MGMT20005 Business Decision Analysis
The Complete Exam Bible · S2 2026MGMT20005 Overview
- Level 2 management subject
- Semester 2, 2026
- Quantitative decision modelling
- Open-book final examination
- Two double-sided A4 pages
- Scientific calculator permitted
MGMT20005 Business Decision Analysis is a quantitative Level 2 management subject at the University of Melbourne.
- Assessed by An individual assignment, a group assignment and an end-of-semester examination weighted 25%, 25% and 50%.
- Current exam Two hours plus fifteen minutes reading time, open book with two double-sided A4 pages and a scientific calculator, but no Excel.
- Hardest step Choosing the correct model before calculating: objective direction, probability structure and constraint meaning control the arithmetic.
- How to prepare Rebuild lecture and tutorial methods with new numbers, then finish every result with a plain-language recommendation.
How MGMT20005 is assessed
| Component | Weight | Format |
|---|---|---|
| Individual assignment 1500 words (equivalent) | 25% | From Week 6 to Week 7. Current subject materials describe a take-home case, an invigilated in-class Canvas quiz and a reflection that together form the assignment. |
| Group assignment (groups of 3 students) 2500 words | 25% | Week 11. The official handbook and PDF guide publish this format; the current Canvas task page describes a different operational submission shell, so follow the live task brief. |
| End-of-semester examination 2 hours | 50% | During the examination period, plus 15 minutes reading time. On campus and open book with two double-sided A4 pages; Casio FX82 calculator allowed; Excel unavailable. |
The official 2026 handbook publishes three components weighted 25%, 25% and 50%, summing exactly to 100%. The examination is explicitly not a hurdle. No positive hurdle statement is published for either assignment, so their rows direct students to the current task page instead of asserting an absence. Assignment 2 has a source conflict: the handbook and PDF subject guide describe a 2,500-word group assignment, while the current Canvas assessment page describes individual or group work submitted as slides, a recorded presentation and an Excel workbook. The live task shell controls submission.
Current MGMT20005 dates
| Date | Item | Control |
|---|---|---|
| 2026-09-01 | Assignment 1 Part 1 | Take-home component due at 9 am according to the current subject calendar. |
| 2026-09-07 | Assignment 1 Part 2 | Invigilated in-class component during the enrolled Monday lecture, shown as 11:30 am in the current calendar. |
Dates are as published in the Semester 2, 2026 Canvas subject calendar and assessment information page. Confirm exact deadlines and submission settings in the live LMS.
What MGMT20005 covers
Follow the decision sequence from payoff tables and probabilities through information, multi-criteria analysis, optimisation, sensitivity and business applications.
Decision Analysis Without Probabilities
Week 1. Objectives, alternatives, states of nature, consequences, payoff tables, maximax, maximin and minimax regret for profit and cost decisions.Week 1 builds the decision before it applies a rule. A complete model names the objective, the alternatives under management control, the uncertain states of nature, and one consequence for every alternative-state combination.
- Objective, alternatives, states of nature and consequences
- Influence diagrams, payoff tables and decision trees
- Optimistic maximax decisions for profit tables
- Conservative maximin decisions for profit tables
- Opportunity-loss matrices and minimax regret
- Direction changes when a table contains costs
Decision Trees and Expected Value
Week 2. Decision and chance nodes, complete probability branches, terminal payoffs, expected value, rollback and risk profiles.Week 2 adds probabilities and time to the decision model. Squares represent decision nodes, circles represent chance nodes, and terminal points carry complete path payoffs. The tree is read chronologically from left to right but solved from right to left.
- Decision nodes, chance nodes, branches and terminal payoffs
- Mutually exclusive and collectively exhaustive chance outcomes
- Expected value as a probability-weighted average
- Right-to-left rollback through a decision tree
- Conditional path probabilities in staged decisions
- Risk profiles and the limits of expected value
Information Value and Bayesian Updating
Week 3. Perfect and sample information, EVPI, EVSI, prior and posterior probabilities, Bayes tables and contingent strategies.Week 3 asks whether uncertainty is worth reducing. Expected value with perfect information allows a different best action in each state and creates an upper bound on what state-revealing information can be worth.
