By Subrata Rath & Abhinav

All corporate, media, governments, establishments are now engaged to find tremendous sense out of the data gathered from various sources about their respective organisations. Fact-based decision making support system as of date have eventually turned up to gather so much data- called big data from different sources at different time points and possibly burdened with non-numeric nature of data. Classical data analysis is almost found to be handicapped to make any headway to extract information from such data…

Precisely this poses the challenge of Data Analytic at its present state of maturity. Unlike Six Sigma, Data Analytic approach is not having a universally acceptable standard approach. This makes the use of data analytic further complicated. Common sense to handle data analytic possible demands the following:

– Data have to be prepared first before to deal with analysis. Undoubtedly this is the most important steps of Data Analytic. Purpose, nature of information, decision making processes, usability of the information coming out of the data analytic process and the likes must be crystal clear at this stage. Selection of target variables and linking the same to the business problem and selection explanatory variables are critical at this stage. This is the first stage-

1. Data Preparation.

– Basic analysis for each variables selected in the data collection plan can be made through visual representation of the data and its corresponding probability distribution, probability based prediction, testing certain important hypotheses envisaged, estimation of variability of parameters of interest in decision making and the likes. Careful and intelligent handling of such data analysis can itself give rise to fantastic information to support decision through parallel adoption of two steps:

2. Data Visualisation and
3. Exploratory Data Analysis

– Most important use of thus collected data perhaps lies in establishing a credible and confident relationship to prescribe explanatory variables to predict target variables. Such prescribed model works in almost all the cases e.g. parametric or non-parametric and classical and Bayesian statistics etc. However, at this stage of Prescriptive Analytics, models may often suffer from more misclassification error or prediction error. Discriminant functions, Regressions, Tree-based algorithms, Naive Bayes classifier, ANN, SVM are few noteworthy models at this stage of

4. Classification

– One of the reasons for high misclassification error in the model is inadequate contribution of the explanatory variables. Search process goes through the intricacies of the explanatory variable in search for better explanatory variables through careful study of the explanatory variables without their linkage to the target variables. This stage of Descriptive Analytics may often offer better explanatory variables through:

5. Clustering and
6. Market Basket Analysis

– Inclusion of the new supervisor may give adequate strength to predict with newly formed model. Time series dimension must be brought at this stage to serve the real purpose of prediction with better accuracy. This stage of Predictive Analytics is no different from the classification with the augmentation of Time series analysis but to firm up the model to predict at the stage of

7. Predictive Analytics

These seven stage analytic process may penetrate to almost all functional domains which include

Functional Domain Analytics Application (Project)
Customer Analytics Predicting Customer Churn
Marketing and Sales Understand Demand to Order Cycle
Supply Chain Total Distribution Cost Assessment and Understanding
Operations Optimisation of Operational Processes – Cycle Time, Defects and Product Mix
HR Understand Attrition Rate
Social Media Company Image and Perception Assessment through Social Media
Retail Realignment of Item Stocking and its Repositioning to increase Sales;
Pricing Strategy
Banking and Finance Understanding Fraud and Bad Debts
Legal Assessment of Potential non-compliances
Energy Energy Mix and Cost Optimisation
Balance Sheet Overall Business Level Health Assessment

And what not?