Engineering Serendipity, Part 2.
Engineering Serendipity, Part 2.
Leveraging Network Analysis to Cultivate Vibrant Start-up Ecosystems.
“Writing is the process by which you realize that you do not understand what you are talking about. Importantly, writing is also the process by which you figure it out.” -Farnam Street
The SKA team seems to regularly discuss the weekly newsletter published by the Farnam Street team led by Shane Parrish, and their posts are definitely a regular inspiration. As a team of quantitative thinkers, we sometimes find setting time aside to write and reflect challenging. However, the process is always incredibly beneficial, and we are working to be more disciplined in this practice.
In our next post, we will update you on some of our recent project work. However, in this update, we will continue the discussion regarding startup ecosystem research.
As a quick reminder, our Managing Director, Dan Evans, has been conducting a long-running research effort at New York University to explore the use of machine intelligence in generating economic development policy recommendations to create more vibrant start-up ecosystems. This work is now officially expanding with a new collaborative effort with researchers from Stellenbosch University in South Africa.
As introduced in the previous post, the most popular approaches to mapping startup ecosystems were less useful for generating insights that lead to actionable policy recommendations than we envisioned. So, we explored other methodologies initially developed by social network analysts (and military experts) and then eventually created a new one.
So, what were our issues with these approaches?
Most importantly, we have a more ambitious end state in mind. In our work at West Point, we developed quantitative approaches to weaken networks. We found that it’s more effective to “prune” a network than eliminate the most central node as a first step, something that many law enforcement organizations have already discovered. If you eliminate the most influential node in an organized crime node, you create more instability and generally know less about the network than before. Instead, a better approach is to slowly roll up the network, making it harder for the network to react to these efforts.
How does this apply?
Our team's “big idea” was, "Now that we know how to ‘more effectively’ weaken a network, can we flip this idea and ‘more effectively’ strengthen a network?”
The Analytical Foundation.
To conduct this type of analysis, we need to:
Determine the best approach to map startup ecosystems,
Compare (and classify) ecosystems, and
Develop network intervention strategies (these will lead to policy/action recommendations) using network analysis methods.
The foundation for this idea is the concept of network topologies.
What is a network topology?
In simple terms, all networks have distinctive shapes, which can be quantified in many ways.
The three topologies illustrated above are simple examples. For example, The Star topology could model an organization with a strong central leader, while the Hub & Spoke could be the flight network of an airline. Finally, the Lattice could visualize a highway network.
In most cases, the world is “messier,” and the topologies of networks with many nodes and links may be harder to discern visually. For example, is it possible to classify the network's topology below? This is where advances in network theory and specialized quantitative methods can find the signal in the noise.
By Martin Grandjean - https://commons.wikimedia.org/w/index.php?curid=29364647
To more effectively differentiate between network topologies, our team has developed numerous methods of classification and visualization.
Innovative heat map visualization developed at West Point to compare network topologies.
A key step of this research effort will be identifying the main differences in the topologies of different startup ecosystems. Analysts can then compare a vibrant ecosystem to a less vibrant one and mathematically identify those differences. This process identifies the absence or presence of particular nodes or links. Based on this knowledge, network analysis and intervention algorithms can generate a list of prioritized courses of action that will more likely yield the desired results. In this case, a more vibrant startup ecosystem.
In the next post in this series, we will discuss the leading approaches to building these network models.
If you have an interest in discussing this research or are interested in “turning complex data into knowledge that matters,” please leave a comment or contact us directly.