Oleg Korshikov, Director, Planning and Supply Chain Solutions at Jume

“Optimization” is one of the most frequently used terms in business. It can refer to anything from downsizing staff to switching software systems. Often, the term implies a pursuit of improvement or increased efficiency—but without a clear understanding of how or why. Some companies implement new accounting systems in hopes of bringing order to their operations. Others reduce headcount to cut costs. Still others simply revise their reporting methods and call it “digital transformation.”

This term is especially popular among IT companies that approach clients with proposals to “optimize” certain processes. As a result, many decisions made in the name of optimization are driven by intuition rather than a precise understanding of how these actions will actually lead to improvement.

Math or Intuition? A Real Tool vs. Abstract Efficiency

To bring optimization down from abstract heights and make it practical, one essential word must be added: mathematical. Mathematics turns vague optimization into a tangible, functional tool. We encounter it every day—for example, when using a navigation app to plan a route. A mathematical model selects the optimal path, minimizing travel time while considering traffic, toll roads, and user preferences. Not by guesswork or expert intuition, but through precise calculations.

Human judgment is prone to bias, narrow perspectives, and reliance on heuristics. Models, on the other hand, may offer unexpected—but always justifiable—solutions. Algorithms evaluate millions of combinations and identify the best option from among them all.

Logic or Data? Mathematical Optimization vs. Machine Learning

Mathematical optimization is often confused with machine learning (ML). While the two fields do intersect—especially in advanced software solutions—there is a key difference: ML relies on the quality and volume of training data, whereas optimization does not learn from the past. Instead, it explores possible scenarios to find the best one.

Optimization doesn’t require feeding historical data into a model. Instead, it depends on how well the business logic is translated into mathematical equations and constraints. This step is crucial to achieving accurate results. The good news: you don’t need to be a data scientist to use a well-built optimization tool. With the right interface and basic understanding, business users can use it effectively.

Still, the boundaries are blurring. Today, emerging systems allow users to adjust model parameters, explore constraints, and identify areas for potential improvement. This frontier overlaps with AI and ML, offering exciting possibilities for the future.

Human Involvement? How to Build an Algorithm

Mathematical optimization always begins with a clear definition of the problem: what variables can be changed, what constraints must be respected, and what should be minimized or maximized. Formalization is essential. To move from an idea to a working solution, three key steps are required:

1. Build an optimization model – Translate all business rules, constraints, and requirements into a system of mathematical equations.
   
2. Connect a solver – This is a specialized algorithm that efficiently evaluates millions of combinations and calculates the optimal solution.
   
3. Create an interface – This could be a full-fledged IT platform or a simple tool for uploading data and launching calculations.
   

Steps one and three can be handled internally or with the help of a specialized IT platform. The solver, however, is typically sourced externally due to its technical specificity. There are several leading solver providers which are used in major IBP platforms, with Gurobi being acknowledged as the most reliable and best-performing solver on the market. 

As for the common question posed by AI skeptics—“Do we even need humans in this system?”—in this context, the answer is clear: absolutely yes. Humans configure the models, define the logic, analyze results, and decide which assumptions need revisiting. Optimization doesn’t replace human thinking—it enhances analytical capability.

Cutting Queues and Milk Costs: How It Works in Practice

Disney uses optimization to reduce guest wait times in its theme parks, allowing visitors to spend more time enjoying attractions. Before their visit, guests can enter their preferences into the Disney Genie app—choosing rides, shows, and restaurants. The algorithm generates an optimized itinerary that minimizes waiting time while considering event schedules.

Another classic example is UPS, the courier delivery service. Their algorithm altered delivery routes to reduce left-hand turns, which tend to increase delays, fuel consumption, and accident risk. This simple adjustment helped cut total mileage by tens of millions of kilometers, reducing both costs and CO₂ emissions.

One of our customers, a leading dairy product producer, used algorithms to optimize raw material procurement, production planning, and product distribution across its logistics network. The tool evaluates millions of scenarios to identify the best solutions, factoring in various supplier prices, volumes, delivery times, and milk characteristics, hundreds of production routes, and countless delivery options. Within months, this approach reduced operating costs by 2–4% and milk costs by 1%—a major gain, especially given milk’s role as a key cost component. In absolute terms, these savings add up to millions of dollars.

Changing the Logic, Not Just Adding a Module

A common misconception is that optimization and IT implementation are essentially the same thing. It’s understandable why this confusion arises: both aim to improve efficiency. However, IT achieves this through technology, while optimization does so through mathematical methods.

When implementing an IT system, it’s difficult to measure in advance how much faster processes will run or how much fewer mistakes will occur. Optimization, in contrast, offers a scientific, measurable approach. It’s formalized, verifiable, and repeatable. Unlike intuitive decisions, which remain speculative, optimization delivers provable and reproducible results.

Optimization is not about adding another module—it’s about changing the decision-making logic. And this new logic is transforming industries. Manufacturers no longer overstock warehouses “just in case.” Finance teams move away from guesswork when allocating budgets. Planners escape the chaos of manual schedules. Instead of working with assumptions, they rely on calculations. Professionals no longer spend hours manually tweaking Excel scenarios. Their job is to compare solutions, test hypotheses, and take informed decisions.

From Theory to Tool: Why Optimization Is Taking Off Now

Although mathematical optimization as a discipline has been around since the mid-20th century, its widespread practical application is only beginning now. In the past, even basic models required too much computing power to deliver timely results. Today, with modern processing capabilities and accessible solutions, optimization has moved from academic theory to a powerful everyday tool.

It’s a tool that justifies decisions with evidence, not just belief. So, next time consultants suggest “optimizing” your business, ask them: How exactly will you do that? What evidence supports your method?
The answers will help you separate genuine methodology from mere rhetoric—and avoid automating for automation’s sake.