Operations Research

Course Description

Operations Research develops the mathematical techniques used to model a system and then find the best decision within it — linear programming, network and transportation models, integer programming, and the modelling discipline that turns a messy operational problem into something solvable.

Within the SCNS taxonomy, ESI is the Industrial and Systems Engineering prefix, and the 4000-level number places this in the upper division. Daytona State publishes it at 3 credits, prerequisites MAC2311C and EGN3443, offered in spring, and describes it as covering "the basic techniques and methods for modeling and optimizing systems" with emphasis on production, logistics, and service operations.

This is the course where mathematics becomes decision-making. The characteristic operations research question is not "what will happen" but "what should we do" — how much to produce, what to ship where, which projects to fund, how to staff a schedule — and the discipline's contribution is that a surprisingly large share of those questions have a provably best answer that human judgement alone will not find.

⚠ Number and title divergence: this one is unusually messy across Florida

Worth checking before you assume anything transfers. The same subject appears under different numbers and markedly different titles at Florida institutions:

Institution	Number	Title
Daytona State	ESI4312	Operations Research
UCF	ESI4312	Deterministic Methods for Operations Research
USF	ESI4312	Foundations of Optimization (formerly Deterministic Operations Research)
University of Florida	ESI3312 / ESI4313	Operations Research 1 / Operations Research 2

Two things follow. UF does not carry ESI4312 at all — it splits the material into a 3000-level deterministic course and a 4000-level stochastic course, so a student moving between UF and another Florida institution has a genuine mapping problem in both directions. And the USF and UCF titles tell you something useful: this course is normally the deterministic half. Stochastic methods — queueing, Markov chains, simulation — usually live in a separate course. Read the catalog description rather than the title.

Learning Outcomes

Required Outcomes

• Describe the scope and method of operations research and its role in engineering decision-making.
• Translate a verbal description of an operational problem into a mathematical optimization model.
• Identify decision variables, an objective function, and constraints, and state the modelling assumptions.
• Formulate linear programming models for production, blending, allocation, and scheduling problems.
• Solve linear programs graphically for two-variable problems and interpret the geometry.
• Apply the simplex method and describe how it moves between basic feasible solutions.
• Recognize and interpret unbounded, infeasible, and degenerate outcomes.
• Perform and interpret sensitivity analysis, including shadow prices and ranging.
• Describe duality and interpret the dual problem in the terms of the original application.
• Formulate and solve transportation and assignment problems.
• Formulate and solve network models, including shortest path and maximum flow.
• Apply project network methods (CPM/PERT) to schedule and identify a critical path.
• Formulate integer and binary programming models for problems with indivisible or logical decisions.
• Describe branch-and-bound and why integer problems are harder than their linear relaxations.
• Solve models using appropriate software and interpret solver output correctly.
• Validate a model against reality and identify where its assumptions break.
• Communicate an optimization result to a non-technical decision maker.

Optional Outcomes

• Apply dynamic programming to multistage decision problems.
• Formulate goal programming models with multiple objectives.
• Describe non-linear optimization and where it is required.
• Describe queueing models and their engineering applications.
• Describe simulation as an alternative when analytical models fail.
• Apply optimization to a supply chain or logistics case study.
• Describe heuristics and metaheuristics for large combinatorial problems.

Major Topics

Required Topics

• Scope and method of operations research
• Model formulation: variables, objective, constraints, assumptions
• Linear programming formulation
• Graphical solution and its geometry
• The simplex method
• Special cases: unbounded, infeasible, degenerate, alternate optima
• Sensitivity analysis, shadow prices, and ranging
• Duality and economic interpretation
• Transportation and assignment problems
• Network models: shortest path, maximum flow, minimum spanning tree
• Project scheduling: CPM and PERT
• Integer and binary programming
• Branch-and-bound and computational difficulty
• Optimization software and solver output
• Model validation and assumption failure
• Communicating results to decision makers