- Expected value without information as the baseline
- Expected value with perfect information and EVPI
- Prior, conditional, joint, marginal and posterior probabilities
- Bayes tables for favourable and unfavourable reports
- Contingent strategies after sample information
- EVSI, information cost and the perfect-information ceiling
Scoring Models and AHP
Week 4. Weighted scoring, pairwise comparison matrices, reciprocal judgments, priority vectors, consistency ratios and overall ranking.Week 4 handles choices that depend on several conflicting criteria. A direct scoring model assigns each criterion a weight, rates each alternative, multiplies weight by rating and sums the contributions.
- Weighted scoring with separate weights and ratings
- AHP goals, criteria and alternatives
- Reciprocal pairwise comparison matrices
- Column normalisation and row-average priorities
- Lambda max, consistency index and consistency ratio
- Overall priorities and defensible recommendations
Linear Programming Models
Week 5. Decision variables, objectives, constraints, domains, feasible regions, graphical solution, corner points and Solver setup.Week 5 translates a business decision into a linear optimisation model. Decision variables name controllable quantities with units and a time basis. The objective is a linear expression to maximise or minimise.
- Decision variables with units and time periods
- Linear objective functions and fixed coefficients
- Capacity, demand, policy and balance constraints
- Non-negativity and other domain restrictions
- Feasible regions and extreme-point solutions
- Binding constraints, slack, surplus and special cases
LP Sensitivity Analysis
Week 6. Objective coefficient ranges, RHS ranges, reduced cost, shadow price, binding constraints and local what-if analysis.Week 6 interprets how a solved LP responds to local changes. Variable-cell rows in the Solver report contain final values, objective coefficients, allowable increases and decreases, and reduced costs.
- Variable and constraint sections of a Solver report
- Objective-coefficient ranges of optimality
- Reduced cost for zero-valued variables
- Shadow prices for constraint RHS changes
- RHS ranges of feasibility
- One-at-a-time limits and re-solving triggers
Business Applications of LP
Week 7. Production planning, inventory flow, portfolio maturities, media allocation, balance equations and economic interpretation.Week 7 reuses LP anatomy in three business settings. Operations models allocate production, workforce, inventory and capacity.
- Production scheduling and product-mix applications
- Inventory balances across a planning horizon
- Workforce and capacity allocation
- Portfolio cash flows and maturity nodes
- Media allocation, budgets and mix restrictions
- Binding constraints and economic interpretation
Its aim is practical: formulate a business decision, select a suitable model, calculate or optimise the result, test the model's sensitivity, and explain the recommended action in language that a manager can use.
The current subject guide names decision trees, optimisation and simulation as core techniques, with applications across entrepreneurship, strategy, operations, marketing and finance.
The first part of the subject asks what to do when future conditions are uncertain.
Week 1 separates objectives, alternatives, states of nature and consequences, then compares optimistic, conservative and opportunity-loss rules when probabilities are unavailable. Week 2 adds probabilities and decision trees. The calculation moves from terminal payoffs to chance-node expected values and then to the best decision branch. Week 3 asks whether information is worth buying.
Students calculate the expected value of perfect or sample information, use Bayes' theorem to turn prior probabilities into posteriors, and build a contingent strategy that can change after a report is observed.
Week 4 introduces decisions with several conflicting criteria. A direct scoring model combines criterion weights and alternative ratings.
The analytic hierarchy process instead derives priorities from reciprocal pairwise judgments, checks whether those judgments are consistent, and combines local priorities into an overall ranking.
Week 5 moves into linear programming: define decision variables with units, write the objective, translate resources and policies into linear constraints, add domain restrictions, and solve a two-variable model graphically or a larger one with Solver.
Week 6 reads the Answer and Sensitivity Reports, distinguishing objective-coefficient ranges from right-hand-side ranges, reduced cost from shadow price, and a local what-if result from a change that requires the model to be solved again.
Week 7 applies linear programming to production, workforce and inventory planning, portfolio cash flows and media selection.
These cases share a conservation logic: product, cash or budget flows into a node, is allocated through decisions, and leaves through demand, maturity or channel use. The official schedule continues with Network Models, Integer Programming and Applications, Monte Carlo Simulation and Simulation-Based Optimisation before revision.
Current teaching modules for those later weeks were not present in the material used for this edition, so this guide does not invent their lecture-specific formulas.
Historical papers confirm binary-formulation and discrete-simulation question types; the practice section uses new scenarios to rehearse only those shapes.
