MODULE I: INTRODUCTION TO DATA ANALYTICS

Curricular Scope & Pedagogical Objective: Module I establishes the foundational theoretical, architectural, and strategic underpinnings of quantitative analytics within the Calicut University B.Com (Honours) curriculum for Data Analytics for Business (COM2MN109). It explores the conceptual genesis of data analytics, contrasts raw facts against actionable business wisdom via the DIKW hierarchy, details cross-industry commercial applications, defines the strategic paradigm of Business Analytics (BA) versus Business Intelligence (BI), delineates the transition from intuitive decision-making to Data-Driven Decision Making (DDDM), categorizes modern data structures and analytical tooling ecosystems, demystifies the multi-disciplinary domain of Data Science using Conway's framework, establishes the standard CRISP-DM lifecycle, and applies empirical decision models alongside corporate case studies and university examination review questions.

Unit 1: Introduction to Data Analytics & Applications of Data Analysis

1.1 Meaning, Definition, and Core Paradigm of Data Analytics

Data Analytics is the scientific discipline of examining, cleaning, transforming, modeling, and interpreting raw datasets with the explicit objective of discovering meaningful patterns, extracting actionable insights, drawing substantiated conclusions, and supporting strategic decision-making. In contemporary commerce, data analytics represents the primary cognitive engine that converts massive, unstructured organizational data streams into quantifiable competitive advantage.

Formally, analytics combines mathematical theory, statistical algorithms, computational technology, and domain expertise. While classical statistics focused primarily on historical inference and descriptive summarization of sampled data, modern data analytics operates on comprehensive enterprise datasets ("Big Data"), utilizing algorithmic learning to uncover non-linear correlations, predict future behavioral trajectories, and automate complex managerial decisions in real time.

1.2 The Data-Information-Knowledge-Wisdom (DIKW) Pyramid

The foundational paradigm underpinning business analytics is the DIKW Hierarchy (articulated by Russell Ackoff and Milan Zeleny), which explains the progressive cognitive transformation of raw digital signals into enlightened commercial execution:

THE DIKW COGNITIVE HIERARCHY VALUE ASCENDANCY MODEL

DATA (Facts) → INFORMATION (Context) → KNOWLEDGE (Understanding) → WISDOM (Action)

Progressive Stages of Cognitive Refinement:

1. Data (Symbols & Unprocessed Signals): Raw, uncontextualized, disorganized facts, numbers, alphanumeric codes, or sensory signals without inherent meaning. Example: A spreadsheet column containing raw integers: "105, 112, 98, 120".
2. Information (Processed Data with Context): Data that has been structured, categorized, aggregated, and formatted to answer questions of "Who?", "What?", "Where?", and "When?". Example: "The daily unit sales of Product X at the Calicut retail branch were 105 on Monday, 112 on Tuesday, 98 on Wednesday, and 120 on Thursday."
3. Knowledge (Synthesized Understanding & Patterns): The cognitive internalization and application of information, combining trends, historical benchmarks, and operational context to answer "How?" and "Why?". Example: "Analyzing 52 weeks of sales data reveals that whenever local outdoor temperatures exceed 35°C, sales of Product X increase by 22% due to heightened beverage consumption."
4. Wisdom (Actionable Strategic Judgment): The ethical, visionary, and proactive evaluation of knowledge, integrating corporate vision, resource constraints, and market dynamics to decide "What should we do?". Example: "Reallocating regional supply chain inventory and increasing dynamic digital promotional spend in North Kerala branches ahead of forecasted heatwaves to capture market share."

1.3 Cross-Industry Commercial Applications of Data Analysis

Across the global economy, data analytics has revolutionized operational workflows and business models:

1. Financial Services, Banking & FinTech

• Real-Time Fraud Detection: Machine learning algorithms evaluate credit card transactions in milliseconds, flagging anomalous geographic locations, spending velocities, or unusual merchant categories.
• Credit Risk Scoring: Quantitative models assess default probability by analyzing credit history, debt-to-income ratios, and alternative digital footprint data.
• Algorithmic Trading: Quantitative hedge funds deploy predictive models to execute high-frequency arbitrage across capital markets.

