Space Planning for Online Community.

Full text

Page 1

Space Planning for Online Community
Danyel Fisher, Tammara Combs Turner, and Marc A Smith

Microsoft Research, Microsoft Live Labs, and Microsoft Internet Research Center
1 Microsoft Way
Redmond, Washington 98052
{danyelf, tcombs, marc.smith}@microsoft.com


Abstract
Several years of consulting with online community hosts
and managers have highlighted a variety of issues that recur
across many online community development efforts. We
summarize those issues in eight points that have functioned
as useful guidelines to working with online communities,
particularly within a corporate
recommendations focus on the location and purpose of the
community, the monitoring of social activity within the
space, the provision of feedback to participants and the
organization and maintenance of the space. While this
collection is particularly focused on issues relevant to
community organizers closely involved in starting,
maintaining or growing online communities, its principles
are generally applicable for analyzing and understanding the
dynamics within a variety of communities.
context. These
Introduction
Online communities can be difficult to handle: there are
real challenges to keeping users involved and happy, while
developing the community in directions that are mutually
beneficial for the hosts and participants. During the last few
years, we have worn several hats within our corporation: in
addition to our academic research, we have been frequently
engaged as internal consultants on the topic of online
community. We have met with organizers of online
communities that were intended to generate a range of
results. These organizers asked us about building,
sustaining, and retiring the communities for which they had
oversight. These communities often have tens of thousands
of participants from around the world, populated by
customers, partners and other users of the company‘s
products or services. Many of these community managers
have responsibility for monitoring and cultivating dozens
or hundreds of these social spaces.
This paper examines social issues in running and
organizing online communities from the perspective of a
corporate or institutional host. Many companies now
sponsor public discussion spaces for a variety of reasons.
Two of the most common are support spaces where users

Copyright © 2007, Association for the Advancement of Artificial
Intelligence (www.aaai.org). All rights reserved.

discuss the company‘s products, and general discussion
spaces, where the company seeks to generate awareness of
features and page-views (and thus advertisement revenue).
While we do not attempt to define community here, we
note that the ideas we discuss will be most helpful for
communities that fit Preece (2000)‘s working definition0.
She defines an online community as consisting of:

“People who interact socially as they strive to
satisfy their own needs…,

“A shared purpose, such as interest, need,
information exchange, or service that provides a
reason for the community,

―Policies … that guide people’s interactions, and

―Computer systems, to support and mediate
social interactions…” (pg 10, emphasis added)
The communities that we consulted on—and have the most
experience with—are open public discussion spaces, and
thus allow pseudonymous or fully anonymous access.
Spaces like MediaMoo (Bruckman & Resnick, 1995), in
which all users use real names, and must be invited, have
noticeably different patterns of interaction and incentives.
In this paper we synthesize key recommendations from
the existing literature with observations drawn from our
experience. Several books (Kim 2000; Preece 2000;
Powazek 2002a, Wenger, McDermott, & Snyder 2002)
have suggested rules to handle, develop, and moderate
online communities. However, we have found that
corporate and institutional hosts have distinct needs from
the organizers of private or social communities. Among
other needs, corporate sponsors often require communities
to demonstrate a return on investment. In addition,
investment in online communities gives product groups
insight into problems other customers may encounter.
Companies with global visibility face a heightened risk that
activity in the communities associated with their products
or services will reflect poorly on their company.
In this paper we offer a guide, relating suggestions to
relevant literature, and illustrating points with case
examples from community teams with which we have
consulted. These are guidelines, not rules, and may serve as
focal points for debates about best practices for community
management.
We also intend this paper to be of interest to the
academic community. While some of the general topics

