi-

Apache web server software. We eventually ended up with a list

cation by anyone having appropriate skills. Anyone may download

of 45 security functions that some or many users might need. So-

open source software from the Internet and use it without charge.

lutions to some were already incorporated in the standard Apache

Users are also explicitly granted the legal right to study the soft-

code downloadable by users, others were available in additional

ware's source code, to modify the software, and to distribute mod-

modules, and a few were not yet addressed by any security mod-

ified or unmodified versions to others. (See chapter 7 for a full

ule generally available to the Apache community. (Security threats

discussion of open source software.)

can emerge quickly and become matters of great concern before a

successful response is developed and offered to the general com-

145

Apache web server software is used on web server computers con-

nected to the Internet. A web server's function is to respond to re-

munity. A recent example is site flooding, a form of attack in which

quests from Internet browsers for particular documents or content.

vandals attempt to cause a website to fail by flooding it with a very

A typical server waits for clients' requests, locates the requested re-

large number of simultaneous requests for a response.)

source, applies the requested method to the resource, and sends

Users of the security functions of web server software are the web-

the response back to the client. Web server software began by of-

147

masters employed by firms to make sure that their software is up to

fering relatively simple functionality. Over time, however, Apache

date and functions properly. A major portion of a webmaster's job is

(CCC) partially addresses this concern by measuring the within-cluster

to ensure that the software used is secure from attacks launched by

homogeneity relative to the between-cluster heterogeneity. It suggests choosing

those who wish illicit access or simply want to cause the software

the number of clusters where this value peaks (Milligan and Cooper 1985).

However, this method appears to be rarely used: Ketchen and Shook (1996)

to fail in some way. We collected responses to our study ques-

found it used in only 5 of 45 segmentation studies they examined.

tions from two samples of Apache webmasters: webmasters who

SiSU

web.mit.edu/evhippel/www/

27

Democratizing Innovation

posted a question or an answer on a question at the Apache Usenet

Even this understates the heterogeneity.

Responding Apache

150

Forum 3 and webmasters who subscribed to a specialized online

webmasters went far beyond the 45 security-related functions of

Apache newsgroup. 4 This stratified sample gave us an adequate

web server software that we offered for their evaluation. In our

representation of webmasters who both did and did not have the

questionnaire we offered an open question asking users to list up

technical skills needed to modify Apache security software to bet-

to four additional needs they experienced that were not covered

ter fit their needs: subscribers to apache-modules.org tend to have

by the standard list. Nearly 50 percent used the opportunity to add

a higher level of technical skills on average than those posting to

additional functions. When duplicates were eliminated, we found

the Apache Usenet Forum. Data were obtained by means of an

that 92 distinct additional security-related needs had been noted

Internet-based questionnaire.

by one or more webmaster users.6

148

The Heterogeneity of Users' Needs

High heterogeneity of need in our sample suggests that there

151

149

Franke and I found the security module needs of Apache users

were very heterogeneous indeed both among those that had the

to each individual's needs. (This step is analogous to the Ward's method in

cluster analysis that also minimizes within cluster variation; see Punj and Stewart

in-house capability to write code to modify Apache and those that

1983.) The “error” is then measured as the sum of squared Euclidean distances.

did not. The calibrated coefficient of heterogeneity, Hc, was 0.98,

We then repeated these steps to determine the error for two optimally positioned

indicating that there was essentially no tendency of the users to

products, three products, and so on up to a number equaling I -- 1. The sum of

cluster beyond chance. (We defined the “heterogeneity of need” in

squared errors for all cases is then a simple coefficient that measures how much

the needs of i individuals can be satisfied with j standard products. The

a group as the degree to which the needs of i individuals can be

“coefficient of heterogeneity” just specified is sensitive both to the (average)

satisfied with j standard products which optimally meet their needs.

distance between the needs and for the configuration of the needs: when the

This means that heterogeneity of need is high when many standard

needs tend to form clusters the heterogeneity coefficient is lower than if they are

products are necessary to satisfy the needs of i individuals and low

evenly spread. To make the coefficient comparable across different populations,

we calibrate it using a bootstrapping technique (Efron 1979) involving dividing

when the needs can be satisfied by a few standard products. The

the coefficient by the expected value (this value is generated by averaging the

higher the coefficient the more heterogeneous are the needs of

heterogeneity of many random distributions of heterogeneity of the same kind).

users in a sample. If the calibrated heterogeneity coefficient Hc

The average random heterogeneity coefficient is then an appropriate value for

equals 1, there is no systematic tendency of the users to cluster.

calibration purposes: it assumes that there is no systematic relationship

between the needs of the individuals or between the need dimensions.

