Operations Research

Operations research (OR) is the discipline of using mathematics and systematic analysis to help people make better decisions about complex systems—how to route a fleet of trucks, how many beds a hospital should staff, where to place warehouses, which projects to fund.

17 frameworks, from 1937 to today · Industrial Engineering · Engineering

still a working view ended · each leads to its entry below

Overview

Operations research (OR) is the discipline of using mathematics and systematic analysis to help people make better decisions about complex systems—how to route a fleet of trucks, how many beds a hospital should staff, where to place warehouses, which projects to fund1. Its practitioners build models of a situation, solve them with algorithms, and use the answers to guide real choices1. What is at stake is not abstract elegance but the allocation of scarce resources: money, time, fuel, people, capacity. OR grew out of wartime necessity, matured into a mathematical science, and now reaches into logistics, health care, finance, energy, and public policy. Its history is a story of problems that resisted simple arithmetic and the ideas invented to crack them.

Wartime Origins and the Birth of a Discipline

The name and the field took shape during the Second World War, though its roots reach back further1. In Britain, the invention of radar in the late 1930s created a problem no single engineer could solve: how to site radar stations, how to filter their signals, and how to coordinate them with fighter squadrons. In 1937 the British physicist Robert Watson-Watt and colleagues had demonstrated radar's feasibility, and by 1940 a small group of scientists—physicists, physiologists, mathematicians—was working with the Royal Air Force on what they called "operational research": research on operations themselves, not on the equipment. They studied the size of convoys crossing the Atlantic, the depth settings of anti-submarine depth charges, and the scheduling of bomber missions. Their finding that larger convoys lost proportionally fewer ships to submarines, for instance, changed convoy policy.

The pattern was repeated across the Allied war effort. In the United States, the Navy and Army Air Forces assembled civilian scientists to analyze antisubmarine warfare, bombing accuracy, and logistics. The mathematician George Dantzig, then working for the U.S. Air Force, was later tasked with planning problems that would lead to his central contribution12. The war taught that teams of scientists from different backgrounds, working on real operations with data, could improve decisions that had previously been made by intuition and rank.

Linear Programming and the Simplex Method

The decisive mathematical breakthrough came just after the war. In 1947, Dantzig, working at the RAND Corporation and the Air Force, formulated the general linear programming problem: maximize or minimize a linear objective subject to linear constraints. The classic illustration is a diet problem—choose quantities of foods to meet nutritional requirements at minimum cost—but the same structure fit scheduling, blending, and transportation. Dantzig invented the simplex method, an algorithm that moves from one corner of the feasible region to another, improving the objective at each step, until no better corner exists2. It was fast, practical, and general.

Linear programming gave OR its first universal tool, and it spread quickly. The economist Tjalling Koopmans applied it to transportation and production planning, and Leonid Kantorovich in the Soviet Union had independently formulated related problems in 1939. The transportation problem—shipping goods from sources to destinations at least cost—became a standard model, and the Hungarian mathematician Harold Kuhn published an assignment algorithm in 1955. By the 1950s, linear programming was being taught in engineering and economics departments and used in oil refining, where blending crude oils into gasoline under quality constraints was a natural fit.

Networks, Integer Programming, and Combinatorial Structure

Many real problems are not smoothly divisible: you cannot ship half a truck or open two-thirds of a warehouse. Integer programming restricts some or all variables to whole numbers, and it proved far harder than linear programming. Ralph Gomory's cutting-plane method in the late 1950s gave a general algorithm, though early versions were slow3. The field of combinatorial optimization grew around such problems—scheduling, routing, packing—where the number of possible arrangements explodes4.

Network flow theory gave these problems a geometric and algebraic home5. In the 1950s, Lester Ford and Delbert Fulkerson developed the max-flow min-cut theorem, which states that the maximum amount of flow through a network equals the capacity of its narrowest cut. Their algorithms, and later refinements, made it possible to solve routing and distribution problems on graphs. The traveling salesman problem—find the shortest tour visiting a set of cities—became the emblem of combinatorial difficulty, and its study drove advances in exact methods and, later, heuristics6.

Uncertainty, Queues, and Stochastic Models

Real operations are full of randomness: arrivals at a service desk, machine breakdowns, demand fluctuations. Queuing theory addressed waiting lines7. The Danish engineer Agner Krarup Erlang had derived formulas for telephone exchange congestion as early as 1909, and his work was rediscovered and extended in the 1950s. The M/M/1 queue—Poisson arrivals, exponential service times, one server—became the textbook model, and networks of queues followed. Stochastic modeling more broadly, including Markov chains and Markov decision processes, gave OR a way to reason about systems that evolve probabilistically1. Monte Carlo methods, developed at Los Alamos in the 1940s by Stanislaw Ulam and John von Neumann, allowed analysts to simulate random processes numerically when closed-form solutions were impossible8.

Simulation and Systems Analysis

When models became too complex for analytical solution, discrete-event simulation stepped in. The idea is to represent a system as a sequence of events—arrivals, departures, failures—and run the clock forward, collecting statistics9. Early simulation languages appeared in the late 1950s and 1960s, and by the 1970s simulation was a standard tool for factory design, hospital planning, and military logistics. It let analysts test policies without disrupting real operations.

In parallel, systems analysis and the broader practice of OR matured into a profession. The RAND Corporation, founded in 1948, became a hub for applying OR to defense, budgeting, and policy. Decision analysis, developed largely by Howard Raiffa and Robert Schlaifer in the 1960s, brought subjective probability and utility theory into OR, giving decision-makers a formal way to weigh uncertain outcomes. The field's practitioners increasingly worked in teams, combining modeling with data collection and communication with managers.

