1. BY ERIC VAN HECK AND
PETER VERVEST
SMART BUSINESS
NETWORKS:
HOW THE
NETWORK
WINS
Realizing scenarios in
which business is conducted
through a rapidly formed
O
network with anyone, anywhere,
nce merely a dream, dig- anytime regardless of different
ital networks are now a
rapidly maturing reality. computer systems and
These digital networks business processes.
can expose unexpected
behavioral properties of the individual
actors. Combined as a swarm, networked
businesses are able to produce exceptional
or “smart” results they were not previously
capable of generating. Companies make
different linkages, combine different capa-
bilities from many different parties, are
more agile, and move positions faster.
What are these “smart” business networks
and why are they important? What should
chief information officers and IT profes-
sionals do to help their companies succeed
in a networked world?
BEING SMART IN THE BUSINESS NETWORK
In less than 10 years, Amazon moved
from electronic book retailing to become
I l l u s t r a t i o n b y KEN ORVIDAS
COMMUNICATIONS OF THE ACM June 2007/Vol. 50, No. 6 29
2. the world’s leading “e-tailer.” Without stores and Consider the example of Kenny’s Bookshop and
with limited inventory, Amazon possibly has more Art Gallery, a family-run small business in Galway,
information on retail goods and their buyers and Ireland (www.kennys.ie). It sources valuable sec-
sellers than many other businesses. Amazon offers ondhand books and sells them to interested collec-
a business platform for the traditional retailer to tors [5]. Established in 1940, Kenny’s began
“make a market” [9]. Within the Amazon business focusing on its bookshop customers and then, in
networks, the retailer can: the 1970s, by mailing paper catalogs to libraries
overseas. In 2003, Kenny’s linked its library man-
Table 1. Characteristics of traditional andto the Online Computer Library
• Facilitate search by buyers and sellers; agement system new
business network approaches.
• Discover and help set pricing and other transac- Center (OCLC), a not-for-profit organization that
tion conditions; owns the largest database
• Manage and coordi- Characteristics Traditional Business New Business of bibliographic records
Network Approach Network Approach
nate the logistical in the world. Kenny’s
Products and Relative simple, unbundled, Relative complex, bundled,
services and slowly delivered and fast delivered products
processes for transfer was the first business to
of the physical or digi- products and services and services have a commercial
tal goods; Value creation Supply chains with long Demand networks with arrangement with
term connected quick connect and
relationships disconnect relationships
• Settle payments and OCLC, which allowed it
arrange fund transfer; to provide a full elec-
Coordination and Hierarchical and central Network orchestration
and control control and decision making with distributed control tronic catalog on virtu-
and decision making
• Authenticate the quality ally any secondhand
Information Information sharing with Information sharing over
sharing direct business partners and with network partners
of the goods sold and book highly efficiently
verify the credibility of and faster than anyone
Infrastructure Actor platforms with Network platform with
buyers and sellers. information silos and networked business else. Kenny’s reaped
systems operating system
instant financial rewards
Thousands of elec- since a cataloged second-
Table 1. Characteristics of
tronic retailers join traditional and new business
hand book is usually valued four times higher than
Amazon every month network approaches. an uncataloged item. Through Bookrouter.com,
for all or some of these Kenny’s published its rapidly growing stock of
VanHeck table 1 (6/07)
functions. At the same time many leave, or are now-cataloged secondhand books on multiple
rebuked by the Amazon system. Amazon facilitates Internet sites such as Amazon, Alibris, and Bib-
product representation, regulatory compliance, liodirect, using their, or other logistics providers,
risk management, and conflict resolution; it has physical delivery capabilities. In 2006, Kenny’s
quickly established a reputation for trustworthy bookshop went completely online and their shop
transactions. EBay, with over 222 million regis- in Galway is now just hosting the art gallery.
tered buyers and sellers, has done the same for auc- Kenny’s could have remained a traditional book-
tions. At the end of 2006, Skype had attracted 171 shop with others—Amazon and eBay—capturing
million registered users. In less than four years, their business. They did the reverse. They became
Skype has surpassed anything traditional telephone smart in their business network by capturing a
service providers have ever achieved. valuable position and leveraging that position
These companies offer platforms on which users across as many links as they could.
