Working in the Connected World -
Managing Connected Assets 

  

Improving Information Flow by Valdis Krebs

Network ties are distributed unevenly. People that work together form networks together -- clusters emerge around established work relationships. Engineers working on Project X form a cluster, those working on Project Y form a cluster, and those working on Project Z form a cluster. Everyone knows everyone else within the local cluster, and yet only a few individuals have boundary spanning ties to other clusters. Strong, frequent, ties are usually found within clusters, while weaker, less frequent ties are found between clusters.

Clusters of concentrated connections appear throughout an organization and throughout industries. Some clusters have many ties to many others, while other clusters have only a few ties. Poor connections between clusters result in very long path lengths throughout the greater network. In such a network it is easy to access those in your cluster but not those in other clusters. This often results in distant clusters not having access to information and knowledge that is available in the organization.

Often the knowledge you need is in clusters other than your own. Networks have a horizon beyond which it is difficult to see what is happening. Research by Noah Friedkin at UC Santa Barbara has shown that this horizon of observability is usually two steps in a human network -- your direct contacts and their direct contacts. Around three steps out, things are real fuzzy -- you do not have a good idea of what is happening there. Beyond three steps, you are blind to what is happening in the rest of the network.

In a network of very long path lengths between clusters, your ability to find the knowledge or information you need is constrained. If the knowledge that you seek is not within your network horizon, then you assume it is not available and you reinvent it or pay for it on the outside. Exasperated with this network horizon in his organization, a former CEO of HP once lamented, "If we only knew what we know".

The natural response in many organizations is to throw technology at the problem. A very poor, yet quite common, solution is to mine the knowledge from employees, codify it, and store it in a knowledge database. Many large consulting firms tried this approach in the early to mid-1990s with usually poor results. They found that people were not always willing to make public their best knowledge and that codifying tacit knowledge was like trying to nail jelly to the wall. A better use of technology is to keep a database of "who knows what" and add a table of "who knows who." In other words, store pointers to the knowledge, not the knowledge itself. A hybrid solution may be to store both pointers and summary information chunks.

Why not use the power of the network itself to create a solution? Improve the organizational network and then use technology to help people communicate across wide spans of the human network. At first blush, improving an organization-wide network may seem an overwhelming task. Where do we start? First, look at the networks and communities of practice/interest/knowledge that have organized around a specific topic, product, service or customer. Usually the whole organization does not have to be included in the problem space. Second, map out the network nodes and their connections (who goes to whom for expertise/knowledge/advice on X?). From this network map, you can see the various clusters and how they are connected. Figure 1 below is a network map of project teams. A line connecting two teams indicates a two-way information flow or exchange of knowledge.

Figure 1

This network of 17 project teams all work on sub-assemblies to a larger product. The teams are composed of mostly engineers, technicians, and project managers. All teams have less than ten members. Three clusters are evident with the cluster composed of teams N, O, and P connecting the other two emergent clusters.

Before we look at how to improve the overall connectivity of the network, let's digress back to social capital. Which team has the best social capital in this network? Which team can access all of the knowledge and resources in the network quicker than the others? (Hint: this network is drawn to reveal the answer.)

Common wisdom in networks is "the more connections, the better." This is usually not true. What is always true is "the better connections, the better." Connections that provide the most network benefits give you access to nodes that you currently do not have access to. Although Team F and Team Q have many connections each and have excellent local access (to the nodes near them), they have only fair access to the rest of the network. Team O has the best social capital (aka network benefits) in this network of project teams. Team O achieves this with only three direct ties -- it is connected to others who are well connected.

The average path length in this network is 3.9 with many paths longer than the network horizon. Even in this small network there are nodes that are nearly blind to what is happening in other parts of the network.

In the summer of 1998, writing in the scientific journal Nature, a stir of excitement was generated by two mathematicians from Cornell, Steven Strogatz and Duncan Watts. Business Week, the New York Times, and many others reported on their research. They received a flood of emails from epidemiologists, computer scientists, marketing people, economists, engineers, government agencies and at least one management consultant (me). What attracted professionals in all of these diverse fields? Part of the attraction was the simple elegance of the solution. While investigating small-world networks (those with many clusters), they discovered that a few randomly added cross-cuts (network links) would improve a network's path length significantly -- very small adjustments could cause large positive changes. The benefits were not just local, but spread throughout the network.

Looking back on our project team network in Figure 1, how can we improve the connectivity with just one added link? Which two nodes would you connect to bring everyone in the network closer together? Although many combinations will increase the access of everyone to everyone else, the greatest measurable effect is when we add a cross-cut between Team Q and Team F. The average path length drops from 3.9 steps to 2.6 -- an improvement of 33%! Information flowing from any node in the network to any other node in the network now arrives much sooner and with less distortion.

Figure 2

The connection between Teams Q and F may be the optimal connection in network efficiency, but it may not be a practical connection. Both of these teams already have many ties and may not have the time and energy to support another one (remember Hansen's research above). What is an alternative connection? If you cannot connect the highly connected nodes, how about connecting their respective network neighbors? Instead of connecting Q and F, how about connecting D and Z? This connection will not reduce the path length as much, but it is between nodes that are not overburdened with connections.[See Figure 3 below].

Figure 3

Copyright © 1999, Valdis Krebs, All Rights Reserved. www.orgnet.com