If it is lower than 1, there is some tendency of the individuals to

6Conceptually, it can be possible to generate “one perfect product” for

cluster. A coefficient of 0 means that the needs of all individuals

everyone--- in which case heterogeneity of demand is zero---by simply creating

are exactly the same. 5 )

all the features wanted by anyone (45 + 92 features in the case of this study),

and incorporating them in the “one perfect product.” Users could then select the

3‹http://groups-beta.google.com/group/comp.infosystems.www.servers.unix›

features they want from a menu contained in the one perfect product to tailor it

4‹http://modules.apache.org/›

to their own tastes. Doing this is at least conceptually possible in the case of

5To measure heterogeneity, Franke and I analyzed the extent to which j

software, but less so in the case of a physical product for two reasons: (1)

standards, varying from [1; i], meet the needs of the i individuals in our sample.

delivering all possible physical options to everyone who buys the product would

Conceptually, we first locate a product in multi-dimensional need space

be expensive for physical goods (while costing nothing extra in the case of

(dimensions = 45 in the case of our present study) that minimizes the distances

information products); (2) some options are mutually exclusive (an automobile

SiSU

web.mit.edu/evhippel/www/

28

Democratizing Innovation

should be a high interest in obtaining modifications to Apache---

To compensate for the likely overstatement of expressed relative

154

and indeed, overall satisfaction with the existing version was only

to actual WTP in our study, Franke and I conservatively deflated

moderate.

respondents' indicated willingness to pay by 80 percent. (Although

the product in question was intended for private use, webmasters

152

Willingness to Pay for Improvements

were talking about their willingness to spend company money, not

their own money.) We asked each user who had indicated that he

153

It is not enough to want a better-fitting custom product. One must

also be willing and able to pay to get what one wants. Those in the

was not really satisfied with a function (i.e., whose satisfaction with

Apache sample who did innovate were presumably willing to pay

the respective function was 4 or less on a 7-point scale, where 1

the price to do so. But how much were the users in our sample-

= not satisfied at all, and 7 = very satisfied) to estimate how much

--the innovators and the non-innovators--- willing to pay now for

he would be willing to pay to get a very satisfactory solution re-

improvements? Estimating a user's willingness to pay (WTP) is

garding this function. After deflation, our sample of 137 webmas-

known to be a difficult task. Franke and I used the contingent val-

ters said they were willing to pay $700,000 in aggregate to modify

uation method, in which respondents are directly asked how much

web server software to a point that fully satisfied them with respect

they are willing to pay for a product or service (Mitchell and Carson

to their security function needs. This amounts to an average of

1989). Results obtained by that method often overestimate WTP

$5,232 total willingness to pay per respondent. This is a striking

significantly. Empirical studies that compare expressed WTP with

number because the price of commercial web server software sim-

actual cash payments on average showed actual spending behav-

ilar to Apache's for one server was about $1,100 at the time of

ior to be somewhat smaller than expressed WTP in the case of

our study (source: www.sun.com, November 2001). If we assume

private purchases (such as in our case). In contrast, they gener-

that each webmaster was in charge of ten servers on average, this

ally find willingness to pay to be greatly overstated in the case of

means that each webmaster was willing to pay half the price of a

public goods such as the removal of a road from a wilderness area.

total server software package to get his heterogeneous needs for

7

security features better satisfied.

Increased Satisfaction from Customization of Apache

155

cannot be both red and green at the same time).

7The difference between actual willingness to pay and expressed willingness

Recall that it takes some technical skill to modify Apache web

156

to pay is much lower for private goods (our case) than for public goods. In the

server software by writing new code. In table 3.2, Franke and I ex-

case of private goods, Loomis et al. (1996) found the expressed willingness to

pay for art prints to be twice the actual WTP. Willis and Powe (1998) found that

amined only the technically skilled users in our sample who claimed

among visitors to a castle the expressed WTP was 60 percent lower than the

the capability of making modifications to Apache web server soft-

actual WTP. In the case of public goods, Brown et al. (1996), in a study of

ware. For these technically skilled users, we found significantly

willingness to pay for removal of a road from a wilderness area, found the

higher satisfaction levels among those that actually did customize

expressed WTP to be 4--6 times the actual WTP. Lindsey and Knaap (1999), in

a study of WTP for a public urban greenway, found the expressed WTP to be

their software---but even the users that made modifications were

2-10 times the actual WPT. Neil et al. (1994) found the expressed WTP for

not fully satisfied.

conserving an original painting in the desert to be 9 times the actual WTP. Seip

and Strand (1992) found that less than 10 percent of those who expressed

interest in paying to join an environmental organization actually joined.