2. Retail, Marketing & E-Commerce

• Customer Churn Modeling: Identifying early behavioral signals (declining session frequency, reduced cart size) to trigger automated retention incentives.
• Collaborative Recommendation Engines: Amazon and Flipkart analyzing co-purchase matrices ("Customers who bought this also bought...") to drive cross-selling.
• Dynamic Pricing: Airlines, ride-hailing apps (Uber, Ola), and hotels continuously adjusting prices based on real-time demand, competitor rates, and weather.

3. Supply Chain & Logistics

• Predictive Demand Forecasting: Utilizing historical sales, macroeconomic indicators, and seasonality to prevent stockouts while slashing carrying costs.
• Route Optimization: Logistics giants (DHL, FedEx) running GPS and traffic algorithms to minimize delivery fuel consumption and idle transit times.
• Predictive Maintenance: IoT sensors on factory machinery predicting bearing wear and mechanical failure before catastrophic line breakdowns occur.

4. Human Resource Management (People Analytics)

• Talent Acquisition & Resume Parsing: Natural Language Processing (NLP) screening thousands of applicant resumes against role competency benchmarks.
• Employee Flight Risk Modeling: Analyzing tenure, compensation ratios, promotion cycles, and engagement scores to proactively retain key executive talent.
• Workforce Productivity Optimization: Balancing shift schedules against store footfall trends.

Unit 2: Business Analytics & Modern Decision-Making Architecture

2.1 Defining Business Analytics (BA)

Business Analytics refers to the systematic use of data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management to drive decisions and actions. While data analytics represents the overarching scientific discipline, business analytics is the explicit application of these analytical capabilities to solve commercial problems, maximize operational efficiency, optimize financial returns, and create sustainable shareholder value.

2.2 Business Analytics (BA) vs. Business Intelligence (BI) vs. Traditional Reporting

In corporate discourse, the terms Reporting, Business Intelligence, and Business Analytics are frequently conflated; however, they signify fundamentally distinct tiers of analytical maturity, temporal focus, and business impact:

Comparative Dimension	Traditional Reporting	Business Intelligence (BI)	Business Analytics (BA)
Temporal Orientation	Purely backward-looking (Historical / Static).	Retrospective & Present (Near Real-Time).	Forward-looking & Prescriptive (Future).
Primary Question Answered	"What happened?" (e.g., quarterly sales ledgers).	"What happened and where?" (e.g., interactive regional dashboards).	"Why did it happen, what will happen next, and what should we do?"
Core Methods & Tools	Manual spreadsheets, static PDF tables, relational SQL dumps.	OLAP cubes, interactive visual dashboards (Power BI, Tableau), KPI scorecards.	Machine learning algorithms, regression, simulation, optimization, predictive modeling.
Business Value & Mindset	Passive compliance and bookkeeping record maintenance.	Monitoring operational performance and trend discovery.	Active strategic disruption, automated execution, and competitive differentiation.

2.3 The Paradigm Shift: From Intuition ("HiPPO") to Data-Driven Decision Making (DDDM)

Historically, corporate management relied predominantly on executive intuition, past personal experience, and the "HiPPO" effect—the Highest Paid Person's Opinion. While managerial experience remains valuable, unvalidated intuition is plagued by subjective cognitive biases (overconfidence bias, confirmation bias, recency bias).

Data-Driven Decision Making (DDDM) replaces speculation with empirical evidence. Academic research from the MIT Center for Digital Business demonstrates that enterprises practicing DDDM achieve 5% to 6% higher productivity and market valuations compared to competitors relying on traditional intuitive methods.

2.4 Gartner's Analytic Ascendancy Model

Gartner defines four progressive tiers of analytical capability, wherein business value escalates in direct proportion to algorithmic difficulty:

GARTNER'S FOUR-TIER ANALYTIC MATURITY FRAMEWORK ANALYTIC ASCENDANCY

DESCRIPTIVE → DIAGNOSTIC → PREDICTIVE → PRESCRIPTIVE

1. Descriptive Analytics (Hindsight): Summarizes historical data to understand past operational performance. (Question: What happened? — e.g., monthly sales reports, website traffic totals).
2. Diagnostic Analytics (Insight): Drills down into data anomalies using root-cause analysis, correlation, and multi-dimensional filtering. (Question: Why did it happen? — e.g., investigating why customer churn spiked in Q3).
3. Predictive Analytics (Foresight): Applies statistical models, machine learning, and time-series forecasting to project future probabilities. (Question: What will happen? — e.g., forecasting next quarter's revenue, loan default probability).
4. Prescriptive Analytics (Optimization / Autonomous Action): Deploys mathematical optimization, simulation algorithms, and decision rules to recommend optimal action courses. (Question: What should we do? — e.g., automated algorithmic supply chain reordering, dynamic airline seat pricing).

