Adopting Analytics Culture: 6. What information is gained from social network analysis? (6 of 7)

PART 6 OF 7 IN A SERIES ON ADOPTING ANALYTICS CULTURE

LINK TO HEADER ARTICLE

LINK TO PREVIOUS ARTICLE (5 of 7)

6.    What particular information is gained from social network analysis and how is it interpreted?

To recap, we have established that adopting analytics culture requires organizational change management.  Beyond analytics technology and expertise, analytics culture depends upon effective organizational decision making practices.  In particular, evidence-based decision making best practices need to be sharpened via a focus on effective processes and supporting organizational architectures. It was asserted that decision effectiveness can be improved via concerted attempts to optimize analytical processes, decision rights, access to information, proper incentives, assessment systems, and communication pathways.

social network

However, it was revealed that the track record for corporate change management initiatives is quite poor.  It was proposed that two factors contribute to ineffective change management: 1) over-emphasizing the organizational chart, and 2) a lack of focus on relational interactions and networks.  Social network analysis (SNA) was proposed as a method for improving understanding of relational factors and engineering organizational network change.  This article goes into detail concerning the information that can be gleaned from SNA.

By characterizing the organization as a set of overlapping networks, beyond a singular management hierarchy or accounting artifact, deep insight into organizational structural dynamics becomes possible.  Collating simple data on communication ties in an organization allows maps of organizational structures to be extrapolated.  SNA provides the capacity to extract detailed maps and network quantitative measures by framing and interpreting data on actors and their relations.

For instance, surveying workers regarding ‘who they go to for advice’ allows one to generate a detailed map concerning exchanges of expertise, some of these exchanges which may be reciprocal, some of which may be one-way.  By aggregating all ‘ties’, a structured set of relations emerges as a network map.  It soon becomes visually and quantitatively apparent who the ‘go to’ people are, as well as those who ‘bridge’ information between subgroups.

Subsequent analysis can be conducted concerning demographics and actor attributes such as gender, seniority, and specialty.  From here, research questions and hypothesis can be framed, such as ‘male workers with higher seniority in this firm tend to have a higher density of incoming connections.’  A change program which aims to increase diversity and involvement amongst younger workers in a particular discipline (i.e. statistical analysis), could then use an understanding of the as-is network to implement a rotation program, incentive system, and/or networking body for a target demographic they wish to work into the organization.

As well, such analysis allows for more sensitivity and care to be taken in reorganizations.  With such insight, it becomes quickly apparent that, for instance, quiet, kindly, and unassuming Lois, the legal secretary, who has been with the organization for 30 years, is a ‘hub’ in conveying key organizational process information throughout the organization, often by passing knowledge through several cohorts.  There are unassuming but crucial ‘Lois’ workers in every organization.

Often, the ‘Lois’ worker is the last person to be picked for the ‘guiding coalition’ in a change management initiative and they may be even at risk for being ‘downsized’ due to age and/or ‘unclear value’ (i.e. lack of visibility at higher management levels).  By second-guessing the biases and interests of the guiding coalition and actually conducting such hands-on quantitative research into the organizational network, change initiatives can avoid the costly mistake of, after giving Lois a cardboard box on Friday, finding on Monday morning that she was running the company from below.

Any type of relational-exchange network can be analyzed within an organization: information flows, respect, reporting lines, cross-functional collaboration, decision making, etc.  The benefit of the SNA approach is that it is an exhaustive method: all agents report the basis of their exchange individually.  The network structure emerges from the aggregation of the network ties. Both formal and informal patterns emerge.  Insight is gained into the existence of cliques, or sub-groups which may be somewhat isolated or disconnected from a process.  Insights from SNA map visualizations often are surprising and lead to rapid insights.  It becomes clear who is not talking to whom, who is hoarding information, who is a bottleneck, etc.

SNA information is quick and relatively painless to gather: a survey is conducted asking each individual for a list of names regarding their collaborations (the object of exchange being targeted). ‘Snowball’ surveys can be conducted in larger organizations, whereby people are asked who their ‘go to’ relations are, those people are surveyed, and the subsequent layers are surveyed in turn.  Pictures of the interconnected network emerge.

Also, data can be gathered from passive sources, especially where electronic records are available (and access is permitted): email exchanges, phone connections, meeting invitations, text messages, etc.  When permitted, electronic records of exchanges can be used to generate rapid, detailed maps concerning who is talking to whom (or not talking too whom) in the organizational network.