SiSU

web.mit.edu/evhippel/www/

29

Democratizing Innovation

157

Table 3.2 Skilled users who customized their software were more

Discussion

161

satisfied than those who did not customize.

Heterogeneity of user need is likely to be high for many types of

162

158

products. Data are still scanty, but high heterogeneity of need is

Users

who

Users

who

Difference

a very straightforward explanation for why there is so much cus-

customized

did not cus-

(one-tailed

(n = 18)

tomize

(n

=

t-test)

tomization by users: many users have “custom” needs for products

44)

and services.

Satisfaction with basic web server functionality

5.5

4.3

0.100

Those interested can easily enhance their intuitions about hetero-

163

Satisfaction with authentication of client

3.0

1.0

0.001

genity of user need and related innovation by users. User innova-

Satisfaction with e-commerce-related functions

1.3

0.0

0.023

tion appears to be common enough so that one can find examples

Satisfaction with within-site user access control

8.5

6.9

0.170

for oneself in a reasonably small, casual sample. Readers there-

Satisfaction with other security functions

3.9

3.9

0.699

fore may find it possible (and enjoyable) to do their own informal

Overall satisfaction

4.3

2.6

0.010

tests of the matter. My own version of such a test is to ask the

students in one of my MIT classes (typically about 50 students) to

think about a particular product that many use, such as a backpack.

159

Source: Franke and von Hippel 2003, table 8. In this table, 45

I first ask them how satisfied they are with their backpack. Initially,

individual functions are grouped into five general categories. The

most will say “It's OK.” But after some discussion and thinking, a

satisfaction index ranges from -21 to +21.

few complaints will slowly begin to surface (slowly, I think, because

160

One might wonder why users with the ability to modify Apache

we all take some dissatisfaction with our products as the unremark-

closer to their liking were not totally satisfied. The answer can

able norm). “It doesn't fit comfortably” in this or that particular way.

be found in respondents' judgments regarding how much effort it

“When my lunch bag or thermos leaks the books and papers I am

would require to modify Apache still more to their liking. We asked

carrying get wet---there should be a water proof partition.” “I carry

all respondents who indicated dissatisfaction of level 4 or lower with

large drawings to school rolled up in my backpack with the ends

a specific function of Apache how much working time it would cost

sticking out. They are ruined if it rains and I have not taken the

them to improve the function to the point where they would judge

precaution of wrapping them in plastic.” Next, I ask whether any

it to be very satisfactory (to be at a satisfaction level of 7). For the

students have modified their backpacks to better meet their needs.

whole sample and all dissatisfactions, we obtained a working time

Interestingly enough, one or two typically have. Since backpacks

of 8,938 person-days necessary to get a very satisfactory solution.

are not products of very high professional or hobby interest to most

This equals $78 of incremental benefit per incremental program-

users, the presence of even some user innovation to adapt to in-

mer working day ($716,758 divided by 8,938 days). This is clearly

dividual users' unmet needs in such small, casual samples is an

below the regular wages a skilled programmer gets. Franke and I

interesting intuition builder with respect to the findings discussed

concluded from this that skilled users do not improve their respec-

in this chapter.

tive Apache versions to the point where they are perfectly satisfied

because the costs of doing so would exceed the benefits.

