The Three Types of Analytics
The Core
Type of
Question It Airline Industry Example
Analytics
Answers
Provides a view of historical data.
Descriptiv "What
e.g., "How many seats on flight
e happened?"
AB123 were sold last month?"
Uses historical data to forecast
future outcomes.
"What is likely
Predictive e.g., "Based on trends, how many
to happen?"
seats will we sell on that route
next month?"
Recommends actions to achieve
desired outcomes based on
"What should
Prescriptiv predictions.
we do about
e e.g., "Helps to fetch the best
it?"
pricing strategy across
flights."
Detailed Explanation of the Airline Example:
The example in the image perfectly illustrates how these analytics work
together:
1. Descriptive Analytics: The airline first analyzes its historical data:
how many tickets were sold for specific routes, at what prices, and
at what times before departure.
2. Predictive Analytics: Using machine learning models, the airline
predicts future demand for each flight. It forecasts how many people
will want to fly from New York to London on a specific date,
considering factors like season, day of the week, holidays, and even
current events.
3. Prescriptive Analytics: This is the most advanced stage. The
system doesn't just predict demand; it prescribes the optimal
action. It automatically recommends or sets the best price to
maximize revenue.
o If predictive analytics forecasts low demand, it might
prescribe lowering prices to fill more seats.
o If predictive analytics forecasts very high demand, it
might prescribe raising prices to maximize profit from the
limited number of seats.
Stages 1-4 are primarily used for Descriptive Analytics ("What
happened?").
Stages 1-4 plus advanced statistical modeling enable Predictive
Analytics ("What will happen?").
The entire pipeline, plus optimization algorithms,
feeds Prescriptive Analytics ("What should we do?").
This image illustrates the Analytics Process Flow — the step-by-step
journey of data from raw sources to insights and decision-making.
Let me break down each stage:
1. Data Sources (ERP, CRM, Excel, etc.)
Data originates from various internal and external systems.
Examples:
o ERP systems → financial transactions, supply chain data
o CRM tools → customer details, leads, sales pipeline
o Excel sheets → manually tracked data
o Other sources: APIs, sensors, web apps, social media, etc.
This is the raw data collection point.
2. Data Acquisition (ETL)
ETL = Extract, Transform, Load
Extract → Pull raw data from multiple sources.
Transform → Clean, standardize, and format data for consistency.
Load → Push the processed data into a storage system (data
warehouse or data lake).
Tools used: Talend, Informatica, Apache NiFi, Airflow, etc.
3. Data Storage (Data Warehousing)
A centralized storage system where cleaned and processed data
is stored for easy access.
Examples: Snowflake, Google BigQuery, Amazon Redshift,
Azure Synapse, Databricks.
Purpose: To maintain a single source of truth for analytics.
4. Data Analysis (BI – Business Intelligence)
Once stored, data is analyzed to derive actionable insights.
BI tools like Power BI, Tableau, Qlik Sense, Looker are used to
explore patterns, KPIs, and trends.
Includes techniques like:
o Descriptive analytics (what happened)
o Diagnostic analytics (why it happened)
o Predictive analytics (what will happen)
o Prescriptive analytics (what to do next)
5. Data Reporting & Visualization (Reports & Dashboards)
Final step → insights are presented visually for decision-making.
Dashboards, KPI scorecards, and reports are built using tools like
Power BI, Tableau, Google Data Studio, QlikView.
Enables business stakeholders to make data-driven decisions
quickly.
Example Use Case
Imagine a retail company:
1. Data Sources → POS transactions, website sales, customer CRM
data.
2. ETL → Extracts all data, cleans duplicates, standardizes currencies,
and loads it.
3. Data Warehouse → Stores unified data for analysis.
4. BI Analysis → Analysts identify best-selling products, regional
trends, and customer churn.
5. Visualization → Management views sales dashboards and decides
where to allocate marketing budgets.
Key Techniques in Advanced Statistical Modeling
These are some of the most common and powerful types of models:
Technique Primary Use Simple Example
Predicting: What will the price
To predict a
Regression of a house be? Based on its size,
continuous numeric
Analysis location, and number of
value.
bedrooms.
Predicting: Will this
Classification To predict a
customer churn (yes/no)? Is this
Models category or class.
email spam or not spam?
To predict future
Predicting: What will our sales
values based on
Time Series be next quarter? What
previously
Forecasting will energy demand be
observed values
tomorrow?
over time.
To discover hidden
Identifying: Grouping customers
groupings or
into distinct segments based on
Clustering segments in data
purchasing behavior for targeted
without predefined
marketing.
categories.
To analyse,
Natural
understand, and
Language
derive meaning
Processing
from human
(NLP)
language.
ML
RPA