can freely move and interact as long as the platform Capturing and leveraging a position in a busi-
provider allows them. These platforms show a strong ness network does not mean one must own, or
network effect: the more users, the more useful the control, the platforms on which those networks
network becomes, the more difficult it becomes to run. For example, TheBigWord company
switch, and the less likely the user will move to (www.thebigword.com) is able to serve the diverse
another network. Albert László Barabási [1] recog- translation needs of large companies worldwide by
nized this by analyzing Internet traffic. He demon- sharing a “translation memory” across a network of
strated that the Net is not democratic, and the its clients and thousands of local “mother tongue”
number of links per node follows a so-called power linguists [7]. It responds almost instantly to trans-
law distribution. A few nodes have many links while lation needs—such as for publishing on Web sites
many more other nodes have very few links. The in many different languages—by posting the work
node with many links attracts nodes with fewer links to targeted groups of qualified translators, dividing
faster than the lesser-connected nodes. As the big get and allocating the work, and managing the process
bigger, what options do the smaller actors have? in a way that is fully transparent to the client.
30 June 2007/Vol. 50, No. 6 COMMUNICATIONS OF THE ACM
4. TheBigWord rewards the translators by paying by tencia. They show how the smart business network
the number of words they translate and by provid- approach with embedded business processes leads
ing their administration and payment services. to substantial business advantages, demonstrating
TheBigWord example demonstrates these business the importance of information sharing in the busi-
networks can respond with much more agility ness network and the design and organizational
together than acting as an individual company. But dynamics of the infrastructure.
what are other characteristics and what are the rea- The pivotal question of smart business net-
sons why these new forms are starting to be devel- works concerns the relationship between the strat-
oped and implemented now? egy and structure of the business network on one
hand and the underlying infrastructure on the
A NEW BUSINESS NETWORK APPROACH other. As new technologies, such as RFID, allow
Organizations are moving, or must move, from networks of organizations almost complete insight
today’s relatively stable and slow-moving business into where their people, materials, suppliers and
networks to an open digital platform where busi- customers are at any point in time, the organiza-
ness is conducted across a rapidly formed network tions are able to organize differently. But if all
with anyone, anywhere, anytime despite different other players in the network space have that same
business processes and computer systems. Table 1 insight, the result of the interactions may not be
provides an overview of the characteristics of the competitive. Therefore, a first critical step is to
traditional and new business network approaches. develop a profound understanding about the func-
The disadvantages and associated costs of the more tioning of the business network.1
traditional approaches are caused by the inability
to provide relatively complex, bundled, and UNDERSTANDING THE NETWORK
quickly delivered products and services. The If two cars drive on the highway with enough dis-
potential of the new business network approach is tance between them they have no relationship other
to create these types of products and services with than to share the same roadway. If, however, these
the help of combining business network insights same cars get very close, they start behaving differ-
with telecommunication capabilities. ently. If the first car brakes, the second car will
The business is no longer a self-contained orga- brake—but with a delay. If the first car accelerates,
nization working together with closely coupled the second car will also speed up; again, with a
partners: it is a participant in a number of net- delay. The drivers of these two cars may not notice
works where it may lead or act together with oth- much more than the distance between them. But
ers. The network includes additional layers of from a helicopter their behavior will appear as a
meaning—from the ICT infrastructures to the wave moving along the flow of traffic. Each partic-
interactions between businesses and individuals. ipant does not see the behavior of the network but
Rather than viewing the business as a sequential responds to the local situation with his or her dri-
chain of events (a value chain), actors in a smart ving logic. The impact of the individual driver’s
business network seek linkages that are novel and actions in response to their specific situation and
different to create remarkable, “better than usual” the road rules they follow creates a collective net-
results. “Smart” has a connotation with fashionable work behavior not seen nor understood by the indi-
and distinguished but can also be somewhat short- viduals. Each driver acts on self-organizing driving
lived. What is smart today will be considered com- logic according to the driving rules of the network.
mon tomorrow. Smart is therefore a relative rather The study of networked behavior beyond the
than an absolute term. Smartness means the net- familiar territories of business and ICT networks—
work of cooperating businesses can create better those of social interactions, ant colonies, bees, and
results than other, less smart, business networks or other biological systems—reveals attributes and
other forms of business arrangement. To be smart characteristics that can be applied to the design of
in business is to be smarter than the competitors smart business networks. The behavior of the indi-
just as an athlete who is considered fast means vidual drivers, as described in the preceding exam-
faster than the other competitors. ple, demonstrates swarm intelligence: the
Another way the new business network emergence of seemingly intelligent or, perhaps,
approach distinguishes itself is the way the network smart, behavior from many individuals [2]. Swarm
is orchestrated. In the sidebar “Multiasistencia:
The Network Orchestrator,” Busquets, Rodón, and 1
See “Decision Making in Very Large Networks,” by Peter J. Denning and Rick Hayes-
Wareham introduce the Spanish Grupo Multiasis- Roth in the November 2006 issue of Communications for more detailed information.