SiSU

web.mit.edu/evhippel/www/

30

Democratizing Innovation

164

4 Users' Innovate-or-Buy Decisions

Three specific contributors to transaction costs---in addition to

168

the “usual suspects,” such as opportunism---often have important

effects on users' decisions whether to buy a custom product or

165

Why does a user wanting a custom product sometimes innovate for

itself rather than buying from a manufacturer of custom products?

to develop it for themselves. These are (1) differences between

There is, after all, a choice---at least it would seem so. However,

users' and manufacturers' views regarding what constitutes a

if a user with the resources and willingness to pay does decide to

desirable solution, (2) differences in innovation quality signaling

buy, it may be surprised to discover that it is not so easy to find a

requirements between user and manufacturer innovators, and (3)

manufacturer willing to make exactly what an individual user wants.

differences in legal requirements placed on user and manufacturer

Of course, we all know that mass manufacturers with businesses

innovators. The first two of these factors involve considerations

built around providing standard products in large numbers will be

of agency costs. When a user hires a manufacturer to develop a

reluctant to accommodate special requests. Consumers know this

custom product, the user is a principal that has hired the custom

too, and few will be so foolish as to contact a major soup producer

manufacturer as to act as its agent. When the interests of the

like Campbell's with a request for a special, “just-right” can of soup.

principal and the agent are not the same, agency costs will result.

But what about manufacturers that specialize in custom products?

Recall from chapter 1 that agency costs are (1) costs incurred

Isn't it their business to respond to special requests? To under-

to monitor the agent to ensure that it follows the interests of the

stand which way the innovate-or-buy choice will go, one must con-

principal, (2) the cost incurred by the agent to commit itself not

sider both transaction costs and information asymmetries specific

to act against the principal's interest (the “bonding cost”), and

to users and manufacturers. I will talk mainly about transaction

(3) costs associated with an outcome that does not fully serve

costs in this chapter and mainly about information asymmetries in

the interests of the principal (Jensen and Meckling 1976). In the

chapter 5.

specific instance of product and service development, agency

considerations enter because a user's and a manufacturer's

166

I begin this chapter by discussing four specific and significant trans-

interests with respect to the development of a custom product

action costs that affect users' innovate-or-buy decisions. Next I re-

often differ significantly.

view a case study that illustrates these. Then, I use a simple quan-

titative model to further explore when user firms will find it more

Preferences Regarding Solutions

169

cost-effective to develop a solution---a new product or service---for

themselves rather than hiring a manufacturer to solve the problem

Individual products and services are components of larger user so-

170

for them. Finally, I point out that individual users can sometimes

lutions. A user therefore wants a product that will make the best

be more inclined to innovate than one might expect because they

overall tradeoff between solution quality and price. Sometimes the

sometimes value the process of innovating as well as the novel

best overall tradeoff will result in a willingness to pay a surprisingly

product or service that is created.

large amount to get a solution component precisely right. For ex-

ample, an individual user may specify tennis racket functionality

167

Users' vs.

Manufacturers' Views of Innovation Opportuni-

that will fit her specific technique and relative strengths and will

ties

be willing to pay a great deal for exactly that racket. Deviations

SiSU

web.mit.edu/evhippel/www/

31

Democratizing Innovation

in racket functionality would require compensating modifications in

manufacturer would have no competitive advantage in---and so

her carefully practiced and deeply ingrained hitting technique---a

no profit from making--- any of these same products from metal or

much more costly overall solution from the user's point of view. In

wood.

contrast, a user will be much less concerned with precisely how

the desired functionality is attained. For example, tennis play-

Specializations in solution types can be very narrow indeed. For

173

ers will typically be unconcerned about whether a tennis racket is

example, thousands of manufacturers specialize in adhesive-

made from metal, carbon fiber, plastic or wood---or, for that mat-

based fastening solutions, while other thousands specialize in

ter, from mud---if it performs precisely as desired. And, indeed,

mechanical fastening solutions involving such things as metal

users have quickly shifted to new types of rackets over the years

screws and nails. Importantly, companies that produce products

as new materials promise a better fit to their particular functional

and solution types that have close functional equivalence from

requirements.

the user's point of view can look very different from the point

of view of a solution supplier. For example, a manufacturer of

171

Of course, the same thing is true in the case of products for indus-

standard or custom adhesives needs chemists on staff with an

trial users. For example, a firm with a need for a process machine

expertise in chemical formulation. It also needs chemistry labs

may be willing to pay a great deal for one that is precisely appro-

and production equipment designed to mix specialized batches of

priate to the characteristics of the input materials being processed,

chemicals on a small scale, and it needs the equipment, expertise

and to the skills of employees who will operate the machine. Devi-

and regulatory approvals to package that kind of product in a way

ations in either matter would require compensating modifications in

that is convenient to the customer and also in line with regulatory

material supply and employee training---likely to be a much more

safeguards. In contrast, manufacturers specializing in standard

costly overall solution from the user's point of view. In contrast,

or custom metal fastening solutions need none of these things.

the user firm will be much less concerned with precisely how the

What they need instead are mechanical design engineers, a

desired functionality is achieved by the process machine, and will

machine shop to build product prototypes and production tooling,

care only that it performs precisely as specified.

specialized metal-forming production equipment such as screw

machines, and so on.

