Tips on collecting data for social network analysis (SNA)

As a healthy but slight change from recent topics, I was going through some old notes and thought it might be worthwhile sharing some tips with you about collecting data for Social Network Analysis (SNA). You might also like to check out my article about social networks, titled Small World!

Just as brief overview, Social Network Analysis (SNA) is a tool that provides a technique for analysing informal networks. SNA is an interesting Knowledge Management (KM) "technology" since it provides one of the few methods for making knowledge activities in an organisation visible through the analysis of relationships between people. However, while the concept of SNA has only become a hot topic in KM recently, it is not a new idea and has origins in social research from back in the 1930's and 1940's. It is the availability of off the shelf SNA software and improvements in desktop computing power that have made it feasible to conduct outside of an academic research setting.

One of the key benefits of applying computing power to SNA is the ability to visualise networks and manipulate these visualisations in real-time. The analysis of social networks using this approach may reveal new insights that might have missed through statistical analysis alone. In the context of KM, SNA acts as a diagnostic gap-analysis tool for social networks in organisations. Broadly speaking, the benefits of interventions based on SNA should include:

1. Promoting greater collaboration within groups or teams;
2. Encouraging boundary spanning communication; and
3. Enabling information to flow where bottlenecks or breakdown previously existed.

So that’s what its all about, but what is the problem with collecting data for SNA?

While the results of an SNA project are focused on the intangible connections between people (and there is a lot of focus on this fun part), the process of collecting data for SNA is actually similar in fashion to data mining and it requires a bit of planning to do it effectively. Now, methodologies from data mining can be used as a point of reference to provide a basic process for SNA:

1. Selection.
2. Pre-processing.
3. Transformation.
4. Data Mining.
5. Interpretation/Evaluation.

Now, building further on the leading practices in data mining, issues for SNA are also similar and of primary concern to SNA projects are the following issues:

• Scope - the dataset size and dimensionality;
• Analysis Process – How will users interact with the results of the SNA study;
• Missing data – unlike statistical analysis, you need to maximise the response rate, else the reported patterns may be misleading;
• Understanding the network patterns – have we really understand what we are seeing;
• Integration - between the survey tool, the SNA tool(s) used and other data that needs to be incorporated into the results; and
• Privacy and security – this goes without saying, and some stakeholder management may be required depending the circumstances.

Putting aside issues around interpretation in steps 4 and 5 (which is an art, more than a science – see my other article), the most typical problems I’ve seen related to data collection in steps 1 to 3 have included:

• Low response rates; and
• Poor data quality.

However, while these might appear to be minor issues and something to be expected with any survey, these can have major impact the validity of the subsequent steps in the SNA process. They also significantly increase the effort (and cost) required to successfully complete an SNA project.

So, unless you are intending to gather data for SNA automatically – e.g. wiki mining – then I offer to two key tips for smoothing the data collection process:

1. Ensure you *really* engage with the community that will be participating in the SNA survey – you might need to offer rewards for taking part and you should allow plenty of time to follow up with people who need lots of encouragement to provide their responses, including individual follow up phone calls. However, depending on your scope, with the right effort up front you can achieve near perfect response rates – it just takes some planned effort.
2. Design your survey collection to maximise the quality of data – in particular you need a high degree of consistency around name data – e.g. James, Jimmy and Jack could all be the same person, but how will you know? If you intend to match SNA data to other organisational data, such as role or location, then you also need those names to match that survey data.

Good luck! Of course, let me know if you need help with your SNA project or contact my good friends over at Optimice.

Technorati Tags: data mining,data collection,Social Network Analysis,SNA,wiki mining,Optimice