The final examination carries 50% and is described as on campus, open book and designed for two hours plus fifteen minutes reading time. Two double-sided A4 pages and a Casio FX82 calculator are permitted; Excel is not.
The subject materials say questions involve modelling or formulation, interpretation and calculations and resemble examples worked in lectures and tutorials.
That makes method visibility essential: label the tree, show the probability update, define each variable, write every constraint and state the validity range before interpreting a sensitivity result.
MGMT20005 cheatsheet for the final exam
AskSia publishes a free MGMT20005 cheatsheet for the final exam, a cheat sheet that follows the official permitted-materials rule for this assessment and condenses the topics in this guide.
Worked example · free
Update demand risk after a warning signal
- 2Compute the volatile-and-Warning joint probability: 0.30 times 0.80 equals 0.24.
- 2Compute the stable-and-Warning joint probability: 0.70 times 0.20 equals 0.14.
- 2Find the Warning probability: 0.24 plus 0.14 equals 0.38. Divide each joint by 0.38, giving 0.6316 volatile and 0.3684 stable.
- 2Interpret the result: the warning raises the probability of volatile demand from 0.30 to 0.6316, but it does not establish volatility with certainty.
Key terms
- State of nature
- An uncertain condition outside the decision maker's control that affects the consequence of an alternative.
- Payoff table
- A matrix crossing decision alternatives with states of nature and showing one consequence for every combination.
- Opportunity loss
- The gap between an alternative's payoff and the best payoff available in the same state of nature.
- Expected value
- A probability-weighted average of possible payoffs used to compare alternatives under risk.
- Risk profile
- The probability distribution over payoffs generated by a chosen decision strategy.
- Posterior probability
- A state probability updated after observing information, calculated from priors and conditional report accuracy.
- EVPI
- Expected value of perfect information, the maximum expected benefit from learning the uncertain state before acting.
- Priority vector
- A set of AHP weights derived from a normalised pairwise comparison matrix and summing to one.
- Feasible region
- The set of decision-variable values that satisfy every model constraint and domain restriction simultaneously.
- Shadow price
- The local change in optimal objective value from a one-unit increase in a constraint right-hand side within its allowable range.
- Reduced cost
- The objective-coefficient improvement required before a zero-valued variable can enter the optimal solution.
MGMT20005 FAQ
How is the subject assessed in Semester 2, 2026?
The official 2026 handbook lists an individual assignment worth 25%, a group assignment worth 25% and an end-of-semester examination worth 50%. The three rows total 100%. Current Canvas materials split Assignment 1 into a take-home case, an invigilated in-class quiz and a reflection, while Assignment 2 has a format conflict between the handbook and the live task page. Follow the current Canvas submission shell.
What are the current final examination conditions?
The current subject materials describe an on-campus, open-book examination designed for two hours plus fifteen minutes reading time. You may bring two double-sided A4 pages and a Casio FX82 scientific calculator. Excel is not allowed. The exam is explicitly not a hurdle, and its exact date and time were still to be advised in the captured subject guide.
Which topics does the subject cover before the final?
The published schedule moves from decision analysis, decisions with probabilities, information value and multi-criteria decision making into linear programming, sensitivity analysis, LP applications, network models, integer programming, Monte Carlo simulation and simulation-based optimisation, followed by revision.
This edition covers the richly sourced Week 1 to Week 7 methods and flags the later current-module gap rather than inventing lecture-specific formulas.
How quantitative is Business Decision Analysis?
It is a genuinely quantitative subject. Students calculate expected values and posterior probabilities, construct weighted scores and AHP priorities, formulate optimisation models, solve two-variable LPs, and interpret Solver reports. The arithmetic is only part of the task: the subject also expects students to interpret solutions, test sensitivity and communicate recommendations in plain language.
What do historical papers show about question types?
Captured papers from 2017, 2019 and 2023 show recurring families including decision trees, EVPI, Bayes updating, AHP, LP formulation, graphical solution, sensitivity reports, binary or fixed-charge models and discrete simulation. They do not prove the 2026 section order or marks. The current materials only say that the exam resembles lecture and tutorial examples and that a sample paper will be provided later.
What belongs on the permitted examination notes?
Use the two double-sided A4 pages as a routing tool. Include decision-rule directions, the Bayes table order, AHP consistency steps, LP formulation checks, sensitivity ranges, reduced cost and shadow-price distinctions, and flow-balance templates. Avoid copying long worked scenarios. Notes should help you recognise and rebuild a method under time, because Excel is unavailable.