Optional Topics

• Dynamic programming
• Goal and multi-objective programming
• Non-linear optimization
• Queueing theory
• Simulation
• Supply chain and logistics applications
• Heuristics and metaheuristics

Resources & Tools

• Introduction to Operations Research (Hillier & Lieberman) — the standard text and the one most programmes adopt.
• Operations Research: An Introduction (Taha) — the common alternative, strong on worked algorithm mechanics.
• Introduction to Linear Optimization (Bertsimas & Tsitsiklis) — for students who want the theory properly.
• Excel Solver — already on your machine, and genuinely used in industry for small and medium models. Learn to read its sensitivity report, not just its answer.
• Python with PuLP or Pyomo — free, open-source modelling libraries; the standard way this work is done now, and worth learning even if the course uses spreadsheets.
• Google OR-Tools — free, fast, and excellent for routing and scheduling problems.
• GLPK, CBC, and HiGHS — free open-source solvers. Gurobi and CPLEX are the commercial standards and both offer free academic licences — get one while you are a student.
• INFORMS (informs.org) — the professional society; its student membership is inexpensive and its case material is a good source of realistic problems.
• NEOS Server (neos-server.org) — free access to commercial-grade solvers over the web.

Career Pathways

• Operations research analyst — the direct title; consistently one of the faster-growing occupations, and it pays well.
• Industrial engineer — process improvement, capacity planning, and facility design across manufacturing and services.
• Supply chain and logistics analyst — network design, inventory policy, and routing; a very large Florida employer base given the state's ports and distribution centres.
• Data scientist and analytics roles — optimization is the part of analytics that recommends an action rather than describing a pattern, and that distinction is increasingly what employers pay for.
• Airline, cruise, and hospitality operations — crew scheduling, revenue management, and fleet assignment are classic OR problems, and Florida has an unusual concentration of all three industries.
• Healthcare operations — capacity, staffing, and patient flow; hospital systems increasingly staff these roles.
• Transportation and public sector — FDOT, transit agencies, and port authorities all model network and scheduling problems.
• Defence and aerospace — the field's origin, and still a major employer; Florida's Space Coast concentration is relevant.
• Graduate study — ESI4312 is the standard entry point to an M.S. in industrial engineering, operations research, or analytics.
• SOC codes 15-2031 Operations Research Analysts and 17-2112 Industrial Engineers.

Special Information

⚠ Formulation is the hard part, and it is the part that is graded lightly and matters most

The single most useful thing to understand about this course.

Solving a linear program is a solved problem. Software does it instantly, correctly, and at a scale no human could approach. What software cannot do is decide what the model should be — which quantities are decisions, which relationships are constraints, what is actually being optimized, and what has been assumed away. That translation from situation to model is where operations research is difficult and where practitioners earn their living.

What good formulation practice looks like:

• Define decision variables precisely, with units. "x = production" is not a variable; "x_i = units of product i produced per week" is. Most formulation errors are traceable to vague variable definitions.
• Write the objective in the same units throughout. Mixing weekly and annual figures, or cost and profit, is the second most common error.
• State every constraint's meaning in words alongside the algebra. A constraint you cannot describe in a sentence is one you do not understand.
• Check that the model would reject an obviously stupid solution. If nothing in the model prevents negative production or infinite capacity, something is missing.
• Write the assumptions down. Linearity, divisibility, certainty, and additivity are assumed by every LP, and each is false in some applications.

The examination consequence: the difficult part of an OR exam question is usually the first ten lines, not the arithmetic that follows.

⚠ Sensitivity analysis is what the client actually asks for

Students underestimate this section reliably, and practitioners rate it as the most-used output of the whole subject.

An optimal solution answers a question nobody asked. Real decision makers rarely want "produce 412 units"; they want to know how much it is worth to relax a constraint, how far a cost can move before the plan changes, and which limits are actually binding. Sensitivity analysis answers exactly those, and it comes free with the solution.

