How to use this guide
This guide follows the course's published calendar description (data requirements for decisions, VBA-enabled spreadsheet models that optimize without and with uncertainty, and Monte Carlo simulation) and the standard topics for a prescriptive-analytics course. Instructors' emphasis and notation vary, so the scope below is a best fit. Check your own course outline.
What "prescriptive" means. Descriptive analytics says what happened, predictive analytics says what will probably happen, and prescriptive analytics says what to do: it picks the best decision from many allowed ones. The usual tool is an optimization model: decisions you control, a goal to maximize or minimize, and rules the decisions must obey. Then you solve it in a spreadsheet (Solver) or with a macro (VBA), read the report, and ask how much the answer depends on shaky data.
Topic order. Work the tabs left to right. 01 Data & formulation is the skill everything else rests on. 02 LP & sensitivity solves linear models and reads the Solver reports. 03 Network models covers transportation, assignment, transshipment, shortest path and maximum flow. 04 Integer & binary adds yes/no decisions and logic. 05 Nonlinear & goals covers curved models and several competing goals. 06 Uncertainty brings in probabilities and scenarios. 07 Monte Carlo simulates risk. 08 VBA models automates all of it.
Inside each topic: read the idea, do each worked example on paper before opening the solution (they run simplest first), then break the lab by changing numbers. Finish with the Drill. Introductory linear programming (the first graph, the first Solver run) is also covered in the MGST 391 guides; this guide starts from the model and goes further.
| Level | What it asks | Example |
|---|---|---|
| Beginner | One definition, or one formula in one step | Convert hours to minutes for a constraint; write "at most 2 of 5" as an inequality |
| Intermediate | A standard calculation with a few steps | Corner points of a two-variable model; expected profit from a demand table |
| Exam level | A full exam-style question with several parts | Formulate and read a Solver report; solve an unbalanced transportation problem |
| Hard | Twists: combining ideas, working backwards, traps | A change that leaves the shadow price's valid range; either-or constraints with big-M |
The drill starts at Beginner and steps up a level after three correct answers in a row; you can also pick a level yourself. The written questions, multiple choice and flashcards are ordered easy → hard, so working through them in order builds each idea from the ground up.
The core formulas
- Model = decision variables + objective + constraints (≤, ≥, =) + non-negativity
- Resource row: Σ (use per unit × units) ≤ available · ratio row: x ≥ p(x + y) → (1 − p)x − py ≥ 0
- Balance row: opening stock + made − sold = closing stock
- Shadow price: change in the optimal objective per +1 of a right-hand side, valid only within the allowable range
- Reduced cost = how much a zero variable's coefficient must improve before it enters the plan
- Network: outflow − inflow = supply (+), demand (−), or 0 for a transshipment node
- Logic: "if A then B" → A ≤ B · "at most k" → Σ ≤ k · fixed charge → x ≤ M·y
- Expected value = Σ probability × payoff · EVPI = expected value with perfect information − best expected value
- Critical ratio = Cu ÷ (Cu + Co) · trials needed n = (z·σ ÷ E)²
Rules people forget: memorize
- Put every row in one unit (minutes with minutes, kg with kg) before typing it into Solver
- Solving a model with fewer rows than the real world is not "simpler": it can give a plan you can't carry out
- A shadow price is per unit of the right-hand side and holds only inside the allowable increase and decrease
- A non-binding constraint has a shadow price of 0 and some slack
- Rounding an LP answer to whole numbers can be infeasible or not optimal: use Integer constraints
- An unbalanced transportation problem needs a dummy source or destination with zero cost
- Each binary needs a link: x ≤ M·y, otherwise the model can produce without paying the setup
- Simulation gives an estimate with sampling error; optimization gives the best answer for the numbers you entered
Exam format, weight and allowed aids differ between sections and terms. Check your own course outline.
Sign in to open the full guide
This is the first section. The full guide has every topic, the labs and the practice questions.
Create an account · Forgot password, or no password yet? Got a code link from your class? Open it and it will ask you to sign in.