What is Analytics?

Clearing the clutter and finding out what it really means

The term “Analytics” has been in use for decades. However, we have been seeing huge influx newer terminologies like Business Intelligence (BI), Data Analytics, Predictive Analytics, Data Science, Machine Learning, Artificial Intelligence, Cognitive Science and so on..

Here is an attempt at clearing the clutter and understanding what it really means.

Analytics is an overarching term that refers to the ability to derive intelligence to make informed decisions at every possible level. Enterprises across industries and geographies are embracing analytics and trying to leverage it to their advantage.

As enterprises function and carry out day-to-day operations, they generate really significant amount of data. This data represents the historical activities and how the organization functioned as an entity. For example, a credit card issuer sells credit cards directly to end consumers. As they do, they accumulate a lot of information — list of customers they have acquired, their profile, type of cards, customer tenure, loyalty, occurrence of default, etc. They capture the entire credit cardholder lifecycle, from marketing to origination,

In their attempt to derive intelligence from data, organizations typically go through the following four phases.

• Descriptive analytics: Know “What happened?”
• Diagnostic analytics: Understand “Why did it happen?”
• Predictive analytics: Predict “What will happen?”
• Prescriptive analytics: Recommend “How can we improve?”

Descriptive analytics

It usually refers to examination of past data to answer the question — What happened? Lots of visualizations, tables, dashboards form the underlying blocks.

For instance, credit card issuer would try to measure and quantify their historical performance, answer questions such as “What is the average rate of delinquency overall?”, “Is the number of new customer acquisitions increasing over time?”, “Which state/region has the highest customer churn?” etc.

Toolkit: Business Intelligence tools such as Microsoft Power BI, Tableau, Qlik Sense, SAP Analytics Cloud, Oracle Analytics Cloud, Looker, SiSense, ThoughtSpot, etc.

Diagnostic analytics

Understanding of “what happened” mostly leads to follow up questions form business users focusing on “why did it happen?” Some of the analytic techniques that help organizations to identify the root causes are drill-down, slicing-and-dicing, data discovery, or what is generally known as exploratory data analysis (EDA). Reasoning and root cause analysis are the key elements here. Insight generation and insight explanation help companies to augment how people explore and analyze data.

The credit card issuer would ask why the rate of delinquency very high in a certain region, or why customer churn was very high last month, etc. The main objective is to reason why something happened in a certain way.

Toolkit: Business Intelligence tools such as Microsoft Power BI, Tableau, etc. are increasingly adding more capabilities on this (i.e. Augmented Analytics — augmenting beyond traditional expectations from BI platforms). ML tools like Python and have been well-proven capabilities to tackle these.

Predictive analytics

This refers to the next phase where companies analyze past data, extract patterns/trends, and try to predict “what is going to happen.” It involves building predictive model to determine what is most likely to happen by using techniques such as multivariate statistics, regression, forecasting, machine learning.

If we go back the example of a credit card issuer, they would want to understand these: i) what is the risk of a prospective customer? ii) is this a fraudulent transaction? iii) will this customer stay or churn?

Toolkit: ML tools like Python, R, SAS, etc. BI tools and business applications (ERP, CRM, etc.) are increasing trying to embed predictive capabilities into their workflows.

Prescriptive analytics

It is the form of analytics that helps to determine what should be done. It extends beyond prediction and often predicted output is used to get to recommended set of actions for end users. In addition to ML, this also involves heuristics, optimization, recommendation engine, etc.

The credit card issuer wants to know what sort of action would result in payment of an overdue balance, on what terms should we accept an application, etc.

Toolkit: ML tools like Python, R, SAS, etc.

Analytics — What do you do to derive intelligence from data?

Enterprises are increasing embracing analytics and are at different maturity levels. When we meet and talk to analytics professional (be it BI developer, Data analyst, ML engineer, Data scientist, or Business analyst), they are doing something to help organizations to derive intelligence from data.

Source: Competitive Advantage vs Analytics Maturity by Gartner

The core objective of Analytics (or any other name you opt for) is “to derive intelligence from data” and it has become critical for organizations to leverage analytics to build competitive advantage.