• A shadow price is the value of one more unit of a scarce resource — which converts directly into "should we buy another machine, hire another shift, lease more space." It is the number a manager acts on.
• A non-binding constraint has a shadow price of zero, and knowing which constraints are slack tells you where not to spend money.
• Ranging tells you how wrong your data can be before the recommendation changes. Since input data is always somewhat wrong, this is what makes a recommendation defensible.
• Shadow prices are valid only within their range. Extrapolating one beyond its range is a common and expensive error.

The professional habit: read the sensitivity report before reporting the answer. A solution presented without it invites the first question you will be asked.

⚠ The model is not the system — say so out loud

The intellectual honesty this discipline requires, and the failure mode that makes optimization dangerous rather than merely wrong.

Every optimization model rests on assumptions that are false in detail: that relationships are linear, that data is known with certainty, that the objective captures what is actually wanted, that everything relevant has been included. The model is a deliberate simplification, and its recommendations are only as good as the simplification.

The specific hazards:

• Optimizing the stated objective exactly is dangerous when the stated objective is incomplete. A model that minimizes cost will happily recommend a plan that is fragile, unsafe, or unacceptable to the workforce, because none of those appeared as constraints. This is a design failure, not a solver failure.
• Optimal solutions sit at corners — they push resources to their limits by construction, which is precisely what leaves no slack when reality deviates. A visibly optimal plan is frequently a brittle plan, and robustness is worth buying explicitly.
• Garbage in, precisely optimal garbage out. The solver's four-decimal answer confers no accuracy on the input estimates, and presenting it that way misleads.
• A model that contradicts everyone's experience is usually wrong, and the professional response is to investigate the discrepancy rather than to insist on the mathematics.

The stance worth carrying: present a recommendation with its assumptions attached, and be the person who says which parts of the answer they would not bet on.

⚠ Learn a solver library, not just the spreadsheet

Practical career advice, because the gap between what the course teaches and what the job requires is narrow and easy to close.

Excel Solver is an excellent teaching tool and is genuinely used for small models. It is also size-limited, hard to version-control, and impossible to embed in a production system. Industry runs optimization from code.

What to do about it while you are still a student: rebuild one course assignment in Python with PuLP or Pyomo, which takes an afternoon once and is thereafter permanent; claim the free academic licences for Gurobi or CPLEX, which are the commercial standards and are otherwise expensive; and look at OR-Tools for routing and scheduling, where it is very strong. A graduate who can say "I have formulated and solved this in Pyomo against a real solver" is describing the actual job.

⚠ Only about three Florida institutions carry this number — hedge accordingly

This course appears at roughly three institutions statewide. Content, credit value, and emphasis vary more than they would for a widely taught course. Read your own institution's catalog description and syllabus rather than assuming this guide describes your section exactly, and have any transfer evaluated in writing.

Course format and transfer

ESI4312 is a lecture course, 3 credits and approximately 45 contact hours. Expect problem-set-heavy assessment with formulation and computation weighted together, software-based assignments, and in many sections a modelling project on a real or realistic system. The prerequisites are not decorative — calculus and probability are both used, and comfort with matrix notation and linear algebra makes the simplex material substantially easier.

How Florida course levels affect transfer

The first digit of an SCNS number denotes the year of offering, not transferability. Courses at the 1000 and 2000 levels transfer transparently between Florida public institutions, and 3000 to 4000 is unproblematic since both are upper division. The boundary that actually matters is 2000 to 3000, where lower-division credit generally cannot satisfy an upper-division requirement.

ESI4312 is upper division and typically sits in the third or fourth year of an engineering or engineering technology baccalaureate. Given the numbering divergence documented above — UF splits the material into ESI3312 and ESI4313 — do not assume any equivalence across institutions without a written evaluation. Students arriving from an A.S. should also note that A.S. degrees are applied and do not carry the A.A.'s guaranteed junior-status transfer.