Introduction to Organisational Network Analysis
I am a data science masters student from Cardiff Metropolitan tasked with researching networks. I am conducting this research as an associate at The Satori Lab, the company helping organisations learn how to deliver excellent public services in the connected age. Armed with tools such as Python, Gephi and Polinode and with much reading to be done, I aim to delve into the world of nodes, edges and algorithms and so we can use it to tell you how your organisation really works. Together with Cardiff Metropolitan University and Welsh Government, The Satori Lab is running a project to further the use of Organisational Network Analysis within the public sector.
Organisational Network Analysis aims to study the connections and communication between entities within a organisation. The organisation is modelled (statistical and graphical) based on people, projects, resources, knowledge. Tools are then used to interrogate the network.
What I’ve already learnt
In my first week I started with the basics of Social Network Analysis and how these applied to organisations. I also looked into centrality algorithms and what they can tell us about networks.
In example 1, people are represented as Nodes. Alice & Bob are nodes that have a connection. This could be that Alice is Bob’s boss, or Bob provides technical support to Alice. This connection is represented with Links.
From these nodes and links, we can start to pull out behaviours and patterns. Centrality is a measure of how individual nodes communicate with each other. There are several different types of centrality, and these can show us different characteristics.
In example 2, the nodes are coloured by who has the highest closeness centrality score. It is calculated on how many connections a node has to others in the network (how central a node is). This indicates which individuals might be good information broadcasters as they have connections to many people.
Betweeness Centrality is scored on the number of times a node acts as a bridge along the shortest path between two other nodes (acting as a bridge). Example 3 colours nodes by betweenness centrality score, the highest scoring being in yellow. This indicates the individual can mediate communications between teams.
Eigenvector Centrality (You’re important if you know important people). It gives weights to the nodes based on the relative weights of their connections (rather than amount of connections). Those with a high eigenvector score could be in positions of power or influence in the organisation. Google’s Pagerank algorithm is a derivative of Eigenvector using website links connected to them to score.
Networks can be directed which gives us more insight into the way the nodes behave. Suppose Alice is a decision maker. Bob and Chris rely on her to make decisions. But she does not rely on them. The link between Alice and Bob is directional: it flows one way. The more directional nodes that point to Alice the higher her “in-degree” The more Bob relies on other people the more links point away from him and the higher his “out degree”. Example 5 shows a central node with a high “in-degree”.
These are trivial examples and in reality the network will be big and messy. The tools we use can run algorithms which do these calculations on the entire network. We need these algorithms to do the work for us and make sense of it all.
Why this is useful?
Your organisation may well be structured using a traditional hierarchy chart or orgchart. Who is whose boss? But this doesn’t give you the real picture of how the organisation works.
Using tools such as surveys, email records and enterprise social media data (slack, yammer) to build the network rather than using the existing org chart gives us the ability to see real patterns, flaws and communication breakdowns within. Is there an over burdened employee? Are different teams not communicating who should be?
Our aim is to track these deviations and use it to give insights into how the organisation can change or harness this resistance.
Whats Next?
So far I have focused on the simple explanations of the networks based on centrality. The next step is to look at more algorithms such as Louvain Community Detection to look at how people group themselves in the organisation.
Our methods of data collection are also important. Increasing mining abilities and tailoring collection methods to the specific business problems can give us richer networks for analysis.
If you wish to keep up to date with this project, follow my medium page
Alternatively, click here to find out how ONA can help your organisation.
Thanks for reading.
I work at The Satori Lab.
I’m new. Help me to settle in on Twitter.
