Business Analytics

Introduction to Business Analytical Processes

Shyam Singhal
 Business Analytics

Scope

• What is BA, Why do we need it

• BI, DA, BA, DS – Improvement of Business through Data Insights

• Cross Industry Process for Data Mining (CRISP-DM) vs TDSP vs

Customization / Tailored

• Types of analysis (Descriptive, Diagnostic, Predictive, Prescriptive)

 Business Analytics

What is it?
 Business Analytics

What is Data Analysis?

 Business Analytics

• Most of the people are biased in

favor of AI?
• The candidate won with a significant
margin?
• I believe I is going to rain today?

• It will take me 20 minutes to reach

from destination – B from A?
• It will take me 20 minutes to reach
destination – B from A, if I take a
car?
• It will take me 20 minutes to reach
destination – B from A, if I take a
car, but during non-peak hours?
 Applications

• Consciously or unconsciously we all use data to make decisions

• Day to day operations also need data
• Which route will best at this time of the day?
• Whether it would rain?
• Do I have a chance to score a rank?
• Your apps like Google, ChatGPT, maps…all work on data
• Proactive maintenance is made possible because of data analytics
• …

Can’t imagine a world without data, analysis, use, improvement

 Business Analytics

• Is it quantitative data
assessment?
• Is it AI?
• Is it data analysis?
• Is it business intelligence?
• Is it data science?
• Is it a pattern?
Generating insights from data to
make informed (data driven)
decisions to aid business /
improve business performance.
 Business Analytics

• Predicting when a machine will break-down

• Predicting a suspicious / fraudulent activity / transaction
• Identifying intruders in the system
• Predicting attrition rate in a company
• Predicting market trend
• Recommending similar / connected products for cross / up sell
• Detecting a potential problem in assembly line
• Diagnosing illness based on patterns
• Which market segment is lucrative for a given product or service
• Segmentation, Targeting, and Positioning
 Knowledge Check

Q. Which are examples of an AI and why?

• Analyze remittance data to identify denials trends.

• Dynamically monitor data and notify key personnel when a known
problem scenario presents itself.
• Automatically send open claim information to a claim status vendor
and then import the results.
• Utilize a bot to login into a payer portal and check a patient’s
eligibility.
 Role of Business Analytics

Generating insights from data to make informed (data driven)

decisions to aid business / improve business performance
(competitiveness, innovation, sustainability…)
 Applicability of Business Analytics

• Decision Support
• Performance Management / Governance
• Problem Solving (Diagnostic / Predictive / Decision Science / Prescriptive…)
• Strategy Formation and Planning
• Optimization (Six-Sigma, Lean, Linear Programming, Prescriptive Analysis….)
• Risk Management (Fraudulent Transactions, Anomaly / Fault Detection…)
• Customer Insights (Recommendations, Personalization, Sentiment Analysis…)
• …
 Types of Analysis
• Descriptive Analysis
• What has happened.

• Diagnostic Analysis
• Why and How did it happen

• Predictive Analysis
• What can happen OR What might happen in future

• Prescriptive Analysis
• How it must be done OR What must we do next
 Descriptive Analysis – what has happened?

Finding out data properties

• Mean, Median, Mode, Range, Quartiles, SD, Variance, Skewness,
Kurtosis
• Cyclicality, Seasonality, Associations, Patterns, Trends
• Scatter diagrams, box-plots, control / specification limits
• Ratios, Frequencies, Spread, Bi-variate/Multivariate analysis
• Grouping, sub-grouping analysis, Vertical, Horizontal analysis

Q. Based on this data, can we make predictions / projections /

forecasting?
 Descriptive Analysis – what has happened?

