Analytics and Big Data

Analytics and Big Data

• Data without any analysis or interpretation is useless. At the same time,
there is no “easy” button to generate a great analysis in one simple
step.
• Analyzing statistics, writing a report, and using a modeling algorithm
are each only one step of several required to obtain a great analysis.
• Not understanding the aim of the analysis and focusing only on these
individual tasks without keeping the end goal or problem to be solved
in mind can lead to wrong decisions and generate a lot of extra work.
• Therefore, instead of working on individual tasks, the data should be
analyzed thoroughly by keeping the goal or objective in mind.
 Analytics and Big Data
• Big Data analytics refers to the process of collecting, organizing, and
analyzing large sets of data to identify patterns and other important
information.
• An analysis of Big Data helps organizations to take strategic decisions
based on the information extracted from Big Data. For example, sales
data can be analyzed to decide whether or not a product should be
sold in a particular month based on the product’s sales trend in that
month.
• Generally, organizations require the information that comes after
analyzing the data.
• As you know, data is available in structured and unstructured formats.
So it becomes a challenge for organizations to perform data analysis.
The data of different formats is collected, combined, and then
analyzed.
 Comparing Reporting and Analysis
• Reporting is a process in which data is organized and summarized in an easy-to-
understand format. Reports enable organizations to monitor various performance
parameters and improve customer satisfaction.
• Analysis is a process in which data and reports are examined to get insights from
them. These insights help an organization to perform important tasks in a timely
manner, such as planning a new strategy, taking important business decisions,
introducing a new product, and improving customer satisfaction.
• In simple words, reporting can be considered as a process in which raw data is
transformed into useful information and analysis as a process that transforms
information into insights.
• Reporting is used as a tool to monitor the day-to-day business operations of an
organization. A good report enables an organization to ask the most relevant
questions related to the business and customers.
 Comparing Reporting and Analysis
• Analysis, on the other hand, enables organizations to answer these
questions based on summarized data at a deeper level and thereby
find ways to implement the insights. However, there exists one major
difference between them.
• Reports enable an organization to understand what is happening but
analysis helps an organization understand why it is happening and what
action can be taken for it.
• An analysis can result in reports, and reports can result in analysis. It is
also possible to have an analysis that is completely based on reports.
Reports are essential for the growth of a business and add value to it.
 Reporting
• A reporting environment is also known as a Business Intelligence (BI)
environment. Such an environment enables users to select and view
reports, run them, and view the results. Reports might contain various
visual elements to aid understandability, including graphs, charts, and
tables.
• The key factors that define a report are as follows:
•  They provide the data that the user has been asking for.
•  They provide the data in a predefined format for easy
understandability.
•  They show the stakeholders what had happened in the past.
•  They allow the data analyst to define several pieces of data with a
few indices.
 Analysis
• An analysis refers to a process that involves a person trying to solve a
problem, finding the data required to obtain the required answer,
analyzing the data, and interpreting the results to suggest a
recommended course of action.
• The following are the important points that define an analysis:  It
provides answers to the questions being raised.
•  It provides the steps required to get answers to the questions.
•  The analysis is specific to the questions being addressed.
•  Analysis requires a person who guides the process.
 Analysis
• Let’s take an example to understand the process of analysis. Suppose the sales of product X of an
organization called OFR Corporation has fallen drastically in the last quarter. The sales manager
needs to know the reasons for the fall in sales figure of product X.
• For this, the manager must analyze the sales data for product X over the last quarter.
• Further, the manager must analyze the feedback of customers gathered from different social media
platforms, such as the Internet, e-mails, and so on.
• Before performing the analysis, the manager must ensure that the data he/she collects is structured
in a predefined format.
• After analyzing the data on the basis of an existing model, the manager would come to know about
the factors that have affected the sales of product X in the market.
• The result of the analysis would help the manager understand the areas that need attention.
• On the basis of the results obtained by the manager, he/she can devise a plan to improve the sales
of the product.