Beyond the visual maps, there are more focused quantitative measures which can be extrapolated from the network.  Based on a long sociological research tradition, a number of key SNA measures have been developed which can be determined from network data.

• Centrality: which agents are particularly powerful in terms of being highly connected in the network? Research has shown that power and control resides in being highly connected within a network.  Several particular measures of centrality exist:
  • Degree: this is a simple measure of the number of incoming connections an individual has. For example, a politician or mafia don would likely have high centrality degree scores in their various networks.  This is an indication that many others in the network are connected to the person.  Powerful individuals may have many ‘incoming connections’ (people who consider themselves tied to them), but fewer outgoing connections (their own connections to others which they reciprocate).  A reciprocal connection is considered stronger.
  • Closeness: this is a measure of one’s closeness to all others in the network.  A strong closeness measure indicates that the agent can connect to all members of the network in a minimal number of steps. Thus, this person may not have the greatest number of connections, but is able to quickly reach others in the network through a minimal number of steps. It would be desirable as a stockbroker, for instance, to have such centrality as it would allow them to quickly gather information or to promote a stock via-via in an investor network.
  • Eigen vector: (aka Bonacich’s power) this measures one’s connection to those who are highly connected. Whereas a politician has a large number of incoming connections, a political advisor who councils many politicians would have a high Eigen vector score.  Indeed, research attests that this is a powerful and desirable measure: many types of natural and social benefits accrue to those connected to those who are connected, including physical health and happiness measures. Whereas those with many direct connections bear many risks and pressures, a high Eigen vector agent can be thought of as the person playing a key role ‘behind the scenes’.  Thus, whereas it may be very risky to be a mafia don (i.e. risk of arrest or being ‘knocked-off’), a person who advises many mafia dons accrues great power and reach in a network while presumably avoiding direct risk.
  • Betweeness: this is a measure of the degree to which an agent stands in a central brokerage position across key paths or vectors connecting the network.  A person playing a gatekeeper or intermediary role would have high betweeness. This would be advantageous in a trading network as the agent can benefit from being a key intermediary in transactions across the network.  While being a ‘bridge’ is often a crucial function in highly distributed network, there is a danger in terms of having this being a bottleneck and point of low redundancy.  In a decision making network, a person being the only bridge can be considered a risk in terms of decision quality (the upshot being that more bridges should be put in place to improve the flow of information).

A number of standard quantitative measures of network characteristics can be extrapolated from aggregate connection data:

• Network density: how interconnected is the organization on a particular measure? This is a measure of the number of connections in the network compared to the maximum number of connections possible. A highly dense network may be difficult to change as there are very tight connections amongst the agents. Very sparse networks may indicate a level of inefficiency as agents must go through several steps via-via to pass information.  Further, directions and decisions may be slow or even lost in the effort to span a low-density network.
• Reachability or distance: this indicates the minimum number of steps to traverse the network. Popularized via the concept of ‘six degrees of separation’, this is a measure of the number of average steps it takes to get from one part of the network to another. A company with many management levels and a lack of informal social ties amongst workers may suffer from inefficiency and inaccuracy in conveying information.
• Clustering coefficient: (aka cliques) this indicates the degree to which two agents connected to a third agent are connected to each other.  High clustering indicates that there are tightly bound cliques of interconnected individuals in the network.
• Cohesion: this is a measure of the tightness with which agents are interconnected. A cohesion measure indicates the ease with which a group can be broken-up by removing members (otherwise of interest in change programs when disconnecting a group of change resistors is desired).
• Cores: how many significant sub-networks exist within the network?  This measure can indicate when there are tight tribes or competing / non-cooperating cliques.
• Largest core:  what sub-group is the largest and most powerful in the organization?  This is often useful in determining where core power resides in the organization. It may be, for instance, that the marketing group network is strong and large, indicating the marketing function guides the rest of the organization.
• Sub-structures: similarly there may be particular types of sub structures within the network – isolates (loners), pendants (groups thinly connected to the organization), or bridges (gatekeepers or mediators between two sub-groups, as discussed earlier).
• Structural holes: this indicates a lack of connection between parts of a network.  This may be due to competitiveness or silos, but the end result is a lack of exchange between parts of the network to the degree that there is potential weakness and waste in the network (when high connectionism is advantageous or otherwise desirable for efficiency or accurate information passage).
• Co-membership:  what is the level of co-membership in multiple sub-groups within the network? Low co-membership may indicate many isolated or ‘siloed’ subgroups.
• Connections: what characterizes the nature of the dyadic ties within the network?
  • Tie strength: this is an aggregate measure of intensity, reciprocity, and intimacy in connections between agents. A manager that has reporting authority but also respect and admiration would have high tie strength to direct reports.
  • Homophily: this is the degree to which agents display a preference for connecting with others based on particular attributes such as gender, race, age, job role, status, values or any other particular shared attribute of importance.
  • Multiplexity: this indicates the aggregate number of connections between agents based on different criteria.  For instance, two agents may be friends and work collaborators, indicating a strong relationship.
  • Mutuality/Reciprocity: this indicates if two agents reciprocate a tie. For instance, an agent may consider another a friend, but the other may not share that assessment.
  • Propinquity: this indicates a tendency for actors to have more ties with geographically proximate members

