In this era of Big Data, massive volume of data is produced at an incredible rate and in various formats. The data volume of most modern businesses is expanding by 30% to 50% annually. An modern institution nowadays handle more than 60 terabytes of data annually, which is about 1,000 times more than a decade ago (Beath, Irma, Ross, & Short, 2012). Exponential growth of data from social media, sensory devices, mobile applications, and personalized information has flood the institutions with myriad amount of data. This data that is highly complex, heterogeneous, uncertain, and dynamic requires institutions to have new breed of analytics tools to reap the benefits of Big Data.
Analytics is a multi-dimensional discipline that arises to meet the needs of greater analytical capability. Analytics comprises a set of data-driven processes, technologies, and concepts which pertinent to gaining insight from data. The outcome of Analytics is actionable insights that can inform decision and devise action (Stubbs, 2011). Figure 1 illustrates analytics from a process perspective which includes three major phases, namely manage, integrate, and analyze data.
What really matters in the end is not how much data can be collected, it is about what users can do with the data that counts. An analysis gap exists given that the capacity to collect and store data has been grows rapidly, but the capability to analyze these data increases at much lower pace. This study focuses on the “analyze” phase (orange-colored oval) where the end-users interact with Analytics applications to glean actionable insights from the data. These applications can be categorized into a) visual analytics, and b) advanced analytics, according to their approach to data analysis. They serve different purposes, deal with different data types, and require different levels of skill to operate. Table 1 provides a comparison view of the three categories of analytics applications.
|Visual Analytics||Advanced Analytics|
|Purpose||§ To gain deeper, qualitative insight of the data||§ To extract hidden knowledge from massive data|
|Targeted Problem||§ Operational
|Input Data Structure||§ Structured to Semi-Structured Data||§ Structured to Non-structured Data|
|Input Data Volume||§ Low to Medium||§ High to Enormous|
|Main Approach||§ Human-driven Sensemaking and Reasoning||§ Machine-driven computations and modeling|
|Techniques||§ Interactive visualization
§ Supporting human cognition
|§ Statistical and mathematical modeling
§ Artificial intelligence and machine learning
|Analysis Result||§ Descriptive
|User Friendliness||§ High (Easy)||§ Low (Difficult)|
|Processing Capability||§ Low||§ High|
|Common User||§ End Users
§ Power Users
|§ Power Users
§ Data Analysts / Data Scientist
§ IT Specialists
|Example of Applications||§ SAS Visual Analytics
§ SAP Visual Intelligence
§ IBM Cognos Insight
§ Tableau (Stanford University)
|§ IBM SPSS Modeler
§ SAS Enterprise Miner
§ R Project
Note that there is a tradeoff between the user-friendliness and the processing capability of the Analytics applications. The applications in the advanced analytics category such as data mining and statistical analysis are known for their powerful processing capability to create knowledge from massive and complex data (Fayyad, Piatetsky-Shapiro, & Smyth, 1996). However, It is widely agreed that they are too complex for wide-spread use because highly specialized knowledge and skills are required to operate them (Nemati, Earle, Arekapudi, & Mamani, 2010). Studies shows that less than 17% of institutions have been actively using advanced analytics for supporting their operations. Whilst visual analytics applications are relatively easier to use, their processing capability is limited to human-driven analysis. Visual Analytics relies on manual knowledge extraction, generally more time consuming, and highly depends on the user’s domain knowledge to derive valuable knowledge.
Visual analytics (VA) application is an emerging breed of Analytics applications which developed to address Big Data problem. Visual analytics applications capitalize on interactive visualization to facilitate human analysts for effective understanding, reasoning, and decision making on the basis of very large and complex data sets (Daniel Keim, 2012). Visual analytics strives to empower non-technical users with considerable level of analytical capability via user-friendly interface (Eckerson, 2009). The self-service analytics enabled by visual analytics extends the scope of analytics beyond scope of dedicated data scientists and business analysts to wider organizational users who have traditionally relied on others for this analytics capability. This idea of pervasive analytics give greater flexibility to users to refine their information seeking strategy on the fly and allow them to react more quickly to their problem and eventually shorten the time to value. This is in conjunction with the ideal situation where the people who need the insight should be the same people who analyze the data given that they possess the domain knowledge which allow them to harness the true value of the information (Mirel, 2004). This is especially important for complex and strategic problem in which human discretion and judgmental heuristic are critical for deriving the solutions. Despite the ideal visions advocated by Visual Analytics, most Visual Analytics applications in existence are nothing more than data visualization tools repackaged under the name of visual analytics as the result of marketing strategy.
Advanced analytics (AA) applications are a group of machine-driven analysis tools which relies mathematics, statistics, and artificial intelligence techniques to automate knowledge discovery (Surma, 2011). Advanced analytics applications can be broken down four types according to their purpose, namely data mining, predictive analytics, simulation, and optimization. With the computational techniques, advanced analytics allows analyses that go beyond human capability. Advanced analytics can handle enormously huge set of complex data, exhaustively and consistently execute the analysis, and provide quantitative and statistically rigor results. Such capability is especially important for Big Data problem that exceeds the human’s attention perception, and cognition ability, even with the computer-aided scaffolding supports. However, the knowledge discovered by advanced analytics are mostly passive knowledge. This knowledge is often highly technical and with little domain context (Cao, 2012). The practical value of such passive knowledge is usually trivial as it tells users very little about how to act upon it.