DATA ANALYTICS

From Zero to Insight — A Complete Beginner's Guide

Covering: Descriptive • Diagnostic • Predictive • Prescriptive Analytics

1. What Is Data Analytics?

Data analytics is the process of examining raw data to find patterns, draw conclusions, and
support decision-making. Think of it as turning a pile of facts and numbers into answers to real
questions.
Every business, hospital, sports team, or government collects data constantly. Without
analytics, that data just sits there. Analytics gives it meaning.

The data journey

• Raw data — Numbers, text, dates, clicks, purchases — unprocessed facts collected
from the world.
• Processing — Cleaning, organising, and structuring data so it can be analysed
meaningfully.
• Insight — Patterns and conclusions that answer real questions and guide smart
decisions.

There are four progressive levels of analytics, each answering a different question:

Level 1 Level 2 Level 3 Level 4

Descriptive Diagnostic Predictive Prescriptive
What happened? Why did it happen? What will happen? What should we do?

Each level builds on the previous one. You cannot reliably predict the future without first
understanding the past.

2. Descriptive Analytics — What Happened?

Descriptive analytics is the most common and most basic form. It summarises historical data to
show what has already occurred. If you have ever looked at a monthly sales report, a
dashboard of website visitors, or an average test score — that is descriptive analytics.
It does not try to explain causes or predict the future. It simply tells you what the numbers say.

Data Analytics — Beginner's Guide | Page 1

 Real-world example
A school principal asks: "How many students passed the maths exam last term?" The answer —
"72% passed" — is descriptive analytics. It describes a past event in summary form.

Common techniques
• Averages, totals, and percentages
• Frequency counts and pivot tables
• Time-series charts and bar graphs
• Dashboards and histograms

Common tools
• Microsoft Excel and Google Sheets
• Power BI and Tableau
• Google Analytics
• SQL queries

Key statistical measures

• Mean — the average value across a dataset
• Median — the middle value when data is sorted
• Mode — the most frequently occurring value

Descriptive analytics answers the question 'what?' — it is the foundation. Without knowing what
happened, you cannot investigate why, predict what comes next, or decide what to do.

3. Diagnostic Analytics — Why Did It Happen?

Diagnostic analytics goes deeper. It examines data to understand the causes and relationships
behind outcomes. It asks 'why?' and looks for correlations, patterns, and root causes.

Real-world example
Sales dropped 20% in March. Diagnostic analytics investigates: Was it a pricing change? A
competitor campaign? A supply chain problem? Analysts drill into data to find the root cause.

Techniques used
• Drill-down analysis
• Correlation analysis

Data Analytics — Beginner's Guide | Page 2

 • Data mining and cross-tabulation
• Root cause analysis
• A/B test review

Key concept: correlation vs. causation

Two things moving together does not mean one caused the other. Ice cream sales and
drowning rates both rise in summer — but ice cream does not cause drowning. Both are caused
by hot weather. This distinction is one of the most important concepts in all of analytics.

4. Predictive Analytics — What Will Happen?

Predictive analytics uses historical data and statistical models to forecast future events. It gives
you a probability — not certainty — about what might happen next. This is where machine
learning algorithms become central.
Predictive models are trained on past data, learn patterns, then apply those patterns to new
data to generate forecasts. The more quality data you have, the better the predictions.

Real-world example
A bank uses a customer's transaction history, income, and credit behaviour to predict the probability
they will default on a loan. If the model says there is an 80% default risk, the bank may decline the
application or offer a higher interest rate.

The basic model-building pipeline

• Step 1: Collect data
• Step 2: Clean and prepare data
• Step 3: Choose an algorithm
• Step 4: Train the model
• Step 5: Test and evaluate accuracy
• Step 6: Deploy into production

Key prediction algorithms

Regression Linear & Logistic Regression

Linear regression predicts a number (e.g. house price). Logistic regression predicts a
category (e.g. yes/no, spam/not spam). It finds the relationship between input
variables and an output.
Tree Decision Trees
Works like a flowchart of yes/no questions, branching until it reaches a prediction.
Easy to understand and visualise, making it a popular starting point for beginners.