Optimization Under Uncertainty and Robustness

Classical optimization assumed known data, but real data are uncertain. Stochastic programming, developed from the 1950s onward, optimized expected performance over scenarios. Robust optimization, formalized in the late 1990s and 2000s, instead sought solutions that perform acceptably under the worst plausible realization of uncertain parameters—a hedge against model error. These approaches changed how OR handled risk, moving from point estimates to ranges and distributions.

Heuristics, Metaheuristics, and Nature-Inspired Search

Exact algorithms guarantee optimality but can be too slow for large combinatorial problems. Heuristics—rules of thumb that find good but not provably optimal solutions—have always been part of OR. From the 1980s, metaheuristics generalized the idea: simulated annealing (inspired by cooling metals), genetic algorithms (inspired by evolution), tabu search, and ant colony optimization. These methods search vast solution spaces by combining exploration and exploitation. They are not guaranteed to find the best answer, but they often find excellent ones quickly, and they brought OR to problems—vehicle routing, scheduling, network design—that exact methods could not touch at realistic scale.

Learning-Integrated Optimization

The most recent turn integrates machine learning with optimization. Rather than assuming a model is given, analysts now use data to learn demand patterns, travel times, or failure rates, and then optimize against the learned model. This learning-integrated optimization—sometimes called prescriptive analytics—blurs the line between prediction and decision. It descends from the field's long tradition of combining statistical estimation with mathematical programming, but the scale of modern data and the flexibility of learning methods have opened new problems, from personalized pricing to real-time dispatch.

The Present

Operations research today is a broad, mature field. Linear and integer programming remain workhorses, solved by commercial and open-source solvers that handle millions of variables. Simulation, queuing theory, and decision analysis are standard in industry. Metaheuristics and learning-based methods address problems that resist exact solution. The field's questions are increasingly about integration: how to combine optimization with real-time data, how to make models robust to uncertainty and to human behavior, and how to deploy OR in settings—health care, sustainability, public policy—where objectives are multiple and contested. The wartime scientists who studied convoys would recognize the ambition, if not the tools.

References

  1. Operations research (introduction, Overview, Second World War, Historical origins), Wikipedia
  2. Simplex algorithm (History, introduction), Wikipedia
  3. Cutting-plane method (introduction, History), Wikipedia
  4. Combinatorial optimization, Wikipedia
  5. Network flow problem, Wikipedia
  6. Travelling salesman problem, Wikipedia
  7. Queueing theory, Wikipedia
  8. Monte Carlo method (introduction, History), Wikipedia
  9. Discrete-event simulation, Wikipedia

The frameworks of Operations Research

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1937-1960

    Operational Research

    How can science improve real military and industrial operations?

    Fly to its tower in Nousfarer

  2. 1937-Present

    Systems Analysis and Operational Research Practice

    How should analysts model and advise on complex operations?

    Fly to its tower in Nousfarer

  3. 1945-Present

    Monte Carlo method

    How can random processes be estimated numerically?

    Fly to its tower in Nousfarer

  4. 1947-Present

    Linear Programming

    How do we optimize a linear objective under linear constraints?

    Fly to its tower in Nousfarer

  5. 1948-1980

    Systems Analysis

    How can complex policy systems be analyzed as a whole?

    Fly to its tower in Nousfarer

  6. 1950-Present

    Queueing theory

    How long will queues and waiting times be?

    Fly to its tower in Nousfarer

  7. 1950-Present

    Stochastic Modeling and Markov Decision Processes

    How should decisions be made in randomly evolving systems?

    Fly to its tower in Nousfarer

  8. 1952-Present

    Dynamic Programming

    How do we solve multistage sequential decision problems?

    Fly to its tower in Nousfarer

  9. 1955-Present

    Network flow problem

    How much flow can a network carry?

    Fly to its tower in Nousfarer

  10. 1955-Present
  11. 1955-Present

    Stochastic Programming

    How do we optimize expected outcomes over uncertain scenarios?

    Fly to its tower in Nousfarer

  12. 1958-Present

    Combinatorial optimization

    How do we find the best arrangement among discrete choices?

    Fly to its tower in Nousfarer

  13. 1960-Present

    Decision Analysis

    How should a decision-maker weigh uncertain outcomes?

    Fly to its tower in Nousfarer

  14. 1960-Present

    Discrete-Event Simulation

    How can complex systems be simulated event by event?

    Fly to its tower in Nousfarer

  15. 1983-Present

    Heuristics and Metaheuristics

    How can good solutions be found quickly for hard problems?

    Fly to its tower in Nousfarer

  16. 1998-Present

    Robust Optimization

    How do we optimize against worst-case uncertainty?

    Fly to its tower in Nousfarer

  17. 2010-Present

    Learning-Integrated Optimization

    How can data and learning improve optimization decisions?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Operational ResearchSystems Analysis and Operational Rese…Monte Carlo methodLinear ProgrammingSystems AnalysisQueueing theoryStochastic Modeling and Markov Decisi…Dynamic ProgrammingNetwork flow problemOptimization Under Uncertainty and Ro…Stochastic ProgrammingCombinatorial optimizationDecision AnalysisDiscrete-Event SimulationHeuristics and MetaheuristicsRobust OptimizationLearning-Integrated Optimization

Beyond the field

Where Operations Research's ideas came from in other fields, and what they led to there: the strongest 12 of 76 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Operations Research is a planet in Nousfarer

Its city has a tower for each of these 17 frameworks. Land, walk in, and climb one concept at a time.

Land on Operations Research

Related fields

Fields its ideas are linked to

More in Industrial Engineering

See where Operations Research sits on the Map of Human Knowledge