Page 2

that we raise are well-known within the academic literature,
we have not found all of them articulated as we have here.
We hope this collection of guidelines may be useful for
researchers both evaluating the success and failure of other
systems, and may be useful directions for academics who
are building research systems (Beenen et al. 2004;
Bruckman & Resnick 1995; Hudson & Bruckman 2002).
These observations gain relevance in the context of the
explosive growth of social media systems which extend the
range of digital objects, access controls, and notification
mechanisms available for online communities. As blogs,
wikis, and ever-more-exotic forms of social media become
available, many aspects of community remain constant:
there will always be users interacting in the context of
various incentives to participate, in some cases yielding
productive cooperation.
Research Setting
Our company is a large one, and maintains a greater-than-
average number of communities. Not only are there spaces
to discuss most of its broad product line, but there are
additional communities related to interoperability between
products, ways to leverage the products, and ways to use
the several programming languages that the company
distributes. There is no single center for community at the
company; rather, individual groups take on the role of
community maintainer for their communities.
As researchers investigating online community, we
became an informal ―go-to‖ point for discussions of how
various communities within the company were, should or
could be doing. We worked closely with product support
teams and community maintainers, watching their methods
of engagement with communities for several years. The
questions they asked informed the technologies we
developed (e.g. (Turner et al. 2005)). Even groups that
neither used our technology nor were our research subjects
would check in with us occasionally to discuss systems that
they had built, or that they expected to construct. Our
partners asked for advice on technical newsgroups, forums
on lifestyles, politics, and gaming, and wikis on a variety of
topics. We addressed questions during all aspects of the
life-cycle for communities: we were involved as
communities created, sustained, and eventually retired.
The online community managers we consulted with had
a variety of intended goals and their communities
demonstrated different levels of interaction and made use
of a variety of technologies. Most of our experience was
around threaded discussion environments like web fora,
discussion groups, email lists and Usenet newsgroups.
We also conducted two specialized ―focus‖ group meetings
(Krathwol, 1998), comprising 15 community leaders who
were non-employee high-value content contributors. We
wanted to understand how these key informants
experienced online communities with a focus on how
information is exchanged, the types of people they interact
with and their motivation for participating in our
company‘s online communities. Their insights are included
where appropriate in our recommendations.
With this paper, we hope to make two main
contributions. First, we provide a mapping between
existing research into online social spaces to the leading
issues we have found in a large scale corporate online
community environment. Second, these points are
illustrated with brief examples that represent interactions
with managers of online communities and other business
partners. The quotations are characteristic of scenarios that
have been presented to us. While not word-for-word, these
quotations allow us to present the terms and issues in
practitioner‘s language and vernacular.
After a brief review of relevant literature, our points are
organized into four categories in the following sections:
1) Location and purpose
2) Social activity within the space
3) Providing feedback to participants
4) Organizing and maintaining the space.
We summarize these points in Table 1.
Literature Review
For
communities is a vital way to create and sustain a strong
customer base (Butler et al. 2008); deliver support and
services in ways that can alleviate barriers of time, distance
and cost; obtain customer feedback for product
improvements and new product design; save personnel cost
by having customers provide help to each other via support
groups; keep in close touch with customers; and promote
the company and brand loyalty. We have addressed the
several practitioner-oriented books (Kim 2000; Preece
2000; Powazek 2002a) oriented toward online community.
In this section, we discuss other research on sustaining or
maintaining online community that have addressed some of
these issues.
Broadening research on online communities has begun to
solidify, linking our findings in practice to projects online.
Constant, Sproull and Kiesler (1997), found that strangers
were willing to share information in online discussions that
public-facing organizations, fostering online
Location and purpose
1. Know the space‘s purpose
2. Build on existing community and brand
Monitor Social Activity
3. Know what the space is doing
4. Embrace leaders, respect lurkers
Provide Feedback
5. Reward users individually
6. Use positive reputation
Organize and Maintain the Space
7. Encourage critical mass
8. Exert gentle control