Use Cases

• Reports, Visualizations, Dashboard for any metrics-based reporting

• Financial analysis – ratios

• Synthetic Test Data Generator

• Anything to do with initial trends, patterns identification

 Diagnostic Analysis – why did it happen?
Finding out the causes – why did it happen?
• First step toward remediation
• Correlation – association and not causation
• Hypothesis testing – establishing the truthfulness of gained knowledge
• T-test, z-test, F-test, ANOVA
• Vetting out reported anomalies
• Vetting out discovered associations
• Regression – causation and structuring of association
• Qualitative – Fishbone diagram

• Regression allows us to gain insights into the structure (shape and slope of
the curve) of that relationship and provides measures of how well the data
fit that relationship” – Professor Jan Hammond, Harvard Business School

Q. But wouldn’t that mean we can also make predictions / projections /

forecasting?
 Diagnostic Analysis – why did it happen?

Use Cases

• Marketplace analysis
• Why did product fail / succeed?
• Why said markets are lucrative?
• Financial analysis
• Fraud / risk detection
• Default detection (?)
• Healthcare analysis
• Narrowing down the illness

But almost all have elements of prediction in them.

Predictive Analysis – why might happen in the
future?

Use Case

• Payment default prediction

• Risky customer assessment
• Fraud detection
• Defects in a unit of work
• Can also be used for betting (unethical, but yes can be done)
• Time it would take to build a unit of work
• Will the patient have a cancer

Q. If we have this power or technique at our disposal then why do we

still make mistakes?
 Prescriptive Analysis – what should we do
next?

Based on analyzed data / predictions, what is prescribed or recommended

• Electricity data analysis suggests peak DD during X – action is with Z

• IoT data analysis predicts machine breakdown – preventive maintenance
is with Z
• Based on your consumption pattern, retailers recommending other
products – decision is with Z
• Loan application rejection based on analyzed data – decision is with Z

Works with predictive modeling. Helps in decisioning. Acts as a decision

support system.
 Structured / Unstructured Data…

• Data having a fixed format

• Data can’t be stored in tabular format
• Data requiring specialized operations prior to being used in
predictive analysis
• Easily understood by business users
• Has a greater number of tools for analysis
• Stored in native format
• Gets accumulated and stored easily
• Requires expertise and specialized tools for processing

Q. Can you do predictive analysis with unstructured data?

 Data Management

• Data Collection – creation, gathering, validation

• Curation – Cleaning (DQM), ETL
• Storage – DW, Data Lake, Security
• Governance – DQM, Business Rules, Access Policies
• Analysis – Inferences via exploration / modeling etc.
• Utilization – Business decisions
• Life Cycle Management – Retention, Expiration, Refresh / Update

• Data Mesh, Data Fabric, Data Federation, Data Virtualization, Data

Mirroring, Data Democratization…
 Data / Business Analytics Life Cycle
• Data Acquisition:
• Data Sources: Identifying and accessing relevant data sources.
• Data Collection: Gathering data from various sources (e.g., databases, APIs,
web scraping).
• Data Integration: Combining data from multiple sources into a unified
format.

• Data Preparation:
• Data Cleaning: Handling missing values, outliers, and inconsistencies.
• Data Selection & Transformation: Normalizing, scaling, and transforming
data for analysis.
• Data Enrichment & Feature Engineering: Creating new features or
transforming existing ones to improve model performance.
 Data / Business Analytics Life Cycle
• Data Mining:
• Model Selection: Choosing appropriate algorithms based on the problem and
data characteristics.
• Model Training: Building and training models on the prepared data.
• Model Evaluation: Assessing model performance using relevant metrics.
• Pattern Evaluation and Knowledge Discovery:
• Pattern Identification: Discovering patterns, trends, and anomalies.
• Pattern Interpretation: Understanding the significance of discovered
patterns.
• Visualization: Presenting insights in a visually appealing and understandable
format.
• Knowledge Extraction: Extracting actionable insights from the patterns.

