Two weeks ago I wrote a post on what the research is saying about small world networks and how these networks can make groups innovative. Just to recap, there has been a great deal of progress in network science that allows us to understand how large groups that don’t appear to have many connections relative to the size of the community (e.g. the population of the USA) can have a handful of ‘degrees of seperation’ between anyone in the network.

One of the surprising results from this work is that some ‘small worldedness’ is good, but too much results in decreased innovation output. What followed that post was an interesting discussion with Greg Satell, who has a very good understanding of the topic. He has since followed up with an excellent description of Uzzi and Spiro’s Broadway study (that actually manages to make network statistics sound sexy!). The key to the main result is the relationship between the small world coefficient, a measure of clustering divided by average path length through the network, and creative output. After reaching an optimum point, more clustered networks become less innovative, even if some connections exist between the clusters.

But quite correctly, in the online discussion, Greg says understanding the dynamics of these networks is more important in helping us manage poorly performing networks. To quote Greg:

As networks age, triads tend to close and clusters form. If newer links aren’t introduced, you get exactly the situation you described.

A recent publication in Organization Science (which publishes some of the best network research) provides solid support for what Greg is saying. In general, as networks age, clusters do tend to turn in on themselves and are less likely to make bridges to other clusters. This has negative implications for the diversity of knowledge and the liklihood of new combinations of knowledge that produce breakthrough product in services. I’ll just share some of the main graphs where the networks are drawn from collaboration data between firms in the US comptuter industry between 1990 and 2005.

This chart shows the rise in the small world quotient as the US computer industry matures and then shows it falling away after 2000.

Rise and Fall of the Small World Network

The main cause of the decline in the small world was the loss of bridging ties between clusters. When clusters lose the critical links between them, small worlds are replaced by industrial islands.

Mature Networks Create Fewer Bridges Between Clusters

This is a really interesting result and the authors, Ranjay Gulati (Harvard U), Max Sytch (U of Michigan) and Adam Tatarynowicz (U of Tilburg) suggest that social factors within the clusters can prevent the formation of bridges with other clusters. Closed clusters have a high level of knowledge homogeneity and this can develop into the ‘not invented here’ syndrome. This makes people within the cluster less likely to look outside the group that they know. Clearly, this type of cluster will be less attractive to outsiders. Anecdotally, this sounds like a plausible explanation.

The result of the collapse in the small world in this industry is quite stark. While firms in the industry are still innovative, the scope of these innovations decreases. Bringing in new ideas from different industries starts to become more infrequent. Given that radical innovations are often the result of connections between different industries and technologies, this has implications for the lomg-term competitiveness of firms in these networks.

The takeaway from this is that older networks (within organizations or industries) will naturally fragment into clusters of firms or people who are similar. Sustaining the small world to encourage diversity and novel combinations of knowledge means creating the conditions for bridging the clusters. It’s not a case of more connections are better, the focus needs to be on the bridges.

In his post, Greg has some good suggestions about maintaining these bridges.

Getting people to collaborate is obviously a key success factor in any business. Through the years, I’ve found three practices to be helpful in keeping “Q” in the range where people can work together effectively but don’t fall into a rut:

Training programs: Besides the obvious knowledge transfer function of training, one often overlooked benefit is that it gets people to meet those they normally wouldn’t.

Generalized graduate training programs can be especially helpful in this regard. After the training is completed, new hires go to different parts of the company but maintain ties to their training class.

Best Practice Programs: Holding monthly meetings across departments for people to show off their best work not only encourages good work, but helps people to mix and develop relationships across the organization.

Forced Switching: Another, more difficult, practice is to force people to switch departments and functions. This often encounters resistance because managers are inclined to want to “protect their team” and people like to stay where they are comfortable.

Thanks to Greg for the conversation on small worlds.