Data Analytics in Operations Research: A Deep but Easy Explanation

Data Analytics in Operations Research (OR) means using data, statistics, and analytical methods to understand real-world problems, estimate important values, identify patterns, and provide useful inputs for Operations Research models and decision-making.

In simple words:

Data Analytics helps us understand what is happening in a system, while Operations Research helps us decide what should be done.

1. What is Data Analytics?

Data Analytics means studying collected data to find useful information, patterns, relationships, trends, and estimates that can help us understand a problem.

For example, a company may have five years of sales data. By analysing this data, we may find:

• How much was sold each day
• Which products sold the most
• Which months had high demand
• How much demand changes over time
• Whether sales are increasing or decreasing
• What future demand may look like

This is Data Analytics.

2. What is Operations Research?

Operations Research is concerned with using mathematical models, analytical methods, and computational techniques to support better decisions.

For example, suppose a factory has limited:

• Labour
• Machine hours
• Raw materials
• Money

The management may want to know:

How should these limited resources be allocated to maximize profit?

This is an Operations Research problem.

3. Data Analytics and Operations Research: The Basic Relationship

Suppose a factory wants to decide how many units of each product it should produce next month.

The factory may have data about:

• Historical demand
• Production costs
• Labour hours
• Machine capacity
• Raw material availability
• Inventory

First, the data can be analysed to understand demand, costs, variability, and trends.

Then Operations Research can use this information in a mathematical model to determine the best production plan.

The basic relationship can therefore be shown as:

Data → Data Analytics → OR Model → Better Decision

4. Why Does Operations Research Need Data?

An OR model requires inputs or parameters.

For example, a Linear Programming model may require:

• Profit per unit
• Cost per unit
• Resource requirements
• Available resources
• Demand
• Production capacity

In a real organisation, these numbers generally come from real-world information and data.

Therefore, data is often the foundation on which an OR model is developed.

5. Complete Data Analytics and OR Process

Real-World Problem

↓

Data Collection

↓

Data Cleaning

↓

Data Analysis

↓

Estimation / Forecasting

↓

Operations Research Model

↓

Optimization / Simulation / Decision Analysis

↓

Decision

↓

Implementation

↓

New Data

↓

Model Improvement

6. Descriptive Analytics in Operations Research

Descriptive Analytics mainly answers:

What happened?

Suppose a hospital has patient-arrival data.

Time	Average Arrivals per Hour
8–10 AM	15
10 AM–12 PM	30
12–2 PM	22
2–4 PM	12

The data shows that 10 AM–12 PM is the busiest period.

This is descriptive analysis.

But management may want to know:

How many doctors or service counters should be available during the busy period?

Now an OR problem begins.

Queuing Theory can be used to study:

• Arrival rate
• Service rate
• Number of servers
• Expected waiting time
• Queue length
• Server utilization

7. Predictive Analytics and Operations Research

Predictive Analytics asks:

What is likely to happen in the future?

Suppose a supermarket has historical sales data.

Analysis may show:

• Normal demand = 500 units per day
• Weekend demand = 700 units per day
• Festival-period demand = 1,200 units per day

A forecasting method may estimate future demand as:

Forecast Demand = 1,100 units per day

This is a prediction.

It does not automatically tell the supermarket how much inventory it should order.

That decision may require an OR inventory model.

8. Prediction Is Not the Same as Decision

This is one of the most important concepts in Data Analytics and OR.

Suppose an analytics model predicts:

Future Demand = 10,000 units

Does this automatically mean that the company should produce exactly 10,000 units?

No.

The company may have:

• Limited production capacity
• Limited raw materials
• Existing inventory
• Overtime options
• Outsourcing options
• Different production costs
• Uncertain future demand

Therefore:

Data Analytics: Demand is likely to be 10,000 units.

Operations Research: Considering demand, capacity, cost, inventory, and constraints, what is the best production decision?

9. Data Analytics and Linear Programming

Suppose a company produces two products.

Let:

x₁ = number of units of Product A

x₂ = number of units of Product B

Suppose data provides the following information:

• Profit from Product A = ₹40 per unit
• Profit from Product B = ₹30 per unit
• Labour requirements
• Machine requirements
• Available resources

An OR model may then have the objective:

Maximize Z = 40x₁ + 30x₂

subject to resource constraints.