172

Manufacturers faced with custom development requests from

users make similar calculations, but theirs revolve around attempt-

Users, having an investment only in a need specification and not

174

ing to conserve the applicability of a low-cost (to them) solution.

in a solution type, want the best functional solution to their prob-

Manufacturers tend to specialize in and gain competitive advan-

lem, independent of solution type used. Manufacturers, in contrast,

tage from their capabilities in one or a few specific solution types.

want to supply custom solutions to users that utilize their existing

They then seek to find as many profitable applications for those

expertise and production capabilities. Thus, in the case of the two

solutions types as possible. For example, a specialist in fabricating

fastening technology alternatives just described, users will prefer

custom products from carbon fiber might find it profitable to make

whatever solution approach works best. In contrast, adhesives

any kind of product---from airplane wings to tennis rackets---as

manufacturers will find it tremendously more attractive to create

long as they are made from carbon fiber. In contrast, that same

a solution involving adhesive-based fastening, and manufacturers

SiSU

web.mit.edu/evhippel/www/

32

Democratizing Innovation

specializing in mechanical fastening will similarly strongly prefer to

Users typically expect a solution they have purchased to work cor-

179

offer to develop solutions involving mechanical fastening.

rectly and reliably “right out of the box.” In effect, a sharp line is

drawn between product development at the manufacturer's site and

175

The difference between users' incentives to get the best functional

routine, trouble-free usage at the purchaser's site. When the user

solution to their need and specialist manufacturers' incentives to

builds a product for itself, however, both the development and the

embed a specific solution type in the product to be developed are

use functions are in the same organization and may explicitly be

a major source of agency costs in custom product development,

overlapped. Repeated tests and repeated repairs and improve-

because there is typically an information asymmetry between user

ments during early use are then more likely to be understood and

and manufacturer with respect to what will be the best solution.

tolerated as an acceptable part of the development process.

Manufacturers tend to know more than users about this and to have

a strong incentive to provide biased information to users in order

A related difference in expectations has to do with field support for

180

to convince them that the solution type in which they specialize

a product that has been purchased rather than developed in house.

is the best one to use. Such biases will be difficult for users to

In the case of a purchased custom product, users expect that man-

detect because, again, they are less expert than the suppliers in

ufacturers will provide replacement parts and service if needed.

the various solution technologies that are candidates.

Responding to this expectation is costly for a custom manufacturer.

176

Theoretically, this agency cost would disappear if it were possi-

It must keep a record of what it has built for each particular user,

ble to fully specify a contract (Aghion and Tirole 1994; Bessen

and of any special parts incorporated in that user's products so

2004). But in product development, contracting can be problem-

that they can be built or purchased again if needed. In contrast,

atic. Information regarding characteristics of solutions and needs

if a user has developed a product for itself, it has people on site

is inescapably incomplete at the time of contracting---users can-

who know details of its design. These employees will be capable

not fully specify what they want in advance of trying out prototype

of rebuilding or repairing or redesigning the product ad hoc if and

solutions, and manufacturers are not fully sure how planned solu-

as the need arises. (Of course, if these knowledgeable employees

tion approaches will work out before investing in customer-specific

leave the user firm while the product they designed is still in use,

development.

such informality can prove costly.)

177

Users' Expectations

Manufacturers also must invest in indirect quality signals that may

181

178

When users buy a product from manufacturers, they tend to ex-

not have an effect on actual quality, but instead are designed to as-

pect a package of other services to come along with the product

sure both the specific user being served and the market in general

they receive. However, when users develop a product for them-

that the product being supplied is of high quality. These represent

selves, some of these are not demanded or can be supplied in a

another element of agency costs that user-innovators do not incur.

less formal, less expensive way by users for themselves. This set

When users develop an innovation for themselves, they end up in-

of implicit expectations can raise the cost to a user of a custom

timately knowing the actual quality of the solution they have devel-

solution bought from a manufacturer relative to a home-developed

oped, and knowing why and how it is appropriate to their task. As

solution.

an example, an engineer building a million-dollar process machine

SiSU

web.mit.edu/evhippel/www/

33