Social network analysis

There are a variety of possibilities for use cases in the deep analysis category. Since IBM has implemented an enterprise social network ...

There are a variety of possibilities for use cases in the deep analysis category.

Since IBM has implemented an enterprise social network analysis to facilitate communication and collaboration with and among employees, the next logical question becomes: What new insights can we derive through an analysis of the conversations? A few possible analytics projects have been proposed, but to date, we haven’t implemented them due to time constraints. We thought it might be of interest to see what kinds of projects would be possible given this rich source of data:

• What is the correlation between the level of social activity and the likelihood of increased innovation?
• What is the correlation between the level of social activity and the likelihood of getting selected for a customer advocate role?

Some of the high-level steps in implementing an internal social analytics practice include:

1. Establish an enterprise social data analytics network inside the corporate firewall. If the intent is to use an external platform, take care to ensure employees understand the risks of inadvertently releasing sensitive information to the public. But if an external platform is chosen, keeping a record of employee identifiers in that platform is critical to enable the retrieval of their information and ensure the analysis contains information from just the employees and not those outside the company. 
2. Ensure a widespread adoption of the capabilities by employees.
3. Establish a metrics program that measures people’s participation in social activities.
4. Map these activities to behaviors and identify key performance indicators (KPIs).
5. Establish algorithms to compute the scores for each key performance indicator.
6. Establish a baseline.
7. Establish a metric for measuring innovation—for example, the number of patents.
8. Establish a metric for the customer advocate role—for example, names of people who got selected for a customer advocate role.
9. Establish a window of time for the study and social media data analysis.
10. Measure the change in KPI values during the window of time.
11. Apply regression analysis and draw conclusions.

Machine capacity

The network bandwidth required is usually low, but the CPU capacity required for this type of analysis is usually high. Table 6.3 shows use cases for internal social media.

Velocity of Data

We broadly divided this into two categories: data at rest and data in motion. In the following sections, we look at the dimension of time from the perspective of analysis. The dimension of velocity of data also can be divided into two parts: data in motion and data at rest. 

Data in Motion

As an example of data in motion, during a US Open tennis match between two players, we might want to understand how the sentiment of the general population is changing about the two players during the course of match. Is the crowd conveying positive sentiment about the player who is actually losing the game? In such cases, the analysis is done as the data arrives. Our assumption, as shown in Figure 6.4, is that for a constant time interval, the amount of detail produced increases as the complexity of the analytical tool or system increases.

Data at Rest

A second type of analysis in the context of velocity is what we call “analysis of data at rest.” For example, we can collect social media conversations around IBM products and services before, during, and after a specific event to understand the public’s opinion. Once the data is fully collected, we can then perform analysis on this data to provide different types of insights; some examples follow:

• Which of your company’s products has the most mentions as compared to others?
• What is the relative sentiment around your products as compared to a competitor’s product?
• Is there a strong correlation between the marketing of a product or service and the number of positive comments about the brand itself?

In these two cases (data at rest and data in motion), there are trade-offs that we need to consider time and the cost-to-deliver those results. For example, if the results of such an analysis are needed in real time or near real time, the amount of time available for processing is, of course, limited and that will have an influence on how deep we can go with the analysis.

We have observed that valuable insights can be derived quickly by providing some lightweight analytics in these real-time use cases. Simple metrics such as popular hash tags, most prolific authors, top mentions, and so on can provide some revealing insights without a large investment in computing (and analysts’) time. This usually translates into lower infrastructure costs because we are processing relatively small amounts of data in a small chunk of time. When we are performing analysis of data at rest, the amount of data available for analysis (the volume) has a strong bearing on the time needed for the analysis to complete and hence the costs of that analysis. The amount of data could range from the one day to several weeks, months, or years. In such cases, we are usually interested in topics in aggregate, such as “How has the sentiment about a company’s brand changed over the given time period as compared to its competitors?” rather than a very specific topic or question.

The business benefits gained from a social analytics project aren’t always directly proportional to the cost. We have observed that, even though a lightweight analytics solution may cost much less than a deep analytics solution, the business benefit is still highly dependent on the specifics of the project that we are working on. For example, if a company is monitoring social media for any negative press about itself, its products, or its executives, and it detects a sudden surge in negative sentiment, the public relations department can be prepared with an appropriate response in a very short time. As any good public relations staff will tell you: “Forewarned is forearmed.”