5 Steps To Consider For Analytics Implementation
Big Data. It is here. With oodles of information on the customer. It is up to smart, forward thinking Chief Information Officers (CIOs) to step up and see how the enormous amounts of data generated by the industry, the consumer, and even society, can be used to find information that will help their organizations achieve their business goals — how to use analytics to generate insightful inputs?
Analytics has progressed technologically, keeping pace with the exponential growth of data. There are now a wide array of analytical tools that can be used on both structured and unstructured data. Choosing the right analytical tools and using them right requires information. Information that ranges from the basic need for the analysis to the kind of data we are looking for, the probable sources of that data, methods of data collection, storing, sorting, sifting… even modeling and review of action taken on the basis of the analysis.
Effective implementation of analytics therefore, requires the adherance to a process. Here’s how we suggest going about it:
The future of business depends upon the ability to access and use the 360 degree views of customers that are now available. It is up to the CIOs to empower the management by taking the lead in analyzing all that information.
- Setting down the goals for analytics:
This is the traditional way of beginning an analysis, with a fixed goal on scope. You define the gaps to target that the data will help to fill and proceed accordingly. This works well for most assignments. However, with the advent of Big Data, there have been many instances where the sheer volume and variety of the data has thrown up patterns/trends that have gone on to impact strategy and thinking outside the traditional analytic mode.
- Collection of data:
Data aquisition from multiple platforms and multiple applications comes next. Is the data internal, external or a mix of both? Structured or unstructured? This is where Big Data comes into the picture with its distinguishing features of sheer volume, variety, and velocity.
- Data Storage:
Yes, we have a multitude of sources for data, continuously pouring in information that is invaluable as customer insights… how do we store it? This depends on various factors including cost; the type of data — structured data like sales figures etc., can be stored in Relational Database Management Systems (RDBMS) and data modeled by means other than the tabular relations used in relational databases can be stored and retrieved from NoSQL databases. Another factor is storage capacity —there are quite a few options on the market now, like Cloudera and Hortonworks (from Hadoop) or NoSQL databases like 10Gen.
Next is the transformation of data into analyzable chunks - placing the data in context and connecting it with other related data to draw learnings. This is pivotal. Analyzing big data and creating valuable insights from it relies heavily on what context the data is seen in. With context comes enhanced individuality of the consumer and his or her behavior. Pig and Hive are platforms used for querying and analyzing large data sets, of both structured and unstructred data.
- Predictive Analysis:
What follows is the inference. The prediction and recommendation of solutions based on presented results. This is the ultimate role of analysis — to extract information from data and make predictions for trends and behavior. This also offers a future rather than historical perspective of the customer. By arriving at data backed conclusions and coming to decisions based on them, management is empowered with realtime tools and actionable insights. Text analysis, statistical data analysis, predictive data modeling and graph engines are some of the techniques used here. SAS, R, Gremlin, Cascading are some of the tools and languages which enables us to build such models or applications.