The Science of Better 30 Years Later… 📐👷🏻‍♀️ ⚙️

METU System Design Course Archive: What 527 Industrial Engineering Projects Reveal About the Future of AI

Non-members can read here. Also subscribe to stay updated whenever I publish…

“The Science of Better” is not a new idea for me. I first wrote about it on September 11, 2023, inspired by a phrase I had seen years earlier on a poster hanging in the corridors of the METU Industrial Engineering department.

Since then, the idea has evolved into a series of articles exploring a simple question: How do organizations systematically make better decisions?

One recent article examined the connection between Industrial Engineering and Value Engineering.

[embed]The Science of Better in Action ⚙️📈 What real industry projects taught me about how value is actually createdatabarezz.com

The premise remained simple. Every improvement initiative follows the same chain:

Decision → KPI → Financial Impact: Improve a decision. Move a KPI. Create value.

While revisiting that idea recently, I started exploring something much larger. A 30-year archive of 527 real industry projects conducted within the METU Industrial Engineering Systems Design course.

A Hidden Archive of Industry Problems 📚🏭

At METU Industrial Engineering, senior students work on real-world Systems Design projects with sponsoring companies for 2 semesters. It’s perhaps the closest thing to a consulting engagement that many students experience before graduation.

Over the years, these projects have involved manufacturers, retailers, banks, logistics providers, telecom operators, and public institutions.

I analyzed an archive containing 527 projects conducted between 1996 and 2024.

Initially, I thought I would find hundreds of different problems. Instead, I found the same problems repeating again and again, for nearly 30 years.

The Same Problems Keep Coming Back 🔄

The most common themes were:

Together, logistics, inventory management, scheduling, production planning, and decision support account for a substantial share of the entire archive. After analyzing 527 projects spanning nearly three decades, I expected to find hundreds of unrelated business problems. Instead, I found a surprisingly stable set of recurring operational bottlenecks.

At first glance this sounds disappointing. Shouldn’t industries evolve? Shouldn’t today’s challenges be completely different from those in the 1990s? Maybe not necessarily.

The deeper insight is that technology changes faster than operational reality. Factories still need schedules, warehouses still need inventory. Supply chains still need planning and all the managers still need decisions.

The problems stayed surprisingly constant but the tools changed.

Three Eras of Industrial Engineering ⏳

Looking at the archive, I see three distinct periods.

1996–2004 was the “The Optimization Era” 📊

This was classic Operations Research. Projects focused on: 📅 Scheduling, 📦 Inventory Control, 🏭 Production Planning, 📐 Facility Layout, ⚖️ Line Balancing. The goal was simple:

Find the optimal answer!

Data was limited. Computing power was expensive. Models were often built for a specific company and a specific problem.

2005–2014 was “The Analytics Era” 📈

ERP systems became widespread. Data started accumulating. Companies wanted visibility. Projects increasingly focused on 🧠 Decision Support Systems, 📊 Forecasting, 🚚 Supply Chain Planning, 🏗️ Warehouse Management, 📋 Performance Measurement…

The question changed. Instead of asking: “What is the optimal answer?” Companies began asking:

“What does the data tell us?”

2015–2024 was “The Intelligence Era 🤖”

This is where things became interesting. Logistics, inventory, planning, scheduling remained important.

But artificial intelligence, machine learning, cloud platforms, and real-time analytics started entering the picture.

The question evolved again:

“What decision should the system recommend?”

The Birth of AI-Native Operations 🚀

What fascinates me most is not what changed. It is what didn’t. The archive suggests that Turkish industry has spent three decades trying to solve five fundamental problems: 📅 Planning, 🚚 Logistics, 📦 Inventory ⏱️ Scheduling, 🧠 Decision Support

And these happen to be the exact areas now being transformed by:

📊 Digital Twins, 🔄 Autonomous Workflows, ⚡ Optimization Engines, 🧠 Large Language Models, 🤖 AI Agents

That is not a coincidence. Those recurring projects are signals. Signals of persistent economic bottlenecks. Signals of problems that software never fully solved. Signals of markets that may still be waiting for better answers.

Then If We Had A Startup Lens 🔍

Most people would look at 527 projects and see student work but I see something else. I see 527 validated business problems.

Each project existed because a company had enough pain to sponsor an external team to investigate it and that’s powerful.

If a problem appears repeatedly across industries and decades, it probably represents structural demand. And structural demand is where great companies are built.

Imagine being a founder in 2026;

Would you rather start with a new technology looking for a problem? Or start with a 30-year archive of problems companies have repeatedly paid to solve?

I know which one I would choose.

The Fourth Era 🌍

If the first era was Optimization and the second was Analytics and the third was Intelligence, then the next decade may belong to something else entirely. Decision Autonomy.

> 1998: A planner creates a schedule.

> 2008: A planner uses analytics to evaluate options.

> 2018: A machine learning model predicts demand.

>> 2028: An AI agent negotiates constraints, runs simulations, generates plans, monitors execution, and continuously adjusts decisions.

The human remains accountable but the system increasingly becomes a decision-making partner.

The Real Discovery 🎯

After reviewing nearly 30 years of projects, my conclusion is surprisingly simple. Industrial engineering was never really about formulas. It was never really about optimization. And now it is not really about AI.

It has always been about one thing: Helping people make better decisions.

The tools evolved, problems endured. And somewhere inside 527 projects, I suspect there are a few future startups waiting to be rediscovered.

That might be the most valuable lesson hidden in the archive.

All the best

Altan

P.s. 📂

If you’d like to explore the archive yourself, the complete dataset is publicly available on GitHub:

[embed]GitHub - aatabarezz/metu-ie-systems-design-archive Contribute to aatabarezz/metu-ie-systems-design-archive development by creating an account on GitHub.github.com

The repository contains:

• 527 Systems Design projects spanning 1996–2024
• Excel and CSV versions of the dataset
• Methodology documentation
• Data dictionary and usage examples
• Citation templates (APA, IEEE, BibTeX, Harvard, Chicago, etc.)
• Open CC-BY-4.0 license

The archive covers 42 distinct Industrial Engineering and Operations Research problem categories and includes estimates for economic impact, market potential, and software productization opportunities.

A note on methodology: project information quality varies significantly across the 28-year period. While many recent projects include abstracts, reports, posters, and richer documentation, some older projects, particularly before 2014, have limited publicly available information. In those cases, I supplemented the archive with reasonable estimates and inferred descriptions based on project titles, sponsoring organizations, industry context, and comparable projects. These enrichments should be viewed as directional approximations rather than historical fact.

If you discover interesting patterns, startup ideas, historical trends, or research opportunities, I’d be delighted to hear your findings…

AA