Analytical Concepts: Must Known For Every Professional & Fresher

Introduction

The use of analytical concept have gained immediate importance in the last few years. The practice of gaining useful insights from data have helped several companies improve their business performance.

Analytics allow companies to obtain a clear picture of events in the past and the future of their performance. A peek into the future helps companies to get prepared for the misfortune (if any) about to arrive.

Using analytics, companies can find answers to three main questions: ‘What has happened’, ‘What is happening’ and ‘What will happen’. It wouldn’t be incorrect to say that the rise in data has fueled this outrageous penetration of analytics use.

Analytics is not just limited to deriving insights from the past, but also enables predicting future outcomes and optimizing business resources. As a result, the more advanced forms of analytics namely, predictive and prescriptive have assumed greater importance in supporting organizations’ decision making needs.

In this article, I’ve explained the 3 major forms of analytics which categorizes all forms of analytical models applied across countries.

According to a study, organizations that focus on basic automation to expand their reporting capabilities can improve their ROI by 188 percent. But, adding advanced analytic deployments that enhance organization strategy can extend their ROI by as much as 1209 percent.

What are the different types of analytics?

1. Descriptive Analytics

Let’s start with the most basic type of analytics i.e. descriptive analytics. The objective of descriptive models is to analyze historical trends and figure out relevant patterns to gain insights on the population behavior. Descriptive analytics involve finding answers to ‘what has happened?’. It is the most commonly used form of analytics by organizations for their day to day functioning and is generally the least complex.

Descriptive models use basic statistical and mathematical techniques to derive key performance indicators that highlight historical trends. The primary purpose of the model is not to estimate a value, but gain insight on the underlying behavior. Common tools used for running descriptive analysis include MS Excel, SPSS, and STATA.

A typical example in the Banking industry would be customer segmentation. Historical data is mined to analyze customers’ spending patterns and wallet share to enable a targeted approach to marketing and sales. Such models are powerful tools to profile the population, but are limited in their predictive ability with respect to the individual members’ behavior from the same population.

Helpful resources:

• Online resources for learning basic descriptive statistics can be found on Khan academy: Link
• Here’s a video on running descriptive statistics in SPSS: Link
• A must-do MOOC on Coursera- Data Scientist’s Toolkit: Link

2. Predictive Analytics

Predictive analytics use statistical modeling to determine the probability of a future outcome or a situation occurring. It involves finding answers to ‘What could happen?’.

Predictive models build up on descriptive models as they move beyond using historical data as principal basis for decision making, often using structured and unstructured data from various sources. They enable decision makers to make informed decisions by providing a compressive account of an event’s likelihood to occur in the future. They encompass various advanced statistical models and sophisticated math concepts like random forests, GBM, SVM, GLM, game theory, etc.

A predictive model builds on a descriptive model to predict the future behavior. However, unlike a descriptive model that only profiles the population, a predictive model focuses on predicting a single customer behavior.

Tools used to run predictive models vary by the nature of the model’s complexity, however some of the commonly used tools are RapidMiner, R, Python, SAS, Matlab, Dataiku DSS, amongst many others. Online resources on using these tools can be found on Coursera.

A typical example in the banking industry would be advanced campaign analytics. It can help predict the likelihood that a customer would respond to a given marketing offer to improve the cross-selling and up-selling of products. Another example would be predicting the probability of fraud on credit cards.

Helpful resources

• MOOC on Coursera on R for beginners: Link
• A comprehensive guide to Python for beginners: Link
• Predictive model building on Coursera: Link

3. Prescriptive Analytics

Prescriptive analytics is the most sophisticated type of analytics that uses stochastic optimization and simulation to explore a set of possible options and recommend the best possible action for a given situation. It involves finding answers to ‘What should be done?’.

Prescriptive models go beyond descriptive models that only address what is going on, and beyond predictive models that can only tell what will happen, as they go on to advise what actually should be done in the predicted future. They quantify the effect of future actions on key business metrics and suggest the most optimal action.

Prescriptive models

Prescriptive models synthesize big data and business rules using complex algorithms to compare the likely outcomes of a number of actions, and choose the most optimum action to drive business objectives. Most advanced prescriptive models follow a simulation process where the model continuously and automatically learns from the current data to improve its intelligence.

These models are typically the most complex in nature and hence are being used by some progressive and large companies, as they are difficult to manage. However, when implemented correctly, they can have a strong impact on a company’s decision making effectiveness and hence, its bottom line.

That said, technical advancements such as super computers, cloud computing, Hadoop HDFS, Spark, in-database processing, MPP architecture, etc. have made deployment of complex prescriptive models using structured and unstructured data much easier. Tools used to run prescriptive models are mostly the same as predictive models however, require advanced data infrastructure capabilities.

Common example of prescriptive models

The retail banking industry is the optimum allocation of sales staff across various branches of the bank to maximize new customer acquisitions. Marrying geo-location information with every branch’s performance and potential, the model can prescribe the most optimum allocation of sales staff across all the branches.

A more sophisticated prescriptive modeling approach is used in the airline ticket pricing systems to optimize the price of air tickets based off travel factors, demand levels, purchasing timing, etc. to maximize profit margins, but also at the same time not deter sales.

According to a research, about 10% organizations use some form of prescriptive analytics at present, this figure has increased from 3% in 2014 and is expected to rise to 35% by 2020. Factors like massive investments in predictive analytics, expansion of IoT capabilities that compliment prescriptive analytics are driving this growth and expanding the scope of prescriptive models.

Helpful resources (in addition to the ones on Predictive analytics):

• Guide to building a recommendation engine in Python: Link
• MOOC on Coursera for practical hands-on Machine Learning: Link
• Guide to learning random forests: Link

Conclusion

In this article, I’ve discussed 3 different versions of analytics used across industries today. These are the building blocks of the analytics industry across the world. It is fair to say that all models, developments, and discoveries made using data can be classified under any of these three categories.

This article is meant to help people who are new to analytics or plan to switch over to analytics to get a clear view of the domain. I hope the mentioned resources help you get started with learning.