Here, Data Analytics may help provide or estimate the model parameters, while Linear Programming determines the best values of the decision variables.

10. Data Analytics and Inventory Management

Inventory decisions often depend on demand information.

Suppose historical analysis gives:

Average daily demand = 500 units

Standard deviation = 80 units

These values describe both the average demand and its variability.

An inventory model can then use such information to help determine an appropriate inventory policy.

Possible decisions include:

• When should an order be placed?
• How much should be ordered?
• How much safety stock should be maintained?
• How can inventory cost be reduced?

11. Data Analytics and Queuing Theory

Consider a bank where customers arrive randomly.

Historical data can be used to estimate:

λ = customer arrival rate

μ = service rate

For example:

λ = 20 customers per hour

μ = 8 customers per hour per server

Queuing models can then help analyse:

• Expected waiting time
• Expected queue length
• Server utilization
• Probability of waiting

Management can then decide whether two counters are sufficient or whether additional counters are needed.

12. Data Analytics and Forecasting

Forecasting is particularly important when future demand or workload is uncertain.

Suppose historical demand is represented by:

D₁, D₂, D₃, ..., D₃₆

A forecasting method may produce future estimates:

D̂₁, D̂₂, ..., D̂₁₂

These forecasts can then be used as inputs to production planning, inventory planning, workforce planning, or other OR models.

The important point is that forecasting and optimization solve different parts of the decision problem.

13. Data Analytics and Transportation Problems

A Transportation Problem requires information about:

• Supply at each source
• Demand at each destination
• Transportation costs

Real-world transportation costs may depend on:

• Distance
• Fuel cost
• Vehicle type
• Historical transportation expenses
• Route conditions
• Delivery requirements

After obtaining realistic cost information, OR can determine the optimal shipment plan.

For example, the transportation objective may be written as:

Minimize Z = Σ Σ c_ijx_ij

where:

• c_ij = transportation cost from source i to destination j
• x_ij = quantity transported from source i to destination j

Data helps provide realistic values of c_ij, while the OR model determines the optimal values of x_ij.

14. Data Analytics and Simulation

Some real-world systems are too complex to represent easily with a simple mathematical model.

For example, an airport may want to know:

What will happen to passenger waiting time if one more security counter is added?

Historical data can help estimate:

• Passenger arrival patterns
• Service-time distributions
• Peak periods
• Variability in service time

These inputs can be used in a simulation model.

Different scenarios can then be compared:

• 4 counters
• 5 counters
• 6 counters

The analysis can compare:

• Average waiting time
• Queue length
• Resource utilization
• Operating cost

15. Data Analytics and Decision Analysis

Many OR decisions involve uncertainty.

Suppose a company is choosing between two machines.

Historical data may help estimate the probability of different future demand conditions.

For example:

P(High Demand) = 0.60

P(Low Demand) = 0.40

Decision-analysis methods can then be used to compare alternative decisions under uncertainty.

Thus:

Data Analytics → helps estimate probabilities and parameters

Decision Analysis → uses those estimates to evaluate alternatives

16. Data Analytics and Machine Learning

Modern Operations Research increasingly interacts with Machine Learning.

Suppose a factory wants to predict whether a machine will fail within the next 30 days.

Historical machine data may contain:

• Temperature
• Vibration
• Operating hours
• Previous failures

A machine-learning model can estimate the probability of failure.

But another question remains:

Should the company perform preventive maintenance now?

This is a decision problem.

The decision may depend on:

• Maintenance cost
• Failure cost
• Downtime
• Spare-parts availability
• Production loss

Thus:

Machine Learning → Prediction

Operations Research → Decision and Optimization

17. The Role of Data Quality

Good analysis requires good data.

A simple chain is:

Bad Data → Bad Analysis → Bad Model Inputs → Bad Decision

For example, suppose a delivery system accidentally records one delivery time as 500 hours instead of 5 hours.

This incorrect value can affect:

• Average delivery time
• Variance
• Forecasts
• Statistical models
• OR model parameters

Therefore, data collection and data cleaning are important parts of a reliable data-driven OR process.

18. Data Analytics Does Not Replace Operations Research

This distinction is very important.

Suppose a machine-learning model predicts future demand very accurately.

It still may not answer:

How should limited production capacity be allocated among 20 products?

That can be an optimization problem.

Similarly, an OR model cannot automatically know future demand if its demand estimates are poor.

