Introduction to Data Analytics
Analytics Framework Case Study and
Upcoming Trends
Learning Objectives
By the end of this lesson, you will be able to:
Explain the customer analytics framework
Explain the phases of customer analytics framework
List the latest trends in data analytics
Case Study: Ernst & Young
Customer Analytics Framework
An analytics framework helps perform data analysis in an organized manner.
The framework allows you to focus on the business outcome.
Case Study: EY
EY created a customer analytics framework for personalized customer experiences
to win more business and drive loyalty in a digital world.
Case Study: EY
To create the customer analytics
framework, company considered
these factors:
● Who are your customers?
● What do they do?
● What do they want?
● How and when to reach them?
Customer Analytics Framework
Phases of Customer Analytics framework
Data
Modeling
understanding
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5
Business Data Model
needs preparation monitoring
Business Needs
Business Needs
Grow Optimize Protect
● Acquire new customers ● Optimize pricing and ● Understand how to
cost to enhance retain customers
● Understand product life customer satisfaction
cycle ● Perform sentiment
analysis
● Develop new products
Data Understanding
Data Understanding
Data understanding is the second stage in the customer analytics framework.
This stage helps to draw patterns by gaining insight from the data.
Data Understanding
Data understanding is highly investigative and diagnostic.
Companies look at customer They identify the most valuable
needs and priorities to attract customers who aid their
potential customers. growth.
Data Understanding
Perform market segmentation for effective marketing and customer engagement by
dividing customers into groups based on:
Age Gender
Interests Availability
Spending habits
Data Understanding
Sentiment analysis is important to identify the sentiments of the customer through
social media.
Influence score measures the degree of influence of each user.
It can be combined with sentiment measure to identify disgruntled customers.
Data Preparation
Data Preparation
Take transformation Determine the data
steps
Collect and consolidate Improve data quality
data and completeness
Standardize data
structure
Data Preparation
Perform data mining
Work with structured and unstructured data
Use various tools and software to transform data.
Integrate data from various sources
Modeling
Modeling
Modeling stage focuses on developing models and can be based on:
Predictive Prescriptive
analytics analytics
Modeling
Predictive Prescriptive
analytics analytics
It helps understand the It helps predict possible
future and answer What outcomes and answer
could happen? What should we do?
As this phase is iterative, revisiting data preparation phase to refine the data is needed.
Modeling
Price
Attrition
Optimization
Model
Model
● Helps calculate how demand varies ● Is created through predictive
at different price levels algorithms
● Uses data to recommend prices for ● Helps companies gain better
improving profits understanding and take preventive
measures for employee attrition
Models can also be made for web analytics and sentiment analysis.
Modeling
Types of training models
Static Dynamic
This model is trained offline This model is trained online
as the model is trained once as the data is fed into the
and used for a while. model to train continuously.
Modeling
Types of predictions from trained models:
Online Batch
It is also called HTTP It is used for processing
prediction and is used accumulated data when
when timely inference is immediate results are
needed. not needed.
Model Monitoring
Model Monitoring
Model monitoring is the final stage where you need to establish, monitor, and meet
service-level agreements.
Example: SLAs for analytics might be the maximum time taken to create or deploy a
model.
Model Monitoring
● Data scientists monitor machine learning
models for drift.
● Drift means the data is no longer relevant
or useful as data is always changing.
● Data scientists ensure that the model
inputs look similar to those used in
training.
Factors in Model Monitoring
Cost Latency Throughput
Model cost needs to It is the delay between the It is the amount of data
analyzed to check whether data transfer instruction successfully moved from
the value generated from and the actual data transfer. one place to another in a
the model is worth the cost. given time period.
Latest Trends in Data Analytics
Latest Trends in Data Analytics
Cognitive Augmented Graph
Computing Reality Analytics
Automated Open
Machine Learning Source AI
Cognitive Computing
Cognitive Computing
Cognitive computing is an advanced type of artificial intelligence in the
cybersecurity domain.
It uses machine learning algorithms and deep learning networks to learn from
human interactions and provides actionable insights.
Augmented Reality
Augmented Reality
According to Gartner Inc., augmented analytics will be the dominant driver of new
purchases of business intelligence and analytics by 2020.
Augmented Reality
Preparing data
Building models
Augmented Analyzing data
analytics
Graph Analytics
Graph Analytics
Graph analytics is also known as network analytics and uses graphs to analyze data.
It is used for detecting crimes, spotting frauds, and applying influencer analysis in
social network communities.
Graph Analytics
Graph analytics highlights dominant edges.
Example: A large number of payments between bank accounts may indicate a
money laundering activity.
Automated Machine Learning
Automated Machine Learning
Machine
Learning Model
Domain knowledge Computer science Skills
Mathematical expertise
It involves a lot of tasks and it is prone to human errors and bias.
Automated Machine Learning
Automated Machine Learning
Enables organizations Helps improve return Reduces the amount of
to use existing on investment time taken to capture
knowledge value
Automated Machine Learning
Automated Machine Learning
Accelerates the Gives power to Delivers the right Exposes the
process of evolving a business users level of same degree of
trained model customization flexibility
Open Source AI
Open Source AI
Open source software has produced iconic innovations like the Firefox web browser,
Apache server software, and the Linux OS.
In open source AI, AI software libraries and algorithms are freely available to developers
and entrepreneurs.
Open Source AI
Many cloud-based technologies have their roots in open-source projects.
AI is expected to follow the trend as companies seek collaboration and knowledge sharing.
Key Takeaways
Customer analytics framework helps perform data analysis
in an organized way and allows to focus on the business
outcome.
Business needs, data understanding, data preparation,
modeling, and model monitoring are the different phases of
the analytics framework.
Cognitive computing, augmented reality, graph analytics,
automated machine learning, and open source AI are some
of the latest trends.