Table 1: Outline of the points on community

Page 3

was useful in constructing technical solutions although they
did not know the person they were helping—and that
information seekers thought the advice they received was
useful.
Normative research has begun to suggest ways to design
and orient communities. A series of papers (Beenen et al.
2004; Rashid, Ling, et al. 2006) have explored ways of
motivating users to increase their participation in online
communities. In those studies, an existing online social
space was partitioned; subsets of the users were then
encouraged to participate in different ways: users tended to
respond in the most positive way when they felt that they
uniquely contributed to the space. Other work from the
same lab examined Usenet Newsgroups, looking at factors
that predicted whether a message would receive a response
(Arguello et al. 2006). In addition, a workshop in 2005
collected a variety of different approaches to incentive
systems to help sustain online communities (Ellis,
Halverson, & Erickson 2005).
These incentives are often designed into the community
fabric itself. Specific design decisions, such as choosing
who can participate, can affect the ways that communities
develop and self-identify (Ren, Kraut, & Kiesler 2007).
Similarly, moderation strategies affect group direction. One
project compared three different oversight techniques
within member-driven communities (Cosley et al. 2005),
and found that community-based oversight led to most
improved contribution quality.
It is less common within the research community to
discuss experiences with maintaining systems (outside,
perhaps, the context of conference keynotes.) One
exception is Erickson‘s (2003) discussion of his experience
designing and evolving Babble to relay some basic
principles of social visualization. We hope to build on this,
as we believe that the research community can benefit from
both testing these experiences and—in building research
systems—can make sure to give themselves every
advantage in attracting research subjects to their systems.
Location and Purpose
We first address the fundamentals of the space: where it is
logically to be ‗found‘, and what it is for.
1. Know the Space’s Purpose
―We wanted to create a perfect thought leader
community,‖ the designer said. ―We‘d have all the
people who are experts in this tool talking to each
other. We weren‘t sure what they‘d talk about—but
get that many good people in one place and great stuff
would inevitably happen. It‘s been six months now,
and no one has shown up!‖
A community without a reason to exist will not come
together. While this is a common observation in the
literature, it seems to be often lost in the excitement around
online community. Too many organizations seem to
embrace the notion of a community without deciding what
the community is meant to accomplish. Powazek (2002a)
asks ―Who is your audience‖ (pg 8), and Preece (2000)
suggests that having a purpose is likely to make a more
―stable‖ (pg 80) environment. Knowing what your users
should expect to gain from the community can help guide
design decisions, and can aid in understanding how much
effort they might expect to put in. Wenger, McDermott, &
Snyder (2002) similarly suggest that considering the
participants‘ goals and interests will help get a community
started.
With the ―thought leader community‖ that the designer
wanted to create, for instance, there would be no purpose
other than socialization. The busy experts did not
necessarily see a benefit for themselves to participate: they
were already well-acknowledged, and already had places
where they could share their ideas; they were not looking
for credit or additional fame.
We see this as an extension of the so-called ―Grudin
paradox‖ (Grudin 1998) which predicts failure for
groupware if the effort that people put in does not match
with the benefit they get out. There will be some set of first
users who will have to put effort into this empty space:
without a purpose or direction, those users will not know
their potential audience. Should they ask a question into the
void, or raise a discussion point? Guidance, clear
descriptions, and especially early posts that model good
behavior can help accelerate this process.
The majority of visitors to open discussion spaces are
one time visitors. Their information behavior is similar to
other new users in that they typically do not know where to
find information to answer their questions, are new to the
technology or product, and may even be new to the concept
of asking questions in an online environment. Often, they
expect an answer to their question to happen as if it were a
commercial service encounter. Having a clear goal for the
community and clear group norms can help calibrate their
expectations while meeting many of their needs.
Online community spaces can take many forms and
structures depending on their purpose. Companies often
sponsor question-and-answer spaces for users of a product
to share information and resolve problems. In those
communities, structured as discussion boards, so-called
―answer people‖ (Turner et al 2005) respond to technical
questions and share their expertise with less experienced
users. It is not uncommon to support community for the
purpose of discussion between members, sometimes in the
hopes of building long-term loyalty. Other communities
form around annotating or expanding a data source. In
these spaces, discussion is kept close to the annotation
source, as in Wikipedia‘s ―talk‖ pages (Riehle 2006).
Knowing and understanding the purpose of your space is
important in identifying, measuring and reporting the
success of the community.
2. Build on Existing Community and Brand
One community organizer came to us for advice in
deciding how to create a new online community. She
explained, ―This should be a place where people who