32 June 2007/Vol. 50, No. 6 COMMUNICATIONS OF THE ACM
5. intelligence studies collective behavior in self-orga- As the analysis is applied to larger and more com-
nizing systems populated by simple individuals plex networks, more advanced ways are required to
interacting locally with one another and with their analyze the structure of the network. In the sidebar
environment without centralized control. However, “Network Horizon and Obtaining a Favorable Net-
in many cases, despite being unpredictable, such work Position,” van Liere and Koppius use social
swarms are able to exhibit impressive capabilities network analysis and simulation techniques to
for problem solving to, for example, seek food or explore the concept of the network horizon: the
respond to an unforeseen problem. number of nodes an actor can “see” from a specific
While these studies provide indicators for net- position in the network [12]. With a larger network
work dynamics—formation, change, decay—and horizon a company can take a more advantageous
for the ways in which the individual intelligence of network position depending on the distribution of
the network actors is combined, the research in the network horizons across all actors and up to a
social network analysis has made a significant con- certain saturation point. The results indicate the
tribution to a more profound understanding of expansion of the network horizon will soon be a crit-
network behavior. Social network researchers take ical success factor for companies.
into account the social relationships and ties of Most network scientists analyze the structure
individuals and therefore the structure of the net- and dynamics of business networks independent of
work. Building on social network analysis using the technologies that enable the networks to per-
complex systems theory Dan Braha and Yaneer form. Instead, researchers tend to concentrate on
Bar-Yam [3] examined the statistical properties of what makes the network effective, the linked rela-
large-scale product development information net- tionships between the actors, and how their intelli-
works for vehicle design. They find that such net- gence is combined to reach the network’s goals.
works have properties (sparseness, small world, Digital technologies play a fundamental role in
scaling regimes) like those of other biological, today’s networks. They have facilitated improve-
social, and technological networks. They demon- ments and fundamental changes in the ways in
strate that the distribution of incoming communi- which organizations and individuals interact and
cation links always has a cut-off—their numbers combine as well as revealing unexpected capabili-
are restricted—while the distribution of outgoing ties that create new markets and opportunities.
communication links is considered scale-free, One need only consider the rapid rise in digital
meaning some nodes act as highly connected hubs. social networks and massive online computer
This would be consistent with Herbert Simon’s games such as Second Life. These are exhibiting
bounded rationality argument that rational agents capabilities that seem well beyond those of existing
experience limits in formulating and solving com- business networks. The next critical step is to
plex problems and in processing (receiving, storing, develop a comprehensive understanding of the
retrieving, transmitting) information. expected smartness of the business network.
Braha and Bar-Yam found it seems easier to trans-
mit information than to process information. Like WHAT MAKES A SMART BUSINESS NETWORK?
individual human beings, a group of people—or The key characteristics of a smart business network
network of nodes—is limited by an inability to are that it has the ability to “rapidly pick, plug, and
digest an intense input of data. It seems that smart- play” business processes to configure rapidly to meet
ness could be related to the capability to organize the a specific objective, for example, to react to a cus-
information flows within the business network as tomer order or an unexpected situation (such as
well as to the topological structure of the network. dealing with emergencies) [11]. One might regard a
Other researchers have shown the attractiveness smart business network as an expectant web of par-
and importance of certain positions in the network, ticipants ready to jump into action (pick) and com-
that is, those nodes that are dominant and those bine rapidly (plug) to meet the requirements of a
that take subservient roles. For example, in 1992, specific situation (play). On completion, the partic-
Ronald Burt identified “bridging positions” where ipants are dispersed to “rest” while, perhaps, being
the network participants link through a focal actor active in other business networks or more traditional
who holds the bridge [4]. This structure brings supply chains.