By clarifying the types of quantitative measures that emerge from a SNA program, it becomes clear how a change program can benefit.  The variables which emerge from SNA each have implications concerning particular goals for a change initiative.  For a change initiative focused on adopting analytics culture, those measures which seek to strengthen decision making effectiveness would be of particular interest.

For instance, an organization could be measured in terms of its financial planning and analysis (FP&A) network.  What would emerge would be a map and quantitative measures representing the degree to which the FP&A process is cleanly connected (in terms of agent communication).  It would also become apparent where certain cliques or sub-groups consolidated.  It might become apparent, for instance, that there was a lack of connection between functional business specialists, corporate finance planners, and IT people managing the FP&A systems.  Such an organization could benefit by putting in place FP&A functional sub-teams which would encourage tighter network relations between the various FP&A decision actors.

Examples abound and are constrained only by the goals and scope of corporate change objectives.  In the next and final article, we will consider more specifically how the above measures can be used to plan and adopt analytics cultural change.

END OF PART 6 OF 7 IN A SERIES ON ADOPTING ANALYTICS CULTURE

Link to previous article in series: 5. How can change management be improved via analytics?

LINK TO HEADER ARTICLE

REFERENCES

Burnes, B., & James, H. (1995). Culture, cognitive dissonance and the management of change. International Journal of Operations & Production Management. Vol 15, No 8, 1995.

Burton, R. M., Obel, B., & DeSanctis, G. (2011). Organizational Design: A Step-by-Step Approach (Second ed.): Cambridge University Press.

Cronin, B. (2011). A window on emergent European social network analysis. Procedia Social and Behavioral Sciences, 10, 4.

Cross, R., Liedtka, J., & Weiss, L. (2005). A practical guide to social networks. Harvard Business Review, 83(3), 8.

Cross, R., & Parker, A. (2004). The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations. Boston: Harvard Business School Press.

Huisman, M. (2012). Software for social network analysis  Retrieved August 6, 2012, 2012, from http://www.gmw.rug.nl/~huisman/sna/software.html

Huisman, M., & van Duijn, M. A. J. (2005). Software for Social Network Analysis. In P. J. Carrington, J. Scott & S. Wasserman (Eds.), Models and Methods in Social Network Analysis (pp. 270 – 316). Cambridge: Cambridge University Press.

Kameda, T., Ohtsubo, Y., & Takezawa, M. (1997). Centrality in Sociocognitive Networks and Social Influence: An Illustration in a Group Decision-Making Context. Journal of Personality and Social Psychology, 73(2), 14.

Kilduff, M., & Tsai, W. (2003). Social Networks and Organizations. London: SAGE Publications Ltd.

Knoke, D., Yang, S. (2008). Social Network Analysis. London: SAGE Publications, Inc.

Krebs, V. (2012). Software for Social Network Analysis & Organizational Network Analysis  Retrieved August 23, 2012, 2012, from http://orgnet.com/inflow3.html

Popov, V. (2003). Social Network Analysis in Decision Making: A Literature Review (W. Time, Trans.): PSIRU University of Greenwich.

Prell, C. (2012). Social Network Analysis:  History, Theory & Methodology. London: SAGE Publications Inc.

Prietula, M. J., Carley, K. M., & Gasser, L. (1998). Simulating Organizations. Cambridge, Massachusetts: MIT Press.

Scott, J. (1991). Social Network Analysis. London: Sage.

Seely Brown, J., & Duguid, P. (2000). The Social Life of Information. Boston: Harvard Business School Press.

Tsvetovat, M., & Kouznetsov, A. (2011). Social Network Analysis for Startups: Finding connections on the social web. Cambridge: O’Reilly.

Tushman, M. L., & Fombrun, C. (1979). Social Network Analysis for Organizations. Academy of Management Review, 4(4), 12.