• Further, effective marketing plans can be created to improve the popularity of the product in the
market.
 Two Perspectives of Big Data Analysis
 Decision-oriented analysis―This approach is also called the traditional
business intelligence approach. Here, analysts try to use the results of the
analysis in the process of making business decisions.
 Action-oriented analysis―This approach is used when a rapid response
or an action is expected to a critical situation. In this type of analysis, a
particular pattern emerges or specific types of data are detected and an
appropriate action is required to be taken. For example, if sales data is
analyzed, you may find negative or missing values that need to be taken
care of by taking the appropriate action.
 The Analytic Process
1. Business understanding―The first phase involves identifying and
understanding the business objectives. It deals with the problems to
be solved and decisions to be made. The main goal is to enhance
business profitability. Once the business objectives are determined, the
analysts evaluate the situation, and identify the data mining goals.
According to the defined goals, a project plan is created between the
analytics team and the IT or development team.
2. Data collection―The process of collecting data is an important task in
executing a project plan accurately. In this phase, data from different
data sources is collected first and then described in terms of its
application and need of the project. This process is also called data
exploration. Exploration of data is required to ensure the quality of the
collected data.
 The Analytic Process
3.Data preparation―From the data thus collected, unnecessary or
unwanted data is to be removed in this phase. In other words, the data
must be prepared for the purpose of analysis.
4. Data modeling―In this phase, a model is created by using a data
modeling technique. The data model is used to analyze the relationships
between different selected objects in the data. Test cases are created to
assess the applicability of model and data is structured according to the
model.
5. Data evaluation―The results obtained from different test cases are
evaluated and reviewed for errors. After validating the results, analysis
reports are created for determining the next plan of action
 The Analytic Process
6. Deployment―In this phase, the plan is finalized for deployment. The
deployed plan is constantly checked for errors and maintenance. This
process is also termed as reviewing of the project.
 Types of Analytics
•  Basic analytics
•  Advanced analytics
•  Operational analytics
•  Monetized analytics
 Basic Analytics
• Basic analytics is used to explore the data whose value is not known. In other words,
the data may or may not prove to be useful. Basic analytics usually involves
visualizations of simple statistics. This type of analytics is used specially when a lot
of disparate data needs to be analyzed. The process of basic analytics includes
investigating what happened, when it happened, and its impact.
• Let us consider an example to understand this better. Suppose a product manager in
an organization needs to know about the sales promotion of certain products,
which was performed last month. He/she also wants to know how many new
customers have been associated with the organization due to the sales promotion.
In this case, to know “what” happened, core analysis would help in understanding
how many customers bought and used the products. To know the “impact” of this,
core analysis would help in understanding how much money the new customers
brought in, and this impacted when compared to the baseline.
 Basic Analytics
• Some examples of basic analysis are as follows:
•  Slicing and dicing―they are referred to the division of data into smaller groups
that are easy to explore. For example, sales data can be divided or categorized
on the basis of regions for better analysis.
•  Basic monitoring― It is to the monitoring of huge volumes of data in real time.
For instance, when your new ad campaign is launched, you might be interested
in monitoring data related to every single minute.
•  Anomaly identification― It refers to the identification of anomalies, for
example, identification of an event that shows a difference between the actual
observation and what you expected in your data. This may be due to some
anomaly in the operations of an organization.
 Advanced Analytics
• Advanced analytics is not just confined to answering questions such as what
happened, when it happened, and its impact. It goes one step beyond by finding
the cause for what has happened and the measures that can be taken to avoid it
from happening again in the future.
• Advanced analytics performs complex analysis of both structured and unstructured
data by using algorithms. It also uses advanced data mining techniques such as
neural networks, statistical models, machine learning, text analytics, and many
more. Additionally, it uses several activities such as complex ad hoc SQL,
forecasting, optimization, predictive modeling, data mining, and so on.
• To determine different patterns in data, complex event processing, forecasting, and
advanced analytics are used.