information and control benefits (a central player) This combination of pick, plug, play, and dis-
but also encourages the dependent actors to find perse means the fundamental organizing capabilities
alternative routes, for example, to disintermediate for a smart business network are: the ability to
the bridger. quickly connect and disconnect with an actor; the
COMMUNICATIONS OF THE ACM June 2007/Vol. 50, No. 6 33
6. Table 2. Critical trade-offs and questions with regard to
smart business networks.
selection and execution of busi- Issues Trade-Offs Critical Questions
ness processes across the net- Network Preferred versus expected outcomes What do customers want in terms of
outcomes products and services (quality,
work; and establishing the Network versus individual outcomes and complexity, bundling, delivery time,
decision rules and the embed- performance price, offline/online)?
ded logic within the business How to value customer demand?
network. Network Individual versus network execution What can we offer to customers in
execution terms of products and services (quality,
Simple versus complex transactions complexity, bundling, delivery time,
QUICK CONNECT AND DISCONNECT price, offline/online)?
Global versus local execution
Quick connect requires that, as What actors will work together to
a result of an event the smart Integrated versus modular products/ fulfill the customer order? How?
services/processes
business network must seek and What level of modularity is required?
select those members who, Network Central governed versus adaptive, self Who is responsible for the customer order?
together, can fulfill the required governance organizing How are decision rights delegated?
goal. This means the network Governance to innovate versus governance
to operate efficiently How is the relational governance evaluated?
logic acts on a menu of poten-
Network
tial fulfillment partners to select design End-to-end control versus individual control How to keep end-to-end control in the
network?
those who can combine to pro- Networked business operating system
versus actor business operating systems How to design and test the networkked
duce the desired results. Once business operating system?
the appropriate participants are Global versus local adaptation
How to deal with the environmental and
found and the connection has Dynamic versus static business network internal dynamics of the business network?
been established, the process of Enabling Open technologies versus closed What technologies will be disruptive for
“play”—performing the busi- technologies technologies the business network?
ness transaction—can begin. Individual versus network technology How to speed up technology adoption in
adoption the business network?
Goldman, Nagel, and Preiss [6]
described this in their discus- Table 2. Critical product design and manufacturing enabling prod-
sion of virtual organizations. trade-offs and uct construction of tailored products from stan-
questions regarding
The capability of quickly con- smart business dardized components: the combination of
nected plug-compatibility networks. VanHeck modules (6/07) combined in a specific
Lego-like table 2 that are
enables a superior response way. Modularity brings the benefits of versatility
speed and greater component variety particularly (the diverse set of products that an organization
for dealing with new requirements. can produce) and agility (the ability to respond
While the ability to quickly connect has received quickly to fulfill an unpredictable customer order)
attention, the capability to quickly disconnect while, at the same time, delivering within the
requires more. Members will join the business net- boundaries of allowed value chain total costs and
work and participate on the basis of risk and lead times. Martijn Hoogeweegen et al. developed
reward. While this can be clear while they are a method to design modular business networks and
active, rules should be agreed upon for when the to optimize the allocation of tasks in a business
actors are disconnected (having completed a spe- network based on modularity principles [8]. The
cific customer order or while they will no longer be network nodes will be considered black boxes pro-
a member of the network). Decision rules and logic viding the functionality required by the business
with regard to connection and disconnection will network and are “played” according to the network
be a crucial component for the success of the busi- rules. However, modular design requires much
ness network. more coordination than non-modular: the greater
the number of components the higher the organiz-
PLUG AND PLAY: BRINGING THE NETWORK TOGETHER ing cost. A crucial decision is the degree of modu-
When selected, the network participants must be larity or granularity of a system, or business
able to interoperate. They must be plugged network, and that is determined by the balance
together to enable the required network outcome. between coordination costs and the complexity of
This means they must act with modularity: the the network.
