How to build a business analytics course: a complete guide for lecturers

Central question: What decision is the organisation trying to improve, and what evidence would change that decision?

Business Analytics should begin with the decision rather than the software. Students need to translate an ambiguous managerial concern into a decision, a measurable outcome, controllable levers, constraints and a statement of what good evidence would look like. This prevents the course becoming a sequence of techniques without a managerial purpose.

What this concept covers

• Decision statements, objectives, choices, outcomes and constraints.
• Descriptive, predictive and prescriptive questions and how they connect.
• Unit of analysis, time horizon, stakeholders and decision rights.
• Success measures, baseline performance and the cost of error.
• A practical analytics lifecycle from question to data, model, decision, action and monitoring.

Learning outcomes

• Frame a managerial problem as an answerable analytics question.
• Distinguish descriptive, predictive and prescriptive questions.
• Identify the target outcome, available choices and material constraints.
• Specify which evidence would change a recommendation.
• Recognise when more modelling will not improve the decision.

How to teach it

• Give students a vague brief such as "customer churn is rising" and require a one-sentence decision question before any analysis.
• Ask groups to classify candidate questions as descriptive, predictive or prescriptive.
• Have students name the cost of a false positive, false negative and delayed decision.
• Require a decision memo skeleton before a model is opened.

Case-style example

Beacon Retail has 240,000 active customers, annual churn of 18% and a retention budget of £1.2 million. Marketing wants to target the 30,000 customers with the highest churn probability, while Finance wants to target only customers whose expected retained margin exceeds the contact cost. Students define the decision, target metric, constraints and minimum evidence needed. They discover that predicting churn and deciding whom to retain are different analytics problems.

Where students struggle

Students often jump from a business symptom to a favourite technique. Push them to state who makes the decision, what can actually be changed and what error matters before they touch the data.

Reading and a quick check

Use Camm et al., Chapter 1. Quick check: ask each student to rewrite a broad management problem as one descriptive question, one predictive question and one prescriptive question.

Where a simulation fits

Use a short case or memo workshop here rather than a simulation. Students need the framing habit before an applied simulation can test it.

Where it leads

The next concept asks whether the data is fit for the question students have just framed.

Central question: Can the available data support the decision, and what changes when the data is incomplete, inconsistent or badly measured?

Most analytics errors begin before modelling. Business data is produced by operational systems, definitions and incentives, so a clean spreadsheet can still encode the wrong unit, duplicated customers or a misleading metric. This concept makes data preparation a visible analytical judgement rather than invisible housekeeping.

What this concept covers

• Measurement validity, data types, units and business definitions.
• Missing values, duplicates, outliers and inconsistent categories.
• Joining files, keys, granularity and time alignment.
• Selection bias, survivorship bias and data leakage.
• Reproducible preparation, data dictionaries and audit trails.

Learning outcomes

• Audit a dataset against a stated decision question.
• Diagnose missingness, duplication and inconsistent definitions.
• Choose defensible treatments for outliers and missing values.
• Explain how data leakage or selection bias can invalidate a model.
• Document transformations so another analyst can reproduce the analysis.

How to teach it

• Give students two files with deliberately inconsistent customer IDs and time periods.
• Ask them to build a one-page data-quality report before any charting.
• Compare different missing-value treatments and discuss which business assumption each treatment embeds.
• Grade the data dictionary and transformation log, not only the final model.

Case-style example

Northstar Travel has 82,000 bookings from three systems. Twelve percent of customer IDs appear in more than one format, 7% of booking values are missing and cancellations are recorded as both zero revenue and negative revenue. Students reconcile the files, calculate how each cleaning rule changes average booking value and write a data-quality note. Two defensible rules produce different revenue trends. Students discover that data cleaning can change the managerial story.

Where students struggle

Students tend to treat cleaning as a mechanical prelude. Ask them to justify every irreversible transformation in business language and to preserve the raw data.

Reading and a quick check

Use Camm et al., Chapter 4. Quick check: present five suspicious records and ask students which should be corrected, excluded, retained or escalated.

Where a simulation fits

A structured data-audit workshop fits best here. The approved simulations become more useful once students have basic descriptive and financial data skills.

Where it leads

With a credible dataset, students can describe performance and communicate patterns without overstating what the data proves.

Central question: What has happened in the business, where is performance changing and which visual or KPI makes the pattern visible?