Data Analytics — Beginner's Guide | Page 3

 Forest Random Forests
Builds many decision trees and combines their outputs for a stronger, more accurate
prediction. Reduces the risk of overfitting (memorising training data too closely).
Neural Neural Networks
Loosely inspired by the human brain — layers of connected nodes process data and
learn complex patterns. Powers image recognition, language translation, and
recommendation engines.
Time Time Series Forecasting (ARIMA, LSTM)
Specifically designed for data that changes over time (sales, stock prices, weather).
Identifies seasonal patterns, trends, and cycles to project forward.
Cluster K-Means Clustering
Groups data points into clusters based on similarity — useful for customer
segmentation. Example: these 3,000 customers behave similarly, so target them with
the same offer.

5. Prescriptive Analytics — What Should We Do?

Prescriptive analytics is the most advanced level. It does not just tell you what will happen — it
recommends specific actions to achieve the best possible outcome. It combines predictions with
optimisation and decision logic.
Think of a GPS navigation system: descriptive analytics would show you the roads you have
already driven. Predictive would forecast traffic jams. Prescriptive would say 'take this route right
now to arrive 12 minutes earlier.'

Real-world example
A hospital uses prescriptive analytics to schedule staff. It predicts how many patients will arrive each
hour (predictive) and then automatically builds the optimal nurse and doctor schedule to minimise
wait times while staying within budget (prescriptive).

Key prescriptive algorithms

Optimise Linear Programming & Optimisation

Finds the best allocation of resources (time, money, staff) given a set of constraints
and goals. Used in supply chains, scheduling, and finance.
Simulate Simulation (Monte Carlo)
Runs thousands of what-if scenarios using random inputs to understand the range of
possible outcomes. Used in risk assessment, financial planning, and engineering.
Rules Decision Support Systems & Rule Engines
Uses business rules combined with predictions to automatically recommend or trigger
actions. E.g. If churn probability > 70%, send a retention offer.
Reinforce Reinforcement Learning
An AI agent learns by trial and error — taking actions, receiving rewards or penalties,

Data Analytics — Beginner's Guide | Page 4

 and improving over time. Used in robotics, game AI, and dynamic pricing.

6. Data Analytics Models — An Overview

A model in analytics is a mathematical representation of a real-world process. It learns patterns
from data and can then make predictions or recommendations on new, unseen data.
Understanding the main categories of models helps you choose the right one for a given
problem.

Supervised Supervised Learning Models

Trained on labelled data — examples with known answers. The model learns to map
inputs to outputs. Used for classification and regression tasks. Examples: spam
filters, fraud detection, price prediction.
Unsupervised Unsupervised Learning Models
Trained on data with no labels — the model discovers hidden structure on its own.
Used for clustering, anomaly detection, and dimensionality reduction. Examples:
customer segmentation, topic modelling.
Time Series Time Series Models
Designed for sequential data over time. Capture trends, seasonality, and cycles.
Used in sales forecasting, weather prediction, and stock analysis.
Deep Deep Learning Models
Learning Multi-layered neural networks that handle extremely complex patterns in images,
audio, and text. Require large amounts of data and computing power, but achieve
remarkable accuracy.

7. Tools Beginners Should Know

You do not need to know all of these — pick one path and go deep before branching out.

• Excel / Google Sheets — Best starting point for descriptive analytics and basic
visualisation
• SQL — Essential for querying and manipulating data in databases
• Python (with pandas, scikit-learn, matplotlib) — The most popular language for data
analysis and machine learning
• R — Excellent for statistical analysis and academic research
• Tableau / Power BI — Drag-and-drop dashboarding tools, no coding required
• TensorFlow / PyTorch — Deep learning frameworks for advanced neural network work

8. Key Takeaways for Beginners

Data Analytics — Beginner's Guide | Page 5

• Start with descriptive analytics — Excel and SQL will take you very far before you ever
need machine learning.
• Correlation does not equal causation — this is the single most important concept to
internalise early on.
• Clean, quality data matters more than a sophisticated algorithm. Garbage in, garbage
out.
• Python with pandas and scikit-learn is the most practical path into predictive modelling.
• Each level of analytics builds on the previous — master one before moving to the next.
• Real business value comes from taking action on insights, not just producing reports.

— End of Guide —

Data Analytics — Beginner's Guide | Page 6