DEMESTIFYING ANALYTICS

Analytics everywhere!

Often people ask (most of the time clients) what is this analytics thing? I can understand the concern. They are (most of the time client's senior executives) used to terms like Data Warehousing , Reporting for most of their working life. Suddenly, there is analytics everywhere….in the news papers, in the television and even in the airport bill boards!! Companies claims all sorts of benefits using ‘cutting edge’ analytics..right from fraud detection to reduction of global warming. That’s confusing!

How it's different from reporting?

I would say reporting is 'analysis' of past data. 'Analysis', as most of the reporting tool provides you slicing, dicing or looking the data by different dimension. Whereas analytics is about predicting things that are yet to come. Reporting is about looking at 'post facto'. While Analytics is about deciphering pattern from historical data and knowing about the future.

Analytics can broadly be divided into 2 types:

A. Predictive modeling: Applied on cross-sectional data (say, transactional data of a bank’s saving account holders at a particular point of time). Most likely some of the saving card holders also have fixed deposit (falls in the common area B). You might be interested to predict which of the existing savings deposit holders are most likely to buy FD from the bank. In this up-sell situation, the key word is ‘most likely’. Analytics reads the pattern of the B area and score the customers in S area in terms of likelihood of buying fixed deposit.

B. Time Series Modeling: Applied on time stamped data this technique deciphers the pattern of the past in different ways. There are several techniques which are as simple as smoothing, averaging etc. Classically however time series analysis breaks the time series in components like long term trend, within year seasonality, cycle and the irregularity. For example, you might be interested in understanding pattern of past sales and want to forecast the future sales. In time series modeling, the sales can be decomposed into components like trend (population growth), seasonal (climate, festivals etc), cycle (GDP growth) and the irregular component.

Besides, analytics also encompasses techniques like basics statistical techniques (ex. Hypothesis Testing, Exploratory technique (ex. Market Basket, Clustering), Text mining, Optimization and many more.

End Note:

In the US, analytics space has emerged in the 1970s with the software coming up for mostly from the university campuses. Supported by emergence of high processing computing technology and user friendly codes (most of the analytics tool runs on 4GL) analytics caught up advanced economies in the late 20th century. Note that analytics necessarily requires data and clean data. The wave of data warehousing overwhelmed Indian business in the early years of this decade. The fast realization happening that we are sitting on the huge data and there is huge potential of enchasing it. This explains why analytics is the buzzword, the starting point of this blog!!