Descriptive analytics is not simply "make a dashboard". Students need to choose measures that answer the decision question, calculate them consistently and design displays that reveal variation, comparison and materiality. This is also the first point where an applied company dataset can be used productively.

What this concept covers

• Measures of location, spread, distribution and association.
• KPI design, denominators, rates, growth and contribution analysis.
• Tables, charts and dashboards matched to the analytical question.
• Segmentation by customer, product, geography, time and operating driver.
• Visual integrity, annotation and avoiding misleading scales or aggregation.

Learning outcomes

• Calculate and interpret descriptive statistics.
• Select KPIs that trace performance to business drivers.
• Choose an appropriate visual form for a comparison, trend or distribution.
• Detect misleading aggregation or denominator choices.
• Build a concise dashboard that supports a managerial conversation.

How to teach it

• Give several chart types for the same dataset and ask which one best answers a stated question.
• Require students to annotate a dashboard with the three observations that matter most.
• Ask teams to redesign a deliberately cluttered executive dashboard.
• Use financial statements to connect ratios with business events, not just formula recall.

Case-style example

Luma Foods reports £96 million annual revenue, but the dashboard hides that the North region grew 14% while the South fell 9%. Gross margin is 34% overall, 41% online and 27% wholesale, while stockouts doubled in the fastest-growing region. Students compute contribution by channel, redesign the dashboard and recommend which issue deserves management attention. They discover that the company-level average conceals opposing operating stories.

Where students struggle

Students often confuse a visually attractive dashboard with an analytical one. Require every visual to answer a named question and every KPI to have an owner, definition and decision use.

Reading and a quick check

Use Camm et al., Chapters 2 and 3. Quick check: give one table and three chart options, then ask which choice minimises interpretation error.

Where a simulation fits

The Financial Statement Analysis simulation fits after this concept. Students work as a Financial Analyst, interpret the income statement, balance sheet and cash flow statement, calculate key ratios across reporting periods and connect numerical changes to business events.

Where it leads

Descriptive patterns create hypotheses. Statistical inference is what lets students judge whether an observed difference is likely to be signal rather than noise.

Central question: How much confidence should a manager place in a sample result, and how should uncertainty affect the decision?

Business decisions are rarely made with complete populations. Students need a practical command of sampling, confidence intervals and hypothesis tests, but the teaching goal is not ritualistic p-values. The goal is to quantify uncertainty and separate statistical evidence from business significance.

What this concept covers

• Random variables, distributions and sampling variation.
• Confidence intervals and estimation.
• Hypothesis tests, Type I and Type II errors, power and sample size.
• Statistical significance versus economic or managerial significance.
• Multiple testing, practical uncertainty and communicating ranges.

Learning outcomes

• Construct and interpret confidence intervals.
• Select and interpret a basic hypothesis test.
• Explain Type I and Type II errors in business terms.
• Distinguish statistical significance from decision relevance.
• Communicate uncertainty without hiding behind a single probability value.

How to teach it

• Translate every hypothesis test into the cost of acting and not acting.
• Ask students to report an effect size and interval before the p-value.
• Use a small sample where a visible difference is not conclusive, then increase sample size.
• Require a one-sentence management interpretation of every statistical output.

Case-style example

Cedar Health tests a reminder message on 1,600 patients. Attendance rises from 74% in control to 78% in treatment, but the intervention costs £1.10 per message and a completed appointment contributes £22. Students estimate the uplift, interval and expected economic value, then recommend whether to scale. One group prioritises statistical caution; another accepts uncertainty because the expected value is positive. Students discover that significance and action thresholds are not the same.

Where students struggle

Students may treat p < 0.05 as a universal decision rule. Make them state the cost of error, plausible effect size and decision threshold first.

Reading and a quick check

Use Camm et al., Chapters 5 and 7. Quick check: ask whether a precise but tiny effect should receive more managerial weight than an imprecise but potentially material one.

Where a simulation fits

A short inference lab or case works better than forcing a simulation here.

Where it leads

Inference prepares students to model relationships and make predictions while keeping uncertainty visible.

Central question: Which variables move with the outcome, how well does the model generalise and what should students avoid claiming from association?

Regression is often the bridge from descriptive analysis to prediction, but it also creates false confidence. Lecturers should emphasise specification, diagnostics, out-of-sample performance and interpretation rather than only coefficient calculation.

What this concept covers

• Simple and multiple linear regression.
• Categorical variables, interactions and transformations.
• Model fit, residuals, multicollinearity and diagnostics.
• Training and validation logic, overfitting and out-of-sample error.
• Association, omitted variables and the limits of causal interpretation.