Unit 3: Key Concepts and Pillars in Business Analytics

3.1 Foundational Analytical Terminology & Variable Architectures

To effectively interpret analytical models, commerce professionals must master core statistical and operational data concepts:

Data Point & Observation: A single individual entity, transaction, or unit of analysis recorded in a dataset (e.g., one customer checkout invoice).
Variables / Features: Specific characteristics or measurable attributes that vary across observations:
- Independent Variable (Predictor / X): The presumed causal factor manipulated or observed to explain changes (e.g., advertising expenditure, discount rate).
- Dependent Variable (Outcome / Y): The performance metric or target variable being predicted or explained (e.g., quarterly sales revenue, customer retention).
- Moderating Variable: A variable that alters the direction or strength of the relationship between X and Y (e.g., customer income level moderating price sensitivity).
- Mediating Variable: The underlying mechanism or intervening psychological state through which X influences Y (e.g., brand trust mediating the effect of advertising on repeat purchases).
Metadata: Structured informational context defining the characteristics, origin, format, and ownership of data ("data about data" — e.g., creation timestamp, table schema, file size, author).

3.2 Classification of Data Structures: Structured, Semi-Structured, and Unstructured

In the modern data landscape, enterprise information arrives in diverse architectural formats:

1. Structured Data

Highly organized data adhering to rigid tabular schemas (rows and columns). Stored in Relational Database Management Systems (RDBMS) like SQL Server, MySQL, and Oracle. Easily queried using Structured Query Language (SQL). Comprises ~20% of enterprise data (e.g., accounting general ledgers, sales transactions, employee records).

2. Semi-Structured Data

Data that does not conform to rigid relational database tables, but contains internal self-describing tags, markers, or organizational hierarchies. Stored in NoSQL databases (MongoDB, CouchDB). Common formats include JSON, XML, CSV, and server event logs.

3. Unstructured Data

Data possessing no predefined conceptual schema or organized data model. Represents over 80% of all newly generated global data. Includes customer service call audio, PDF contracts, video surveillance feeds, social media comments, and emails. Processed using NLP, Computer Vision, and Deep Learning.

3.3 The Five Dimensions of Data Quality

The classic computer science aphorism "Garbage In, Garbage Out" (GIGO) applies universally to business analytics: if input data is contaminated or inaccurate, even the most sophisticated machine learning model will produce disastrous decisions. Rigorous data governance enforces five core quality dimensions:

Quality Dimension	Operational Definition	Commercial Failure Manifestation
1. Accuracy	The degree to which data correctly reflects the true real-world entity or transaction without measurement error.	Incorrect pricing entry leads to massive billing disputes and margin erosion.
2. Completeness	Ensuring all mandatory data fields and expected records are present without missing values or truncated entries.	Missing customer postal codes prevents delivery routing and geographical segmentation.
3. Consistency	Data values across multiple enterprise systems (e.g., ERP, CRM, Billing) remain synchronized without contradiction.	CRM shows customer active while billing database flags them as cancelled, causing erroneous credit charges.
4. Timeliness	The availability of data within the required operational decision window; freshness versus latency.	Detecting fraud 24 hours after fraudulent funds have been wired internationally is useless.
5. Validity	Data conforms strictly to expected syntax, domain ranges, format rules, and data dictionary constraints.	Entering alphanumeric characters into a phone number field causing system crashes.

3.4 The Analytical Tooling Ecosystem

Modern business analysts deploy an integrated toolchain to extract value from data:

Spreadsheet Modeling (Microsoft Excel, Google Sheets): The ubiquitous workhorse of business finance. Used for quick ad-hoc analysis, financial modeling, PivotTables, VLOOKUP/XLOOKUP, and what-if sensitivity analysis via Solver.
Relational Querying (SQL): The universal standard language for querying, filtering, joining, aggregating, and extracting datasets from enterprise relational databases and cloud data warehouses (Snowflake, BigQuery).
Statistical & Machine Learning Programming (Python & R): Python (with libraries like Pandas, NumPy, Scikit-learn, Matplotlib) and R are the premier languages for advanced statistical computing, automated predictive modeling, and machine learning pipelines.
Data Visualization & BI Dashboards (Power BI & Tableau): Transform complex datasets into interactive visual storytelling reports, enabling non-technical corporate executives to explore KPIs intuitively.