Page 4

are using the Management Console can talk with each
other.‖
―What do they do now when they need to talk to each
other?‖
―Most of them are System Administrators who
already use the Installation Console message boards.‖
In this example, there was already an established
community within the company that the organizer could re-
use. Creating a new space would only have complicated
matters: users would need to choose where to look, and
would have encountered all the same people. Instead, she
simply extended the existing space by adding a special-
topic forum to cultivate discussion on the Management
Console issues. Participants flowed into it naturally, and the
moderation already in place continued. Participants were
not confused about which place to read or post to;
organizers did not need to coax users into their new space.
The organizer had not originally thought to use the existing
boards because those boards were run by a different
organizational structure within the company. Different
product groups and different organizational structures are
important divisions from within an organization, but they
are not as important from outside.
Structuring around internal organizational differences
can be a problematic way to organize online social spaces.
Outsiders will often ‗roll up‘ an entire organizational
hierarchy and view a company as a monolithic entity. A
site might be seen as being run by ―the university‖, not by
the psychology department, no matter how it is labeled.
Internally, the funding for a community can be sensitive to
the distinctions between reporting structures, product lines,
or university departments. Users, however, are unlikely to
navigate organizational structure to find their community.
If it makes sense, take advantage of shared branding to find
ways for users to connect across the organization. Consider
whether online help or technical support can find ways to
direct potential users to the information in your community
with minimal effort.
Monitor Social Activity
While placing the community initially is important, it is
critical to be able to follow the development and growth of
the community. Knowing the distribution of users can help
a community organizer make wise decisions about how to
develop or change the community space.
3. Know What the Space is Doing
One maintainer asked, ―What tools can I use to
measure our return on investment in communities?
How can we gauge reach and impact to current and
potential customers? Is this community going to
survive? Is it worth what we pay to maintain it?‖
Marketers want to know how many customers they ―touch‖
through community and what the impact of the interaction
is. Business managers want to know whether communities
are driving loyalty or adoption, or whether customer
support costs are being reduced. Community organizers
want to know if their community is thriving.
We have found only limited research addressing the
financial effects of online community (Cothrel & Johnston
2007), and none from academia. Counting members of a
community with any precision can be a challenging task;
estimating how much they are worth is even more complex.
Our own approach has been to assess the opposite, by
estimating the effect of discontinuing (or never building)
the community. For example, how would call center
volume change if there was no source online for answering
questions? Knowing what the space is doing entails on-
going monitoring of whether the community is broadly
fulfilling the purpose for which it was designed; and
keeping track of immediate issues that may be time
sensitive and require targeted intervention.
Preece (2000) notes that it is possible to assess a
community through interview, observational, and log- and
database- quantitative methods.
observational methods can be valuable tools for
understanding the basic feeling of the community and spot-
checking what activity is occurring. A community
organizer who doesn‘t read their messages cannot be
completely aware of what is going on.
But only log- and database-backed quantitative methods
can scale well: in our organization, some community
managers track dozens or hundreds of discussion spaces at
once. Databases that store community records can be a
critical tool, as an organizer can find out basic statistics
with a small handful of tools. Community analysis tools are
now becoming available to measure what users do within
the space. Sophisticated metadata viewers and processors,
such as Netscan (Turner et al. 2007), can provide detailed
information on users and their participation. Even if
databases are not available, community monitoring can be
done with web log analysis, reporting the number of page
views and the number of distinct users, and supplemented
with observation. (The number of page views for
submission and commenting pages may be a useful proxy
for comments posted.)
Figure 1 illustrates the results of one analysis that helps
community maintainers understand their community. Using
a visualization based on ―Newsgroup Crowds‖ (Turner et
al. 2007; Viégas & Smith 2004), we examined six different
Usenet newsgroups. The Newsgroup Crowd visualization
requires several basic statistics to be collected about each
participant in a newsgroup: how many different threads
they have participated in; how many times they have
posted; how many distinct days they have posted on. Each
bubble represents a single user: the x-axis then shows (in
log scale) the number of messages per thread; the y-axis
shows the number of days. In Figure 1, we show these
averages for six different groups.
The different behaviors for these groups—all originally
intended as ―question and answer‖ groups—gives us useful
insight into how their user population differs. In the bottom
Interviews and

Page 5

left, we see a newsgroup dedicated to discussing a database
product. The most active users have shown up virtually
every day of the year, and have posted very few messages
per thread: they are ―answer people,‖ joining a question-
and-answer, and
The ―MSN Messenger‖ group (top center) seems to only
have one person answering questions.
Last, in the ―MCSE exam‖ newsgroup, the most active
users are also involved in the longest threads: these groups
are clearly debating contentious issues. Understanding this
dynamic can help the community organizer decide whether
this is desirable behavior.
To date, there is little research available on the ‗health‘
of communities that offer useful predictions of their
likelihood to survive. Community organizers may still wish
to try to compare their month-to-month populations and
patterns of participation to understand how the community
is changing. Whether or not there are formal tools for
measuring behavior and interaction in the community, it is
important to collect some level of basic statistics to
establish baseline trends and patterns.
4. Embrace Leaders; Respect Lurkers
answering promptly.
The support specialist was worried. ―We‘re getting
thousands of page views a day from thousands of IPs,
but only a few dozen posts! What do we do about the
free riders?‖
Participation in most online groups will be heavily skewed:
some members will contribute a great deal, while many
others will contribute only
participation in many groups can be described with a long-
tailed power law. This pattern recurs in nearly all spaces we
have studied.
occasionally. Indeed,
Rather than attempt to change this behavior, or chase off
the infrequent contributors, it is worth thinking about this
curve in terms of the different roles played by members
with different participation levels. Lurkers fall at one
extreme: they are the ―read-only‖ members who consume
content without contributing any themselves. Nonnecke
(2003) has argued that a healthy population of lurkers is a
natural part of the ecosystem—that passive readership
keeps ideas flowing between communities, and that a lurker
on one system is likely to be an active contributing member
on another.
Lurkers‘ use of the system can also be harvested as a
passive labeling of page popularity or interest. Increasingly,
systems leverage lurkers by making their use visible,
showing how many times an article has been read or a
video watched as a form of ―read wear‖ (Hill 1992). Some
e-commerce sites, such as Amazon.com, have been
unusually innovative about utilizing lurker behavior: there
are a number of different levels of involvement that the site
mines, ranging from the very active (lists, reviews) all the
way out to the highly passive (books that users have
purchased together, or merely surfed between).
Frequent posters can be valuable members of a
community. Not only do they provide the majority of the
material and help set the direction for the community‘s
conversations, but they can nurture new members as well.
Arguello et al (2006) show that new members are far more
likely to stay if they are greeted upon entering a
community. Kim (2000) suggests that new users be
directed to a welcoming or visitor‘s center populated by
experienced users (pg 129).
One way to acknowledge their status is to give them a
separate place to work. We ran a series of focus groups
with frequent posters who were not employed by our
company in order to understand how they coordinated their
efforts. They had been given private discussion spaces for
them to ask questions or bring visibility to hard questions
that have not been answered in public discussion, and
found that space very helpful. They were able to form a
subculture of community around their status of being high-
value members of a larger community. Finding ways for
these key members to communicate behind the scenes helps
those members feel supported.
Provide Feedback
This next section addresses the sticky questions of
providing feedback systems for users. An over-generous
reward system can incent poor behavior, but a system that
does not allow users to show off their contributions will be
impoverished.
5. Reward Users Individually
We were working with a user-annotated reference set
in beta. A user could edit an annotation, or create a
new one. ―We‘d like to reward creators,‖ the creators
told us, ―so we‘ve put up a top ten list of contributors.
The more you add, the better your rank.‖