decomposition of a system by grouping elements
into a smaller number of subsystems with rules EMBEDDED BUSINESS LOGIC
governing the architecture for mixing and match- Each business network participant has specific
ing these components. capabilities captured in its business processes (own
The concept of modularity has a long history in business logic) that it executes according to this
34 June 2007/Vol. 50, No. 6 COMMUNICATIONS OF THE ACM
7. NETWORK HORIZON AND OBTAINING A FAVORABLE NETWORK POSITION
By Diederik van Liere and Otto Koppius
n the networked world, decision makers must ensure their These findings might explain why network orchestrators in
I companies understand their position in the business net-
work. They should see their company as a node in much
larger and extensive networks of interdependent firms that are
a wide variety of industries are so successful. Companies such
as Li & Fung in the apparel industry, Solectron and Flextronics
in semiconductors, and Grupo Multiasistencia in home insur-
constantly being reconfigured. This perspective implies that deci- ance claims and repairs were among the first to see them-
sion makers must be acutely aware of their companies’ unique selves as nodes in a larger business network. In a
position in the network; not only the direct (van Liere) figure caption: Bridging position and extensive network
Sidebar (supplier or client) homogenous, low-horizon network, their
contacts but also the contacts’ contacts, their contacts, and so
network horizon in aenabled them to position themselves as network
horizon
network.
forth. The number of nodes that an actor can “see” from a spe- orchestrators connecting otherwise disconnected parts of the
cific position in the network is called the network horizon. The network and become significant companies in their respective
figure here illustrates how the industries while others
network horizon of Firm A is struggled to make the tran-
expanded using the opportuni- sition to network thinking.
ties of the bridging position in G G We foresee two chal-
a business network. Our stud- lenges for organizational
ies have indicated the network B B decision makers seeking
horizon is an important deter- F E F E the networked approach.
minant of both individual firm A A The first challenge is to
performance and the overall start to think about a net-
network dynamics and perfor- work strategy, rather than
D D
mance. C C seeking competitive advan-
Firm A occupies a bridging position Suppose firm A has a network horizon
Research was carried out tage based upon internal
by developing a management because it can broker information and that includes firms E, F and G and capabilities. By choosing
resources between firms B, C and D recognizes the opportunity to
while these are dependent on firm A. strengthen its bridging position.
exercise called the Business and maneuvering to the
Networking Engine [2] that can Therefore it establishes a relationship correct network position a
be used to test the effects of with firm F. firm can reap even more
new, networked organizations benefits from its current
based on the concepts of mod- capability-based competi-
ularity and loosely coupled busi- Bridging position and tive advantage. The second challenge is to identify which net-
ness processes. A series of network horizon. works are relevant for their business and where they can
experiments with executives expand their network horizon. As business networks continue
from the insurance industry to grow, the entire network cannot be scanned. Some parts of
showed that firms with a higher network horizon achieved higher
VanHeck favorable (Van Liere)more relevant than other the expansion of the
performance. These firms were better able to identify
sidebar the network are fig. (6/07)
advantage is increasingly network-based,
parts. As competitive
network positions where they could act as a bridge between dif- network horizon will be a crucial success factor in the near
ferent parts of the network: a position rich in structural holes [1]. future. c
Firms with a lower network horizon rarely identified such oppor-
tunities and hence had lower performance. References
As the network adapts to its members’ responses to 1. Burt, R.S. Structural Holes: The Social Structure of Competition.
changing market conditions and competitor actions, the bridg- Harvard University Press, Cambridge, MA, 1992.
ing position advantage may be temporary. So when does a 2. Hoogeweegen, M.R., van Liere, D.W., Vervest, P.H.M., Hagdorn
network position confer a sustainable competitive advantage? van der Meijden, L., and de Lepper, I. Strategizing for mass cus-
Extensive simulation revealed that a key factor is the way the tomization by playing the business networking game. Decision
network horizon is distributed across the different firms (in Support Systems 42, 3 (Mar. 2006), 1402–1412.
other words, the network horizon heterogeneity). The results
are striking. In the traditional supply chain where all firms More information about the Business Network Engine is available at:
bne.rsm.nl.
have a low network horizon (that is, the firms know their
direct upstream and downstream partners but are aware of
little else), a competitive advantage can be sustained for some Diederik van Liere (dliere@rsm.nl) is a Ph.D. candidate at
time. the Department of Decision and Information Sciences of RSM
However, when the network horizon becomes more Erasmus University in Rotterdam, The Netherlands.
heterogeneous across firms, for instance because some firms Otto Koppius (okoppius@rsm.nl) is an assistant professor at
start to think in network terms instead of supply chain terms, the Department of Decision and Information Sciences of RSM
the situation changes significantly. In heterogeneous networks, Erasmus University in Rotterdam, The Netherlands.