Therefore, the two areas can complement each other:

Data + Statistics + Analytics + Prediction + OR + Optimization

→ Better Decision Support

19. Classical OR and Modern OR

It is important to understand that Data Analytics is not a completely new replacement for classical Operations Research.

Classical OR already uses areas such as:

• Probability
• Statistics
• Forecasting
• Data collection
• Estimation
• Simulation
• Decision analysis
• Sensitivity analysis

What has changed is the scale of available data and the computational methods used to analyse it.

Modern OR increasingly combines:

• Statistics
• Data Analytics
• Machine Learning
• Artificial Intelligence
• Optimization
• Simulation
• Computational methods

However, the central question of OR remains:

How can we make better decisions under constraints and uncertainty?

20. Data Analysis vs Data Analytics

Data Analysis	Data Analytics
Usually focuses on examining data	Broader concept involving analysis and decision support
Finds patterns and relationships	Can include descriptive, predictive, and prescriptive approaches
May focus on historical data	Can also involve prediction and future decision support
Uses statistical techniques	May combine statistics, computing, modelling, and optimization

21. A Complete Real-World Example

Consider an e-commerce company.

Step 1: Data Collection

The company collects:

• Order date
• Product
• Quantity
• Customer location
• Delivery time
• Price
• Returns
• Inventory level

Step 2: Data Cleaning

The company identifies:

• Duplicate orders
• Missing values
• Incorrect dates
• Impossible quantities

Step 3: Data Analysis

The company studies:

• Average demand
• Demand variability
• Seasonal patterns
• Regional demand
• Delivery patterns

Step 4: Forecasting

The company predicts future product demand.

Step 5: OR Modelling

The company decides:

• Inventory levels
• Warehouse allocation
• Shipment quantities
• Delivery routes
• Workforce requirements

Step 6: Optimization

The company may try to minimize:

Total Cost = Inventory Cost + Transportation Cost + Labour Cost

subject to demand, warehouse capacity, vehicle capacity, and workforce constraints.

Step 7: Implementation

The recommended plan is implemented.

Step 8: Feedback

New operational data is collected and compared with the model's results.

This creates a continuous improvement cycle:

Data → Analytics → Prediction → OR → Optimization → Decision → New Data

22. What Should a Student Learn?

A student who wants to study advanced Operations Research does not need to become a complete Data Scientist first.

A practical learning sequence is:

Stage 1: Data Fundamentals

• Types of data
• Variables
• Observations
• Population and sample
• Primary and secondary data
• Data collection
• Sampling
• Data cleaning

Stage 2: Statistics

• Mean
• Median
• Variance
• Standard deviation
• Probability
• Probability distributions
• Correlation
• Covariance
• Statistical inference

Stage 3: Data Analysis

• Descriptive analysis
• Exploratory Data Analysis
• Data visualization
• Regression
• Hypothesis testing
• Forecasting
• Time-series basics

Stage 4: Computing

A practical path is:

Excel → Python

Python can later be used for:

• Data cleaning
• Statistical analysis
• Visualization
• Forecasting
• Optimization

Stage 5: Operations Research

• Linear Programming
• Integer Programming
• Transportation Problems
• Assignment Problems
• Network Models
• Inventory Models
• Queuing Theory
• Simulation
• Decision Analysis
• Dynamic Programming
• Nonlinear Programming
• Stochastic Programming

Stage 6: Research-Level Integration

At the research level, the powerful combination is:

Statistics + Data Analytics + Operations Research + Optimization + Programming

23. Final Conclusion

Data Analytics in Operations Research is not simply about making charts or calculating averages.

It is about using data and analytical methods to understand a real system, estimate important parameters, identify patterns, make forecasts, represent uncertainty, and provide reliable inputs for Operations Research models.

Operations Research then uses mathematical modelling, optimization, simulation, and decision-analysis techniques to determine better decisions.

Data Analytics helps us understand and estimate what is happening in a real system.

Operations Research uses that information, together with mathematical models and constraints, to determine what decision should be made.

Therefore, the complete idea can be summarized as:

Data → Analysis → Estimation/Forecasting → OR Model → Optimization → Decision

References

1. Hamdy A. Taha, Operations Research: An Introduction, 11th Edition, Pearson.
2. Frederick S. Hillier and Gerald J. Lieberman, Introduction to Operations Research, McGraw Hill.