Figure 1: Newsgroup crowds for six Usenet groups. Each
bubble represents a single user. On the x-axis, a log scale
stretches from one post per thread to 100; on the y-axis, a
linear scale stretches from one day of posting to 365 distinct
days of posting in a given year. Large bubbles represent large
message volumes. Note that ―win98.gen_discussion‖ has a
population of people who post very few messages per thread,
but has posted nearly every day of the year.

Page 6

We checked the top ten list. The top three had each
repeatedly gone to existing annotations and added a
new one: ―The above annotation is incorrect.‖ Did
they misunderstand the ―edit this annotation‖ button?
Perhaps. But there would have been little recognition
for editing the existing annotation—while adding a
new one would gain recognition with minimal effort.
Not everything that can be counted, counts. A ranked ―top-
ten‖ list of participants can be a tempting target for gaming
or cheating the system, and virtually any ranking system
invites manipulation. In an online game, displaying a list of
the top ten players makes sense because game scores map
neatly to skill; in contrast, being in the ―top ten‖ posters
usually equates to contributing the highest number of
posts—but does not guarantee that those high numbers of
posts will be of high quality. Users that devote themselves
to manipulating ranking systems for the sake of being on a
high score list or having a high rank seem to be pervasive
in almost every public site (Powazek 2002b). Indeed, the
best of intentions and most carefully designed systems
often become complex and byzantine (Adler & Alfaro
2007; Lampe & Resnick 2004) in an attempt to find ‗fair‘
ratings. Informants in the focus groups referred to those
who try to become visible through gaming the system or
who try to make money from being visible as ―wannabes‖,
―greedy‖, ―spammers‖, and ―assholes‖. The general feeling
was that those people were not part of the community to
help others but rather to seek monetary gain – to get a free
subscription to the company‘s developer network or exploit
business opportunities (i.e. sell a solution or book, or gain
consulting prospects). Participants cautioned organizations
who host communities not to support this type of behavior
by viewing the quality of top responders‘ posts and not just
the quantity of responses to messages.
In systems where people are compared directly to each
other—―top ten lists‖, for example—the desire to increase a
ranking may be particularly strong. It may be better to
avoid lists that explicitly compare users to each other;
instead provide information about individuals on their
distinct profile pages.
An interesting case study comes from Slashdot (as
discussed by Powazek 2002b). Slashdot has a notion of
reviewing articles, and giving them points (Lampe &
Resnick 2004; Lampe, Johnston, & Resnick 2007); these
points give credit to the poster of the comment. This system
was ‗gamed‘ by users intent on having highly-reviewed
articles; they would find friends to review their articles
well. Slashdot responded by adding a meta-review system;
meta-reviewing earns ―karma‖ points. This multiply-tiered
system was intended to reduce gaming. Of course, this
created a new game: trying to accumulate as many karma
points as possible. The extrinsic goal of obtaining karma
points was not necessarily aligned with the goals of high-
quality conversation or high-quality reviewing. Over time,
Slashdot has adjusted their system; new forms of games
have arisen in a continuing arms race.
Ratings that reflect users‘ behavior can help them
recognize their role in the community and reinforce their
commitment to maintain it. Public recognition can also be
an incentive to help encourage posters to contribute more
or better posts, whether it is represented in statistical form
or textual form (such as ebay‘s ―star‖ levels). Even a basic
search function on user names can provide a basic form of
essential information about history and reputation; more
sophisticated portals can provide user reports that combine
a user‘s self-description along with their logged behavior
history. Users appreciate the ability to prove their value to
their employers, and to prove their status to other members
of the community.
While we do not have a complete answer, we can note
some positive examples: Powazek (2002b) suggests that
explicit rankings are more prone to gaming than implicit
ones: it may be better to label a user a ―frequent poster‖
than to report that their precise message count. Rashid et al
(Rashid et al 2006) suggest that displaying a non-numerical
value of a contribution to other users—finding ways to
highlight the users‘ unique added-value —may also be a
positive motivator.
6. Use Positive Reputation
For each post they place, we‘ll let other people rate it.
If they get a negative reputation, we‘ll put that by their
name so everyone knows to ignore them. That way,
the bad people won‘t sully our boards. Unless they
register under another name, I guess.
It is tempting to build a reputation system to control
misbehaving users: users who behave badly will be
punished by the community and receive a negative
reputation. If they want to get back in the communal good
graces, they need to repent, behave better, and ultimately
return to a state of grace.
In practice, these sorts of systems can easily backfire.
Resnick et al (2006) have shown the relative cost of a
‗negative reputation‘ is roughly equivalent to the cost of re-
entering the site with a new identity: that is, no punishment
can be worth more than the cost of re-entering the site from
scratch.
An intentional disrupter is unlikely to be dissuaded by a
negative reputation: instead, they will simply drop out and
create a new account under a new name. A financial
scammer on a site like eBay can build reputation cheaply
(by selling something cheap in quantity), and then
profitably use-up that reputation (by executing a scam to
sell a costly item).
In contrast, a sincere user who misunderstood the posting
rules or context may be scared away by the stigma now
attached to their name. Neither of these outcomes really
attains the positive result of getting improved behavior
from a mis-behaving member.
Frequent posters to our company‘s online communities
discussed the ways to handle posters who break the
unwritten rules of the newsgroup, such as using profanity,
knowingly deceiving others, or personally attacking
someone else. Although some offenders were publicly
corrected, these active community members preferred to
chastise misbehaving members privately, allowing the