 Advanced Analytics
• The following are some examples of advanced analytics:
•  Predictive modeling―It refers to a data-mining solution that provides algorithms
and techniques to use on structured and unstructured data to ascertain future
outcomes. For instance, a telecommunications company can identify the customers
who are about to drop its service by using a predictive model.
•  Text analytics―It is to the process of examining unstructured text, extracting
important information out of it, and transforming the information into structured
information that can be leveraged in different ways.
•  Other statistical and data mining algorithms―These algorithms use techniques
such as advanced forecasting, optimization, cluster analysis for segmentation, and
affinity analysis.
 Advanced Analytics
• Consider a company that monitors the Web activities of its customers. It carries out
an analysis to ascertain whether viewing a product online increases the probability
of that product being purchased. It might be a daunting task to analyze the
complete Web data and integrate it with other customer data. This is because Web
data is constantly being generated and added to existing data. As a starting point,
correlation analysis can be performed. To analyze data for the first time, an
expensive and elaborative model and process are not required. However, if a strong
correlation exists between browsing data and sales data, then the company can
market its products easily to users who browsed the website but did not make any
purchases. May be later on, the analysts might think of quantifying the relationship
more precisely and might make use of advanced analytic models.
 Advanced Analytics
• Many statisticians and mathematicians have used techniques such as advanced
analytics for several years but it was never a part of the analytics landscape. Let’s
take an example of companies that took the help of statisticians 20 years back to
identify customers who might unsubscribe their services in the future.
• These statisticians predicted customer behavior on the basis of machine learning
algorithms or advance survival analysis. Most employees in the company found it
difficult to understand their meaning and how prediction can be used for benefit.
Further, it was difficult to achieve the computational power and analyze changing
data.
• In today’s world, advanced analytics has become more conventional. With the ever-
increasing advancements in computational power, new developments in algorithms,
enhanced data infrastructure, and the requirement for better insights from huge
volumes of data, companies are now using advanced analytics as a part of their
business-making process.
 Operationalized Analytics
• Operational analytics means making analytics an important part of the business
process. For instance, an insurance company can use a model to predict the
probability of a claim being fraudulent. This model can be used by the company to
form an important part of its claims processing system to flag claims having a high
probability of fraud. The fraudulent claims are then sent to a separate unit for
investigation, which reviews the claims further. In some cases, the model will work
for the end users from behind the scenes.
• For instance, in a company a model may be used to predict customers who can be
good targets for selling products when they make calls. While an agent of the
company is taking a call of a customer, he/she might receive a message about
additional products to sell to the customer. In such cases, the agent might not be
aware that a predictive model is working behind the scenes to make product
recommendations.
 Monetized Analytics
• Monetized analytics helps businesses to take important and better decisions and
helps earn revenues. However, Big Data analytics is also used to derive revenue
beyond the insights it provides. You might be able to get a unique data set that is
valuable for other companies. For example, credit card providers can use this data
to offer value-added products. Similarly, telecommunications companies like
financial institutions have started selling location-based insights to retailers. Data of
different types such as text-messaging data, location data, billing data and Web data
are used to find inferences about customer behavior patterns. These patterns will
help the retailers in finding useful information.
 Characteristics of Big Data Analysis
 It can be programmatic―Big Data analysts must use code to handle raw data.
Analysts can use the code to manipulate or explore the data. This is specially useful
when you have large volumes of data.
 It can be data-driven―A hypothesis-driven approach is used by several analysts to
develop a premise and gather data to see if the premise is correct or not. However, Big
Data uses a huge volume of data to drive the analysis.
 It can use a lot of attributes―Analysts in the past used hundreds of attributes of the
data set for analysis, and today Big Data uses thousands of attributes and millions of
observations to conduct analysis.
 It can be iterative―Big Data analysts use iterations on models that provide more
computational power. These models allow the analysts to do iterations to the required
level of analysis. Because of this, new applications are designed and developed to
analyze requirements and time frames to complete the analysis.
THANK YOU

15