Unit 4: Concept of Data Science & Core Areas of Study

4.1 Demystifying Data Science: Drew Conway's Venn Diagram

Data Science is an interdisciplinary field that utilizes scientific methods, processes, algorithms, and computational systems to extract actionable knowledge and empirical insights from structured and unstructured data. In 2010, data scientist Drew Conway conceptualized Data Science as the dynamic intersection of three foundational disciplines:

DREW CONWAY'S DATA SCIENCE VENN DIAGRAM INTERDISCIPLINARY ARCHITECTURE

COMPUTER SCIENCE (Hacking) ∩ MATH & STATISTICS ∩ SUBSTANTIVE DOMAIN EXPERTISE

Intersection Zones:

Computer Science + Math & Statistics = Machine Learning: Developing algorithmic models that learn from data without explicit business domain context.
Math & Statistics + Domain Expertise = Traditional Research: Classical academic hypothesis testing and econometric analysis conducted without scalable software engineering.
Computer Science + Domain Expertise = The "Danger Zone": Writing code and extracting patterns without statistical rigor, leading to spurious correlations and false conclusions.
The Center (All Three Intersected) = TRUE DATA SCIENCE: Combining software engineering capability, mathematical-statistical rigor, and deep commercial domain understanding.

4.2 Data Science vs. Business Analytics vs. Data Engineering

In corporate talent architectures, these three complementary roles form the data value pipeline:

1. Data Engineer

Architect of the Plumbing: Builds and maintains reliable data pipelines, ETL (Extract, Transform, Load) workflows, distributed data lakes (Hadoop, Spark), and cloud warehouses to ensure clean data arrives reliably.

2. Data Scientist

Algorithmic Pioneer: Formulates custom machine learning models, neural networks, and NLP algorithms to solve complex, unstructured business questions and discover hidden predictive patterns.

3. Business Analyst

Commercial Translator: Bridges the gap between technical data models and executive decision-makers. Translates business problems into analytical queries and converts analytical findings into ROI-driven corporate strategies.

4.3 Core Interdisciplinary Areas of Study in Data Science

An academic curriculum in Data Science encompasses five vital intellectual domains:

1. Applied Mathematics and Probability Statistics: Linear algebra (matrix operations), multivariable calculus (gradient descent optimization), probability distributions (normal, binomial, Poisson), inferential statistics, hypothesis testing (t-tests, ANOVA, chi-square), and regression analysis.
2. Computational Algorithms & Software Engineering: Object-oriented programming, data structures (trees, graphs, hash tables), cloud infrastructure (AWS, Azure, Google Cloud), and parallel processing for massive data.
3. Machine Learning (ML) & Artificial Intelligence (AI): Supervised learning (classification, regression), unsupervised learning (clustering, dimensionality reduction via PCA), reinforcement learning, and deep learning architectures.
4. Data Mining & Information Retrieval: Extracting hidden association rules (market basket analysis), anomaly detection, web scraping, and text mining.
5. Data Visualization & Cognitive Narrative Design: Principles of visual encoding (Edward Tufte's design principles), chart selection heuristics, and narrative dashboard design that avoids visual distortion.

4.4 The Industry Standard Process: The CRISP-DM Lifecycle

The universal methodology guiding data analytics and data science projects across global industries is the CRISP-DM (Cross-Industry Standard Process for Data Mining) framework:

CRISP-DM SIX-STAGE ITERATIVE LIFECYCLE STANDARD ANALYTICS WORKFLOW

BUSINESS UNDERSTANDING ↔ DATA UNDERSTANDING → DATA PREPARATION → MODELING → EVALUATION → DEPLOYMENT

Iterative Phase Execution:

1. Business Understanding: Uncovering the commercial objective, defining the problem from a business perspective, setting project success criteria, and translating goals into a formal data analytics problem definition.
2. Data Understanding: Collecting initial datasets, exploring data distributions, identifying data quality issues, and discovering first insights or interesting subsets.
3. Data Preparation (Consumes ~70-80% of project effort): Cleaning raw data, handling missing values, encoding categorical variables, normalizing scale, and engineering new analytical features.
4. Modeling: Selecting and applying appropriate modeling techniques (e.g., decision trees, logistic regression, neural networks) and calibrating hyperparameters to optimal values.
5. Evaluation: Rigorously evaluating model performance against business objectives. Ensuring the model does not overfit and truly achieves the desired commercial KPIs.
6. Deployment: Integrating the finalized analytical model into live production systems (e.g., embedding real-time fraud scoring into point-of-sale checkout APIs).