Page 7

community to self-police, rather than having a corporate
entity discipline action.
In the spaces we examined, frequent and long-time
community members had a voice of authority. In other
spaces, semi-official moderators, selected from the
community‘s ranks and given some editing power, are
often a successful voice of authority.
There is a surprising counter-example to our discussion
of negative reputation. The discussion site Slashdot
provides negative reputation: messages that are judged by
the community to be ―negative‖ simply fall below the radar
of most readers; a user needs to set a particular setting to
see those (Lampe, Johnston, & Resnick 2003). Those
negatively labeled users take particular pride in being the
dark underbelly to Slashdot, and so continue to post using
their same user name and reputation, providing a back-
channel to the main conversation.
Positively incenting good behavior can be useful, as it
encourages and models those users who have done well.
Consider ways to incent the good while applying sanctions
in a graduated manner that matches the scale of the
infraction. This recommendation echoes those made by
political scientist Elinor Ostrom‘s study of self-organizing
resource cultivation and management groups (1991).
Organize and maintain
As an online community is developing, some design
choices can help the community gain and keep members;
others will cause the group to shrink. Tables of contents
can help users figure out where to look for a useful
reception and decide where to post. In this section, we
discuss the importance of helping groups maintain an
appropriate size.
7. Encourage Critical Mass
―We were ready,‖ the manager told us. ―We‘d built
one message board for each make and model of car on
our boards. A few thousand people stopped in, but
they never posted! Even when our front page would
promote one discussion board, that board would get
attention—but no one would do anything on the
others! Several of our boards have three messages on
them:



