Big data can originate in many industries. Big data is also created by social networks. Collecting big data is relatively easy, and storing it is relatively inexpensive, especially in the cloud. The cloud also offers huge computing resources that can be used to analyze the big data.
Social networks are communications networks. The enterprise, with a much smaller number of internal social network users than the public social networks like Facebook, is still generating big data, and enterprises should therefore consider what to collect and how to use the collected data to improve their internal operations.
There are three parts to the big data environment:
* The individuals that are members of the social network deposit data into some repository. This is deliberately or automatically accomplished.
* The repository, which is most likely part of the enterprise, either on premises or in the cloud, is where the big data analysis occurs.
* The big data analysis produces intelligence that is fed back to the enterprise and/or individuals.
The platforms that an enterprise should adopt are scale-out systems. Scale systems include commodity servers, an architecture that does not share resources such as memory and storage, and an IP network connection. Amazon's Elastic MapReduce is an example of this environment. A pay-per-use charge is well suited to big data analysis.
Speed vs. Accuracy
The traditional enterprise approach is to process the transaction. The resultant data is then passed to a system which extracts the relevant information. usually in a batch mode. This causes delay in gaining the results. The data is then processed to produce intelligence and insights into the operation and behavior of the individual network members and the enterprise. This method produces data accuracy but delays the delivery of the intelligence and insights. This methodology fits well within the way enterprises currently carry on their business operations.
There are exceptions to this methodology. When financial fraud occurs or an enterprise wants to react to changing market conditions, delayed intelligence cannot be tolerated. The batch method cannot deliver just-in-time responses to queries. The real-time queries can produce rapid results but are not necessarily delivering accurate results, because not all of the relevant data may be analyzed, because some of the data is still in the batch processing system. The enterprise will probably have to add data collection points in their business processes to obtain the big data sooner, so that more accurate and timely intelligence can be generated. This means less batch processing and more real-time processing.
Social Network Interactions
User interaction is a major factor in the success of a social network. This is tempered by trust, the trust that exists between individuals but not with strangers. The more closed the social community, the more likely there will be trust among the users. This benefits those who want to analyze the big data generated on the enterprise social network, because it is likely that the enterprise users may create more relationships and interact to a greater extent than they would on a public social network.
The cloud service that is most attractive for processing big data needs to provide large inexpensive storage. The cloud provider should focus on scalability and elasticity on commodity storage and servers. This type of platform is attractive to operations that are relatively simple, while not introducing latency, especially when the scale of analysis becomes large.
The hardware available in the cloud appears adequate to handle big data analysis. The limiting factor is the software that performs the analysis. I think that we will see two approaches: Software used within the enterprise's on premises systems; and cloud-based services. I anticipate that the cloud approach will be more attractive to both small and large enterprises.
Big data analysis will be new to most enterprises; they will have little or no experience. Big data analysis service is a perfect candidate for cloud residence.
What Can Enterprise Social Networks Tell Us?
Telecom professionals have performed data analysis on a small scale when they process and review telephone Call Detail Recordings (CDR). They looked for traffic patterns, calling abuse, verifying bills, and for planning network capacity. Big data analysis can go much further. I would collect information from the social network system, Unified Communications system, and even human resources. There are benefits for combining all of this information:
* Discovering communications patterns and relating those patterns to the media used, the frequency of its use, who uses the communications, and relating all this information to the profile of the individuals participating in the communications.
* By analyzing the social connections, the enterprise will probably discover that the existing organization structure is partially bypassed so that the individuals can work with others that are not part of their group or department. The enterprise may be able use the connection information to reorganize their internal organization structure to reflect the social network connections and thereby improve enterprise productivity.
* Analyzing the big data can help the enterprise discover the most effective means of communicating, and how these may be changing over time.
* Determining how relationships have been created and how they affect productivity.
* Learning how individuals discover who to add to their social network.
* Determining what forms of social media are the most popular, and ensuring that those media have the best support.
* Creating programs that teach others how to use social media for their benefit at work.
* Monitoring for social media abuse.
* Determining how social media may have changed the costs of doing business.
* Has social media helped to accelerate project completion and success?
These are just a few benefit examples. I am sure that creative people will create many more uses for analyzing social network big data.