How Market Professionals Can Self-Learn Operations Research?

Operations Research (OR) is a discipline that combines applied mathematics, statistics, computing, and analytical thinking to help organizations make better decisions. For professionals already working in the industry, mastering this field , even if only partially, can transform how production is planned, how resources are optimized, and how costs are reduced.

The good news: it is entirely possible to enter OR as a self-learner, even if you come from a non-technical background. This article presents a clear, practical, and realistic path, and also includes insights and advice from Arnaldo Gunzi, an experienced optimization professional with roles at companies such as Klabin, Stone, and Mercado Livre.

The Self-Learning Path: 5 Practical Steps

1. Start by Understanding the Fundamentals (Without Heavy Math)

Before diving into equations, understand the purpose of each area within OR.

Suggested introductory topics:

• Linear Programming (LP)

• Integer / Mixed-Integer Programming (ILP/MIP)

• Queueing Theory

• Machine Scheduling Problems

Initial (free) resources:

• Introduction to Operations Research — Hillier & Lieberman (introductory chapters)

• Coursera: Various Operations Research courses

Institutions:

• INFORMS

• Operational Research Society

2. Learn Modeling Before Algorithms

Professionals seeking optimization solutions don’t need to develop new algorithms—that’s for researchers.

Your focus should be on:

• Identifying the problem

• Formulating constraints (rules or limits the model must obey, such as machine capacity, available materials, maximum number of employees, total available time, etc.)

• Defining variables (elements the model can decide or adjust, such as how much of each product to produce, machine operating hours, staff allocation per shift, raw material purchases, etc.)

• Defining the objective (minimizing cost, maximizing production, etc.)

Typical example:
Minimize total production delay considering 3 machines and 12 orders.

3. Choose a Practical Tool

Today, you can use OR without being a coding expert.

Ideal tools for self-learners:

No-code

• Excel Solver (great for beginners)

• Harumi Platform (with the assisatnce of our AI for simple problems)

With code (for growing in the field)

• Python + PuLP

• Python + OR-Tools (Google)

• Python + Pyomo

• Harumi Platform (for more complex problems)

These tools allow you to:

• Maximize or minimize functions

• Solve scheduling problems

• Run simulations

• Test “what-if?” scenarios
  
  

4. Solve Real Problems in Your Company

This is where the magic happens.

Choose a small but real problem.

Typical suggestions:

• Optimizing the weekly production plan

• Defining the product mix that maximizes margin

• Reducing inventory while maintaining service level

• Scheduling shifts to minimize overtime

Start small → model → solve → compare with reality.
This creates powerful and immediate learning.

5. Build a Portfolio and Move to Deeper Theory

After solving 2–3 real world problems, move into more analytical topics:

• Simplex algorithm

• Branch and bound

• Metaheuristics (e.g., simulated annealing, genetic algorithms)

• Stochastic optimization

You don’t need to become a mathematician, but understanding what happens “behind the scenes” helps significantly.

Insights and Suggestions from Arnaldo Gunzi

Arnaldo Gunzi has a strong technical background in engineering and project management. He graduated from ITA in Aeronautical Infrastructure Engineering, completed a master’s degree in electronic engineering at UFRJ, and earned an MBA in Project Management from FGV.

Arnaldo has worked as Analytical Projects Coordinator at Klabin S.A.; as a Data Scientist Expert on the Shipping Optimization team at Mercado Livre; and now as a Senior Staff Data Scientist at Stone Co., always leading initiatives in Data Science, Optimization, and Operations Research applied to industry.

He also teaches statistics, optimization, and data science, helping train analysts and data professionals.

First, an observation: the world increasingly needs analytical solutions. Over the past decade and a half, we’ve had more and more sensors collecting data (telemetry, vibration sensors, temperature, electric current, etc.), and in the future even more data will be collected and stored. However, data by itself is useless if it doesn’t generate decisions, actions, problem prediction, cost reduction, or productivity gains.
Someone must handle this flood of data, process it, and generate useful decisions. Be the person who does that. It’s not as hard as it seems.

Below are some of his tips, in his own words:

1. Master the Foundations

Tools come and go, but fundamentals remain. Linear algebra, probability & statistics, and basic computing may seem boring, but they’re foundational to everything.
Once you understand the basics, move to OR topics such as Linear Programming, Integer Programming, Metaheuristics. Those who master the foundations have a superpower most people don’t—it's like having X-ray vision into the Matrix.

2. Solve a Practical Problem

Theory and practice are like wheels on an axle—if only one turns, you go in circles. Theory sharpens the axe; practice uses it. Look for an attackable problem at work.
Are inventories too high? Is production planning always struggling? Are processes unbalanced? Is there something you could improve with more optimization and technology? Even if it’s not your job to solve problems—just ask and take initiative. I’ve never seen a good manager reject a solution that reduces costs and increases productivity.

3. Simplicity

Between two solutions, choose the simpler one—Occam’s razor. Companies don’t need a super-LLM to solve their problems. Most problems are far simpler.
A linear regression may be enough. A pure linear formulation may capture 80% of the value. Simple descriptive statistics may provide context. Start simple, build bottom-up. I’ve never seen those “big consulting” PPT mega-solutions work in practice.

4. Understand the Process

Our work is technical, but paradoxically, the more we understand real processes, the better.
What does each column mean? Why are there outliers? Are the data trustworthy? What’s the relationship between features and the desired output?
As Judea Pearl said: data is dumb—it doesn’t see causality, only correlations. The one who orchestrates data and models is you.

Problems never come pre-formulated like in textbooks. And there is never a single “correct answer.” What you exclude from a model is often more important than what you include. Everything has a cost: an “optimal” solution might be expensive in licensing or computing time; a very simple solution may not capture all value. The real challenge is always formulating the problem, not solving it. That is the Art of Modeling.

5. A Title Doesn’t Mean Real Learning

A diploma is just paper if it doesn’t translate into real skills. Study deeply, concentrate, absorb the content. It’s not the teacher who must teach—it’s the student who must learn.

6. Networking

Connect with people in the field, and with their connections. Attend events like SOBRAPO and INFORMS.
Real “learning by osmosis” comes from interacting with high-level people. Something good will always come from these connections.

Master the techniques and the processes, and you’ll be unstoppable!

The Reward

Financially, in the long run, positions and salaries are proportional to your ability to generate real value.
Study hard, work hard, and take initiative to apply ideas in practice—society benefits from it.

But more than that, nothing beats seeing your work being used. I remember walking through the office and seeing someone I didn’t know using a tool I created.
That’s ideas taking shape in the world—changing processes, assisting decisions, and generating value!

As Morgan Housel wrote: “Be proud of what you create, not what you accumulate!”

Thanks to Miriam Koga for the invitation, and congratulations to the Harumi team for exploring such a noble, difficult, and important field as Operations Research.

Conclusion

Operations Research is not an academic-only field. On the contrary—it is born from real problems, and market professionals have a huge advantage because they already understand real-world bottlenecks and constraints.

With curiosity, accessible tools, and hands-on practice, anyone can master the fundamentals of OR and use them to transform their operations—or even transition into more analytical roles within their company.