Where are marks usually won in a model answer?
The visible requirements in historical questions and current teaching materials reward labelled models, shown calculations, explicit variable definitions, complete constraints, interpretation and a justified recommendation. For sensitivity, test the relevant allowable range before applying a coefficient or shadow price.
For probability questions, show the joint and marginal values before the posterior so the direction can be checked.
Questions students ask in each MGMT20005 chapter
- Why can three decision rules recommend different alternatives?
- They optimise different summaries of the same payoff table. Maximax focuses on the best possible outcome, maximin focuses on the protected worst outcome, and minimax regret focuses on the largest missed opportunity. A disagreement is not an arithmetic contradiction. It reveals how the preferred action changes with the management attitude built into the criterion. Chapter 01 · Decision Analysis Without Probabilities
- Why is a decision tree solved from right to left?
- A decision made today depends on the values of later branches. Starting at terminal payoffs lets you calculate each final chance-node value, carry that value back to an earlier decision, retain the best branch, and continue until the root has one optimal contingent strategy. Left-to-right calculation would reach decisions before their consequences were valued. Chapter 02 · Decision Trees and Expected Value
- What does the expected value of perfect information limit?
- It limits the rational price of information that perfectly reveals the uncertain state represented in the model. EVPI is the expected value with state-contingent best actions minus the best expected value without information. An imperfect survey or pilot should not be assigned greater decision value than that perfect-information ceiling. Chapter 03 · Information Value and Bayesian Updating
- What is the difference between scoring and AHP?
- A scoring model accepts criterion weights and alternative ratings directly, multiplies them and sums contributions. AHP derives priorities from reciprocal pairwise judgments, checks the consistency of those judgments, repeats the process for alternatives under each criterion and then combines the local priorities. Both produce rankings, but their input structures differ. Chapter 04 · Scoring Models and AHP
- What makes an optimisation model linear?
- Every objective and constraint term uses a fixed coefficient multiplied by one decision variable. The model contains no products between decision variables, no powers of variables and no nonlinear functions. A percentage rule can remain linear when the percentage multiplies a sum and the expression is rearranged without multiplying variables together. Chapter 05 · Linear Programming Models
- What stays unchanged inside a range of optimality?
- The current decision-variable values remain optimal when one objective coefficient changes inside its interval. The objective value does not stay fixed; substitute the new coefficient into the objective and recalculate. At a boundary there may be an alternative optimum, but the current solution remains optimal. Chapter 06 · LP Sensitivity Analysis
- What connects production decisions across periods?
- The inventory balance. Opening inventory plus production must equal demand plus ending inventory, adjusted only for flows the model explicitly includes. Ending inventory then becomes the next period's opening stock. Omitting that link treats each period as independent and can create or destroy product inside the model. Chapter 07 · Business Applications of LP
How to study for the exam
Treat the subject as a sequence of modelling decisions rather than a formula list. For each topic, start by naming the objective and the information available. Without probabilities, identify whether the table contains profits or costs before applying an optimistic, conservative or regret rule. With probabilities, draw node shapes and label branch probabilities before folding back.
For information questions, build a Bayes table in the fixed order prior, conditional, joint, marginal, posterior, and choose the best action separately inside each report branch. For AHP, make reciprocity automatic and retain enough decimals to test consistency before combining priorities.
For LP, define every variable with a unit and time period, read each constraint back into business language and check dimensional consistency. Solve graphical models at all extreme points, and keep the decision values separate from the objective value. In sensitivity questions, first classify the proposed change.
An objective-coefficient change uses the range of optimality and preserves current decision values inside the range. An RHS change uses the range of feasibility and preserves the shadow price inside the range. Anything structural or outside the range requires re-solving. Build the permitted examination notes from these routing decisions.
One side can hold uncertainty rules and trees, one Bayes and AHP, one LP formulation and geometry, and one sensitivity plus flow balances. Then work original questions without Excel, using the same calculator permitted in the exam. After every calculation, write a closing sentence that selects an action, quotes the computed basis and states the limitation.
That final move is where quantitative output becomes business decision analysis.
Your AI Management tutor for MGMT20005
Stuck on a hard MGMT20005 question? Sia is AskSia’s AI Management tutor — ask any MGMT20005 Business Decision Analysis question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.