Analytical Application: Quantitative Exercises & Calculations

PRACTICAL PROBLEM 1: EXPECTED VALUE OF PERFECT INFORMATION (EVPI) IN DDDM
Case Scenario: A retail enterprise is deciding whether to launch a new product line. Market research indicates two possible future market states: High Demand (Probability = 0.60) or Low Demand (Probability = 0.40).
• Decision Option 1 (Launch): Yields a profit of ₹50 Lakhs under High Demand, but incurs a loss of −₹20 Lakhs under Low Demand.
• Decision Option 2 (Do Not Launch): Yields ₹0 under both states.
A data analytics consultancy offers to build a predictive market intelligence model that provides 100% accurate advance information on market demand. Calculate the Expected Value of Perfect Information (EVPI) to determine the maximum fee the enterprise should pay for this analytics model.

Step 1: Calculate Expected Monetary Value under Intuition / Uncertainty (EMV_Max):
EMV (Launch) = (0.60 × ₹50 Lakhs) + (0.40 × −₹20 Lakhs) = ₹30 Lakhs − ₹8 Lakhs = ₹22 Lakhs
EMV (Do Not Launch) = ₹0
• Without analytics, management would launch the product with an expected payoff of EMV_Max = ₹22 Lakhs.

Step 2: Calculate Expected Value with Perfect Analytics Information (EVwPI):
• If Analytics predicts High Demand → Optimal decision is Launch → Payoff = ₹50 Lakhs.
• If Analytics predicts Low Demand → Optimal decision is Do Not Launch → Payoff = ₹0.
EVwPI = (Probability of High Demand × Best Payoff) + (Probability of Low Demand × Best Payoff)
EVwPI = (0.60 × ₹50 Lakhs) + (0.40 × ₹0) = ₹30 Lakhs

Step 3: Calculate Expected Value of Perfect Information (EVPI):
EVPI = EVwPI − EMV_Max = ₹30 Lakhs − ₹22 Lakhs = ₹8 Lakhs
• Managerial Decision Rule: The maximum economically rational amount the enterprise should pay for the predictive data analytics solution is ₹8 Lakhs. Any fee below ₹8 Lakhs creates positive net financial surplus.

PRACTICAL PROBLEM 2: EMPIRICAL CLASSIFICATION OF BUSINESS DATA TYPES
Case Scenario: An analyst at an e-commerce platform compiles a customer dataset containing the following attributes. Classify each variable into its correct statistical scale (Nominal, Ordinal, Interval, Ratio) and data structure category (Structured, Semi-Structured, Unstructured).

Classification & Operational Rationale:
1. Customer ID (e.g., CUST-9842): Nominal Scale | Structured Data. Serves solely as a categorical identifier; arithmetic operations are meaningless.
2. Customer Satisfaction Rating (1 = Poor, 2 = Fair, 3 = Good, 4 = Excellent): Ordinal Scale | Structured Data. Possesses ranked order, but mathematical intervals between ranks are non-uniform.
3. Annual Purchase Spending (₹45,200.00): Ratio Scale | Structured Data. Quantitative continuous metric possessing a true absolute zero point (₹0 indicates zero spend).
4. Customer Product Review Text ("The product arrived damaged and late"): Unstructured Data. Natural language text lacking a predefined relational schema; requires Natural Language Processing (NLP).
5. User Clickstream Log (JSON format containing session timestamps and IP tags): Semi-Structured Data. Contains self-describing key-value tags without a fixed tabular database schema.

Real-World Empirical Corporate Case Studies

CASE STUDY 1: Netflix — The $100 Million Big Data Bet on "House of Cards"

In 2011, traditional Hollywood television networks greenlit television series based on executive gut feelings, pilot screenings, and focus group surveys. Netflix disrupted this entire model by leveraging granular user engagement analytics from 33 million subscribers.

The Analytical Synthesis: Netflix's data algorithms identified three distinct, overlapping high-engagement clusters: subscribers who consistently watched films directed by David Fincher, subscribers who completed movies starring Kevin Spacey, and subscribers who enjoyed the British political drama mini-series House of Cards.