Hello, is anyone there?
Buy Cialias Cheap!
'I‘m here. Is anyone else?‖
Potential contributors who show up at a message board and
find it abandoned are unlikely to contribute. Optimally,
they would join a group that is bustling, but not so crowded
that their voices would not be heard. Finding a way to keep
groups full, but not crowded, is an important balance for
moderators.
Moderators who try to initiate too many online spaces at
once will dilute community involvement—questions will
be answered too late, if ever; discussions will move at such
a leisurely pace that participants will not check back to see
if they have gotten a response. This is what happened with
the discussion forum example above: a user would show
up, check on one or two car groups, and see that none of
them were particularly active, and wander away. Without
clear cues as to which boards have active participation and
which don‘t, users will find themselves lost.
Consider ways to help concentrate community, at least in
the early stages, in order to help users find a group with
active participants. Communities are often quite vocal
about wanting to split when they have more topic diversity
than they desire: the BMW owners will let you know if it is
important to separate out one particular model. A far
greater real risk is not allowing critical mass to form before
splitting a group.
This can be partially alleviated if the community has
access to cues suggesting where to look for content: for
example, it is not uncommon for many forums to label
which discussions are most active. This sort of presentation
of metadata to the reader can give them meaningful
information on where to look for content.
8. Exert Gentle Control
A team had decided to consolidate several groups
together in order to help make sure they kept critical
mass. ―We‘ll put the fitness discussion groups—male
and female—together, to help make sure they have
lots to talk about.‖
They posted notes to the forums, saying that the forum
would be merged in two months. The women‘s fitness
group sent a petition: they wanted to talk about diets
and post-pregnancy issues, and didn‘t think the men
could contribute. The men‘s fitness group simply
moved their conversation to a different provider, one
who was purely dedicated to men‘s fitness issues.
It is tempting to think of the participants in an online
community as ―your own‖. The organizer has provided the
space where they communicate and put a great deal of
effort into sustaining it. However, most users participate
out of their own free will, and may have strong opinions
about their host‘s product or service.
Discussion forums, then, resemble a casual party: the
host can change the music, but guests might leave. The host
can suggest they do something else—but others may go to a
different party entirely. As such, too heavy a hand in
control—trying to censor content or direct conversation—
can alienate users.
Hosts who develop a reputation for throwing good
parties may have more influence over their guest‘s
behavior. Even with unpopular rules, users may continue to
participate if community is valuable enough. A TiVo
community, for example, was able to sustain their user-base
even though the popular topic of stealing service was
repeatedly deleted by the TiVo-employed organizers.
(Powazek 2002a, pg 228)
When a forum really does need to be discontinued, it is
worthwhile to consider ways that the membership can

Page 8

maintain connections, especially if the managers intend to
host communities on similar topics. Leave a marker of
where the forum used to be, and if at all possible, leave
archives available (Powazek 2002a, pg 250). If these
optimal suggestions cannot be followed, try to give
participants enough notice that they can create their own
archives. Participants who are not left with this assurance
may not trust this hosting in the future, and may be more
withdrawn about posting to other fora.
Conclusions
A growing number of online communities are being
deployed with the hopes that their investment will be
returned with improved customer satisfaction, customer
loyalty, better communication, and a sense of membership
or association with a company‘s products or services.
Benefits to the customers from being able to communicate
with one another are enormous.
Despite these potential benefits, many investments in
community can underperform. We hope that this collection
of observations about online community cultivation and
management offers useful guidance to practitioners that can
improve the benefits of community while avoiding
common pitfalls.
We encourage researchers and community organizers to
test our suggestions, and publish their own experiences in
order to build a growing knowledge base about how online
communities interact in practice.
We close with the enthusiastic words from one
participant, discussing the social benefits of being an expert
member of a newsgroup:
―[T]hat group of people that goes out there to
newsgroups and sends questions and answers every
day, after awhile are more than simply technical guys
trying to find answers to their questions. They‘re some
kind of group of friends. You can see from the text
messages that it goes more to the friendship area, it‘s
not only trying to solve the question, it‘s also trying to
help people – trying to help friends that are passing
through the same path you passed before. And after
that it‘s great to meet those people anywhere else
because you feel like you have friends out there. …
Where your only contact is your keyboard and your
screen. It is something that far more than just typical
contact. At least, that‘s what I feel.‖
Acknowledgements
Our thanks to AJ Brush and Paul Johns for extensive
discussion of these issues. In addition, we thank our focus
group participants who shared their experiences – both
positive and those that need improvement. We further
thank community leaders throughout Microsoft who have
explored these situations with us.
References
Adler, B, and Alfaro, L. A Content-Driven Reputation
System for the Wikipedia. 2007. In WWW 2007,
Proceedings of the 16th International World Wide Web
Conference, ACM Press.
Arguello, J., Butler, B. S., Joyce, L., Kraut, R., Ling, K. S.,
& Wang, X. 2006. Talk to me: Foundations for successful
individual-group interactions in online communities. In
CHI 2006. ACM Press.
Beenen, G., Ling, K,, Wang, X., Chang, K., Frankowski,
D., Resnick, P., and Kraut, R.. 2004 Using Social
Psychology to Motivate Contributions to Online
Communities. In CSCW 2004.
Bruckman, A., and Resnick, M. 1995. The MEDIAMOO
project: Constructionism and professional community.
Convergence 1,1 (Spring).
Butler, B., Sproull, L., Kiesler, S., Kraut, R., 2008.
―Community Effort in Online Groups: Who Does the Work
and Why?‖ Leadership at a Distance, S. Weisband & L.
Atwater (Eds.)
[http://www.katz.pitt.edu/fac_pages/Butler.htm],
forthcoming.
Constant, D., Sproull, L., Kiesler, S., 1997 ―The kindness
of strangers: on the usefulness of electronic weak ties for
technical advice‖. In Culture of the Internet, S. Kiesler
(Ed.), Lawrence Erlbaum Associates, New Jersey.
Cosley, D., Frankowski, D., Kiesler, S., Terveen, L., and
Riedl, J. 2005. How oversight improves member-
maintained communities. In Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems
(Portland, Oregon, USA, April 02 - 07, 2005). CHI '05.
ACM, New York, NY, 11-20. DOI=
http://doi.acm.org/10.1145/1054972.1054975
Cothrel, J. and Johnston, B. 2007. Online Community
Return on Investment: State of the Practice 2007.
Presentation at the Online Community Business Forum.
ForumOne Communications, Sonoma, California.
Retrieved November 15, 2007, from
http://www.onlinecommunityreport.com/images/presentatio
ns/Business_Forum_ROI_final.pdf
Ellis, J., Halverson, C., Erickson, T, 2005. Sustaining
Community – Incentive Mechanisms in Online Systems:
Final Report of the Group 2005 Workshop. Retrieved May
16, 2007, from
http://www.visi.com/~snowfall/Group05IncentivesReport.p
df
Erickson, T. 2003. Designing visualizations of social
activity: six claims. In Proceedings of CHI 2003. ACM
Press.
Grudin, J. 1998. Why CSCW Applications fail: problems in
the design and evaluation of organizational interfaces. In
Proceedings of CSCW. ACM Press.
Hill, W., Hollan, J., Wrobleski, D., McCandless, T, 1992.
Edit Wear and Read Wear. In Proceedings of CHI 1992.
ACM Press.