a low network horizon confers little advantage but having even
a modest network horizon can sustain a competitive advantage
for some time. However, once all firms have a high network
horizon, the network becomes homogenous again. Any
opportunity can be spotted by many firms and any competitive
advantage is therefore short-lived. 2007 ACM 0001-0782/07/0600 $5.00
COMMUNICATIONS OF THE ACM June 2007/Vol. 50, No. 6 35
8. Figure 1. The traditional business network approach.
logic. Traditionally, ing each participant to
AGENT
PRODUCER FORWARDER SHIPPING LINE RETAIL
when such participants execute in its own way
Transaction
combine they create Layer according to this logic.
interfaces between capa- BANK INSURER
This means that, to be a
END CONSUMER
bilities: translating from RFID
data
member of the network,
Scanning
Scanning
RFID data
one business logic to INLAND SHIPPING
an organization must be
data
Logistics
another and executing PRODUCER
able to absorb the
accordingly. This can be Layer shared logic and execute
SEASHIP
seen in the outsourcing ROAD TRANSPORT accordingly. This is the
DISTRIBUTION TERMINAL
phenomenon: carve out “own business logic” of
the total function of a the network that can be
particular business oper- Figure 1. The traditionalapproach. enabled by a Networked Business Operating Sys-
network
business
ation and hand it over to tem (BOS). Based on the service-oriented architec-
another party. As indicated earlier, traditional busi- ture it resolves the problem of information silos by
VanHeck fig 1 (6/07)-19.5 picas
ness network approaches lack the ability to rapidly loose coupling of underlying systems, which are
pick, plug, and play to configure rapidly to meet a connected together in a business operating layer.
specific objective, for example, to react to a cus- This layer allows process execution and manage-
tomer order or an unexpected event. Figure 1 pre- ment “from a distance” from the underlying appli-
sents part of a global business network. Its focus is cation systems. The enabling interorganizational
Figure 2. The new business network approach.
on the actors and relationships from manufacturers technology architecture must reflect this loose cou-
via multi-modal transportation (road, train, sea- pling. Loose coupling is not synonymous with
ship) to retailers. In decentralized processes.
most current business Networked Business
It is quite the opposite,
networks, companies Business Layer
Operating
Operating System
where the processes are
are developing capabili- more tightly coordinated
Standards
ties at the logistics layer EPC
because the rigidity of the
Global
and the transaction PRODUCER FORWARDER
IT architecture is no
AGENT SHIPPING LINE RETAIL
layer. As discussed in Transaction longer a constraint [10].
Layer
Table 1, these actors BANK
The business operating
INSURER
END CONSUMER
focus on their direct RFID Scanning
layer can become rather Scanning
data data
RFID
INLAND SHIPPING data
partners and are not complex due to the fact
able to have the end-to- Logistics
Layer
PRODUCER
that business logic is
SEASHIP
ROAD TRANSPORT
end management of DISTRIBUTION developed related to such
TERMINAL
processes running issues as:
across many different
Figure 2. The new business
organizations in many network approach. • Membership Selection: The capabilities to
different forms. The decide which business entities can act as nodes
actor platforms are of the network;
dominated by information silos residing either in • Linking: The positioning and connecting of
VanHeck fig 2 (6/07)-19.5 to the other parts of the network. The
different places within an organization as “islands,” nodes picas
or in two or more different organizations. Individ- linking processes can include the directories
ual actors are orchestrating processes in their part (search and select) and routing (path finding)
of the supply chain. through the network as well as typical commu-
Figure 2 presents an example situation illustrating nications tasks such as handshake, authentica-
the use of the new business network approach. The tion, and trust establishment;
central idea is that linking partners is on the basis of • Goal Setting: The coordination mechanisms
linking processes but allowing individual execution that determine goals in the business network
according to those processes: they act individually and the tasks and responsibilities assigned to
according to the joint rules of the network. each member node;
As shown in Figure 2, each of the smart business • Risk and Reward Management: The division of
network participants becomes a “smart insect” in a material results (profit and loss in a monetary
goal-seeking swarm. The network separates process but also in a fairly loose and generic sense) and
from execution. It shares the processes required to the perceived value by each of the participating
achieve its goals (the shared business logic) allow- business entities of its share;
36 June 2007/Vol. 50, No. 6 COMMUNICATIONS OF THE ACM