posted 7 Sep 2005 in Volume 9 Issue 1
Social-network analysis Part III
A masterclass covering the use of social-network analysis in an organisational setting. Part three: ONA – Getting to the good questions.
By Patti Anklam
In previous articles in this series, I provided some basics on setting the context for using Organisational/Social Network Analysis (ONA/SNA) and the required elements for designing a successful ONA project, including a summary of available software and tools. I’m a firm believer in the adage, “Measure twice, cut once.” I’ve also learnt from experience that if you launch a survey without paying attention to the details, you risk having a low response rate, disgruntled staff and incomplete results. But once you have completed the successful design and launched the survey, you can get to what for many is the exciting part: those first looks at the drawings of the network, and those first insights from some of the key metrics. Then comes the “hard” part: working with the people in that network you’ve analysed to pull out commitments to meaningful action.
Starting the analysis – getting organised
Getting started on the analysis work requires:
Some disciplined file organisation;
A checklist of views you want to look at.
When you work on the analytic side of the process, you will have your source input files from the survey. Often these will be spreadsheet data files or simple data files. You’ll want to be sure that you have a clean folder on your desktop to keep track of the output files you produce. I typically structure my project files as outlined in Figure 1.
Once you’ve got the structure set up, you begin working through the views that you think are important to look at. Your work files will reflect the views that you’ve listed as your starting points. Organise this in whatever way makes sense to you, but you might find it helpful to start with a grid that lists the network questions that you asked in the survey and the demographics that you collected, as shown in Table 1.
You see from this that just at the outset of the analysis you have twenty potential views. For example, you might want to look at the impact of job level in the decision-making network, or the impact of geographic location on the exchange and flow of information. Your pre-existing understanding of the organisation will guide you toward the “most interesting” questions to start with. While this approach gives you twenty views to start with, you also want to consider the differences that you will see when you “cut off” the data at particular tie strength (see Figure 2).
Each software tool has its own sequence of commands and keystrokes to create views based on the tie strength. The software also lets you change the width or colors of lines to show the difference in frequency or strength.
Probing for meaning in the visual patterns
As you work through the result grid, you should keep notes of the interesting patterns – that is, the patterns that lead you to meaningful questions. Most of the questions will surface around fragmentation, isolation and the roles of specific individuals. The demographic attributes that you collect in the survey process are usually selected so you can show potential sources of fragmention – things that have prevented people from connecting and developing relationships. The two maps in Figure 3 show the same network – of frequent to very frequent information exchange – highlighted using two different demographic attributes.
The fun part of the analysis – for you as well as the people you are working with – is the way that you can work with hypotheses and scenarios. The “what if” views you can create often lead to more meaningful insights, for example: “What if we remove the manager of the group?” Often the key roles of people in certain roles can be seen by removing them. For example, Figure 4 shows what happened in one analysis when we removed all the senior managers and the administrative assistants.
As the analyst and consultant, it is your role to look at the possible variations in presenting the data so that the people in this network are moved to say, “Aha!” or to ask more probing questions. For example, a pattern like the one illustrated in Figure 5 leads to the question, “What do we need to know about the context behind this?”
Getting an understanding of the background context is the most essential part of an analysis. I feel that I cannot emphasise enough to students in my masterclasses and to my clients that the survey results generate questions, not answers. The goal of the initial analysis part is to look for the interesting questions and then probe deeper. One of the deeper places to look is in the metric data created by the analysis software.
Delving into the metrics
The network maps and their patterns are very revealing and as a presentation device speak to the “right side” of people’s brains. Relationships are easy to visualise and humanise the data. However, the visuals provide only part of the results. Metrics provide additional insights – and offer more questions – but they can also provide baselines: statistics about the networks that you can use for comparing organisations and subgroups (or comparing your organisation to a similar organisation in another company), or that you can use as a starting point to measure improvements over time.
There are two categories of metric:
Positional, or “centrality” metrics – used to identify the roles of specific people within a network and the extent to which people are either very central to the network or on the periphery;
Structural metrics, or what I like to call the “baseline” metrics – used to provide data about the network as whole.
Analysis software provides the data based on the network you specify (which can be a network that has been subdivided in some way). There are numerous statistical analyses available from the software; some of the specific measures that we use to describe organisational networks are summarised in Table 2.
Just as you look for interesting visual patterns with the network maps you create, you will want to generate metrics to correspond to the maps that you highlighted. For individuals you want to look at in networks, the degree and betweenness metrics can be particularly revealing. Figure 6 shows an example of how you need to look at different metrics to make sure you are asking the right questions.
Baseline metrics will help guide people in the organisation to look for areas for improvement. For example, Table 3 shows a set of metrics for several different organisations within a professional services business.
In the example of the groups in Table 3, measures for the innovation group indicate a high group centrality, one of the highest distance averages and one of the lowest densities. Although none of these metrics would be alarming in and of itself, seeing them in this context enables a more thoughtful analysis. Comparative metrics like this help not only to highlight potential areas for improvement but also potential best practices. For example, the Product Line A exhibits high density, very low distance and group centrality and a good average degree. What, we inquired, was this group doing that others were not that resulted in such positive metrics?
Answering the questions
As you can see, the analysis of the data can take you into many different directions; you will need to be guided by the concerns of the client with respect to where to delve deeply. These concerns arise through the process of following up the many questions that the ONA provokes. You’ll want to review the results with the project sponsor as soon as you can – the second stage of the analysis is to get into the context of the organisation to:
Ensure that the data (both maps and metrics) doesn’t give any false impressions. Typically, people who are new to the organisation are not very connected (if at all). You want to be sure that these people are identified as newcomers;
Identify key people in the network to get an understanding of the impact of specific patterns on the individuals and the organisation. For example, you’ll want to interview people who are very central (and between!) many others to determine how they are managing the role.
Interviews to obtain context in either of these areas is vital to the success of the ONA project. Working with the client or sponsor team, you can identify the people whom you should talk to and set up time to ask them a few basic questions. You can show these people some of the maps that indicate their position in the network to prompt them to reflect on their role and give you the context you need. Recall the “kit” network shown earlier. We may have been tempted to conclude that this project manager was being a bottleneck of information flow in the network, but when she described the nature of the work with the contractors, we understood that this was appropriate for the work at hand.
After interviews with the key people, you can get ready for the conversations for action – and prepare to facilitate those conversations. We’ll get to that in the next and final article in this series.
Acknowledgment: Thanks to Andrew Parker and Rob Cross for supplying one of their training datasets for me to use in developing examples for this article.