Page 9

Hudson, J. and Bruckman, A. 2002. "IRC Francais: The
Creation of an Internet-Based SLA Community." Computer
Assisted Language Learning (CALL) 15(2), 109-134.
Kim, Amy Jo. 2000. Community Building on the Web.
Berkeley: Peachpit.
Krathwohl, David, 1998. Methods of Educational and
Social Science Research: An Integrated Approach. 2nd
Edition. New York: Addison Wesley Longman.
Lampe, C. and Resnick, P. 2004. Slash(dot) and Burn:
Distributed Moderation in a Large Online Conversation
Space. In Proceedings of CHI 2004. ACM Press.
Lampe, C., Johnston, E. Resnick, P., 2007. Follow the
Reader: Filtering Comments on Slashdot. In Proceedings of
CHI 2007. ACM Press.
Nonnecke, B. & Preece, J. 2003. Silent Participants:
Getting to Know Lurkers Better. In C. Leug & D. Fisher
(Eds.) From Usenet to CoWebs: Interacting with Social
Information Spaces. Amsterdam:Springer-Verlag.
Ostrom, E. 1991. Governing the Commons: The Evolution
of Institutions for Collective Action. New York: Cambridge
University Press.
Powazek, D. 2002a. Design for Community: The Art of
Connecting Real People in Virtual Places. Indianapolis:
New Riders
Powazek, D. 2002b. ―Gaming the System: How
Moderation Tools Can Backfire‖. Retrieved from the
Internet Archive on December 1, 2007.
http://web.archive.org/web/20030604072605/http://designf
orcommunity.com/display.cgi/20020531108
Preece, J. 2000. Online Communities: Designing Usability,
Supporting Sociability. Chichester, England: John Wiley.
Rashid, A., Ling, K., Kraut, R., & Riedl, J. 2006.
Motivating participations by displaying the value of
contribution. In Proceedings of CHI 2006: ACM Press.
Ren, Y., Kraut, R., Kiesler, S., 2007. Applying Common
Identity and Bond Theory to Design of Online
Communities. Organization Studies. Vol 28, No 3. 377-
408.
Resnick, P., Zeckhauser, R., Swanson, J., and Lockwood,
K., 2006. The Value of Reputaiton on eBay: A Controlled
Experiment. Experimental Economics 9:2 (June 2006). Pg.
79-101.
Riehle, D. 2006. How and why Wikipedia works: an
interview with Angela Beesley, Elisabeth Bauer, and Kizu
Naoko. Proceedings of Wiki WikiSym ’06.
Turner, T., Smith, M., Fisher, D., and Welser, H. 2005.
Picturing Usenet: Mapping computer-mediated collective
action. In Journal of Computer-Mediated Communication
10(4) article 7.
Viégas, F. and Smith, M. 2004. "Newsgroup Crowds and
AuthorLines: Visualizing the Activity of Individuals in
Conversational Cyberspaces", in HICSS 2004: IEEE.
Wenger, E., McDermott, R., and Snyder, W. Cultivating
Communities of Practice: A Guide to Managing
Knowledge. Cambridge: Harvard Business School Press.
2002.