The Strategic Result: Relying entirely on predictive correlation, Netflix bypassed the pilot episode process and committed an unprecedented $100 Million to produce two full seasons of House of Cards without testing a pilot. The series became a worldwide cultural phenomenon, validating that big data analytics can de-risk multi-million-dollar corporate capital investments.

CASE STUDY 2: Amazon — Dynamic Pricing, Recommendation Engines & Anticipatory Shipping

Amazon's global retail supremacy is built on continuous real-time analytics. In 2023, Amazon's proprietary algorithms updated retail prices over 2.5 million times per day across millions of SKUs, adjusting to competitor prices, local demand elasticity, inventory levels, and profit margins.

Commercial Impact: Over 35% of Amazon's total consumer purchases are directly generated by its collaborative filtering recommendation algorithms. Furthermore, Amazon patented "Anticipatory Shipping"—a predictive logistics model that packages and dispatches products toward regional distribution hubs before the customer has even placed the order, based on search history and historical order cycles.

CASE STUDY 3: Capital One — Transforming Banking via Information-Based Strategy (IBS)

Founded by Richard Fairbank and Nigel Morris, Capital One pioneered the concept of testing consumer credit card offers through systematic statistical experiments. While legacy banks charged uniform 19.8% interest rates to all applicants, Capital One deployed data analytics to run tens of thousands of real-time risk experiments.

Risk Segmentation: By tailoring interest rates, credit limits, cash-back rewards, and fee structures to individual credit risk profiles, Capital One transformed from a small regional division into one of the largest and most profitable credit card issuers in the world.

Exam Preparation & University Review Questions

CALICUT UNIVERSITY B.COM (FYUGP) HIGH-FREQUENCY EXAMINATION QUESTIONS

Part A: Short-Answer Questions (Conceptual - 2 Marks Each)

Define "Data Analytics".
Explain the Data-Information-Knowledge-Wisdom (DIKW) hierarchy with an example.
What is meant by the "HiPPO" effect in managerial decision-making?
Differentiate between "Structured Data" and "Unstructured Data".
State the four progressive tiers of Gartner's Analytic Ascendancy Model.
Define "Metadata" with two examples.
What is the CRISP-DM framework? Name its six stages.
Define the "Expected Value of Perfect Information" (EVPI).

Part B: Short-Essay / Analytical Questions (5 Marks Each)

Compare and contrast Business Intelligence (BI) and Business Analytics (BA) across temporal focus, questions answered, and tools used.
Discuss the Five Dimensions of Data Quality and explain the business consequences of poor data hygiene.
Explain Drew Conway's Data Science Venn Diagram and describe what occurs when only two circles intersect.
Detail the primary applications of data analytics in Financial Services and Retail Marketing.
Describe the roles and skill distinctions among Data Engineers, Data Scientists, and Business Analysts.

Part C: Comprehensive Essay Questions (10 / 15 Marks Each)

"Data is the new oil of the digital economy, but only analytics transforms it into actionable fuel." Elucidate this statement in detail. Examine the concepts, evolution, and strategic importance of Business Analytics in contemporary corporate decision-making.
Critically evaluate the CRISP-DM lifecycle for data mining projects. Detail the operational requirements, challenges, and iterative relationships between each of the six stages. Illustrate how an e-commerce firm would apply CRISP-DM to predict customer churn.
What is Data-Driven Decision Making (DDDM)? Contrast it with traditional intuitive decision-making. Analyze the economic value of predictive analytics using the Expected Value of Perfect Information (EVPI) model, and illustrate your answer with reference to the Netflix and Amazon case studies.

Key Takeaways for Examination Success:

Always illustrate the DIKW hierarchy as an ascending pyramid: Data → Information → Knowledge → Wisdom, providing clear commercial examples for each level.
In essay questions on BA vs. BI, construct a multi-column comparative table highlighting temporal perspective (backward-looking BI vs. forward-looking BA).
Memorize Gartner's 4 tiers: Descriptive (What happened?), Diagnostic (Why?), Predictive (What will happen?), Prescriptive (What should we do?).
Draw Conway's 3 overlapping circles: Computer Science, Math/Stats, and Domain Expertise, explicitly labeling the "Danger Zone" and "Machine Learning".