• Knowledge Deployment:
• Decision Support: Integrating insights into decision-making processes.
• Deployment: Implementing models into production systems for real-time or
batch processing.
• Continuous Improvement: Continually improve / train the model or
underlying analysis techniques to environment changes
 Data Mining – CRISP-DM

Cross Industry Standard Practices for Data Mining (CRISP-

DM)

• Understand Business (Business Understanding)

• Understand Data (Data Understanding)
• Prepare Data (Data Preparation)
• Perform Modeling (Modeling)
• Conduct Evaluation (Evaluation)
• Deploy Solution (Deployment)
 Data Mining – CRISP-DM

Cross Industry Standard Practices for Data Mining (CRISP-

DM)

• Understand Business (Business Understanding)

• What are business goals?
• What challenges are being faced by the business?
• What are the current capabilities?
• What are the desired level / capabilities?
• Form the problem statement – What do we need to do?
• What are the available alternatives?
• Explore and evaluate alternatives
• Form the chosen approach to the solution
 Data Mining – CRISP-DM

Cross Industry Standard Practices for Data Mining (CRISP-

DM)

• Understand Data (Data Understanding)

• What are the Data Sources?
• How do we collect / gather the data?
• Basic data analysis – primarily Descriptive Analysis
• What is our data quality level?
 Data Mining – CRISP-DM

Cross Industry Standard Practices for Data Mining (CRISP-

DM)

• Prepare Data (Data Preparation)

• Clean the data – missing / NULL / invalid values, duplicates,
referential integrity, Outliers…
• Transform the data – ETL / ELT for Summarization,
Segmentation, Consolidation, Normalization / De-
normalization…
• Feature Engineering – creating new data from primary data,
i.e., new features / attributes / factors / columns
 Data Mining – CRISP-DM

Missing Value Imputation:

• Mean/Median Imputation: Replace missing values with the mean or median of the
respective feature.

• Mode Imputation: Replace missing categorical values with the most frequent category.
•
• K-Nearest Neighbors (KNN) Imputation: Impute missing values based on the values of
the k nearest neighbors.

• Multiple Imputation by Chained Equations (MICE): A more sophisticated technique that

generates multiple imputed datasets to account for uncertainty in the imputation process.

• Predictive Modeling: Train a predictive model to predict missing values based on other
features.
 Data Mining – CRISP-DM

Cross Industry Standard Practices for Data Mining (CRISP-

DM)

• Perform Modeling + Conduct Evaluation (Modeling + Evaluation)

• Statistical / ML / Neural Networks (DL) / Gen AI models
• Classification / Clustering / Anomaly Detection /
Recommendation
• Build  Train  Evaluate  Deploy  Improve
• Evaluation – Accuracy, Fitment (Data + Objective), Precision,
Recall, F1 Score
 Data Mining – CRISP-DM

Cross Industry Standard Practices for Data Mining (CRISP-

DM)

• Deploy Solution (Deployment)

• Deploying the solution in production
• Monitoring the outcomes / outputs fitment for Data and
Objectives
• Reporting the outcomes
• Continually improving the deployed solution
 TDSP

Team Data Science Process

• Business Understanding
• Data Acquisition and
Understanding
• Modeling
• Deployment
• Customer Acceptance
 CRISP-DM vs TDSP

Cross Industry Standard Team Data Science Process

Practices for Data Mining
(CRISP-DM)
• Business Understanding
• Understand Business
(Business Understanding) • Data Acquisition and
• Understand Data (Data Understanding
Understanding)
• Prepare Data (Data
Preparation) • Modeling
• Perform Modeling (Modeling)
• Conduct Evaluation • Deployment
(Evaluation) • Customer Acceptance
• Deploy Solution
 Summary
• Business Intelligence (BI), Data Analytics (DA), Business Analytics
(DA), Data Science
• All are different names to generate data based insights to solve business
problems or to improve business performance

• Data insights can be generated through simple Descriptive Analysis

to complex AI/ML/DL based modeling

• Data analysis requires Data Preparation

• Your analysis and model is as accurate and/or helpful as your data

• Types of Data Analysis

• Descriptive, Diagnostic, Predictive and Prescriptive

• Data Mining Techniques

• Classification, Clustering, Anomaly Detection, Recommendation

• CRISP-DM, TDSP…Common sensical approach to Data Analysis

Q&A