Learning outcomes

• Estimate and interpret a regression model.
• Diagnose basic model specification problems.
• Compare in-sample fit with out-of-sample usefulness.
• Translate coefficients into business implications.
• Avoid causal claims that the design cannot support.

How to teach it

• Give students a model with an impressive R-squared but obvious leakage.
• Ask teams to remove variables one by one and defend the trade-off between simplicity and fit.
• Require a prediction interval or error metric alongside the point forecast.
• Ask what omitted factor could reverse the business interpretation.

Case-style example

Orbit SaaS models monthly churn for 18,000 accounts using seat count, support tickets, discount rate and product usage. A regression suggests every extra support ticket raises churn by 0.7 percentage points, but high-value accounts both use support more and renew more often. Students compare specifications, test an interaction and write a cautious interpretation. They discover that an apparently actionable coefficient may reflect customer mix rather than a causal effect.

Where students struggle

Students can mistake a fitted relationship for a mechanism. Ask them what the model predicts, what it explains and what it cannot establish.

Reading and a quick check

Use Camm et al., Chapter 8. Quick check: show two models with similar validation error and ask which is easier to defend to a manager and why.

Where a simulation fits

Use a modelling lab and memo rather than a simulation. Later simulations give students richer contexts in which model outputs inform decisions.

Where it leads

Regression provides one route to prediction. Forecasting adds time structure, seasonality and the discipline of rolling evaluation.

Central question: What is likely to happen next, and how should forecasts change planning when the future is uncertain?

Forecasting is valuable because many business decisions are timed before demand, cash flow or capacity is known. Students should learn to build a baseline, recognise trend and seasonality, compare accuracy across horizons and convert forecast error into planning consequences.

What this concept covers

• Naive and moving-average baselines.
• Trend, seasonality, smoothing and decomposition.
• Forecast accuracy metrics and holdout evaluation.
• Hierarchical and multi-period business forecasts.
• Prediction intervals, scenario ranges and forecast bias.

Learning outcomes

• Build and compare baseline forecasts.
• Identify trend, seasonal and irregular components.
• Evaluate forecasts using appropriate error metrics.
• Explain why the best average error may not be the best decision model.
• Translate forecast uncertainty into inventory, staffing or cash planning.

How to teach it

• Always benchmark a complex model against a naive forecast.
• Use rolling holdouts rather than evaluating on the training history.
• Ask what over-forecasting and under-forecasting cost the business.
• Have students present a forecast as a range plus decision rule rather than a single number.

Case-style example

Verde Kitchens must forecast weekly meal-kit demand for 52 stores. Last year averaged 420 units per store, holiday weeks were 35% higher and a new promotion is expected to add between 5% and 15%. Students compare a seasonal naive forecast with an exponential-smoothing model, calculate validation error and produce base, low and high capacity plans. They discover that the operationally best forecast depends on the asymmetric cost of waste versus stockouts.

Where students struggle

Students often optimise a statistical accuracy metric without asking which errors hurt the business most. Put the cost function next to the forecast metric.

Reading and a quick check

Assign Makridakis, Spiliotis and Assimakopoulos (2022) alongside Camm et al., Chapter 9. Quick check: ask why a naive benchmark is essential even when sophisticated models are available.

Where a simulation fits

Use a forecasting lab or planning case here. Working Capital Management becomes more useful later when forecast uncertainty has to be turned into operating choices.

Where it leads

Forecasts predict what may happen. Experimentation is about learning what an intervention causes.

Central question: Did the intervention cause the outcome, and what experiment or quasi-experiment would make that claim credible?

Business Analytics becomes more valuable when students can move from correlation to intervention. Randomised experiments are the cleanest teaching vehicle because they make counterfactual logic visible. The aim is not to turn the course into advanced econometrics, but to help students distinguish prediction from causal evidence.

What this concept covers

• Potential outcomes and the counterfactual problem in plain language.
• Randomisation, treatment and control groups.
• A/B tests, sample size, stopping rules and guardrail metrics.
• Selection bias, confounding and common threats to validity.
• When quasi-experimental reasoning is needed because randomisation is impossible.

Learning outcomes

• Design a basic randomised business experiment.
• Choose primary and guardrail metrics.
• Interpret treatment effects with uncertainty.
• Identify common threats to internal and external validity.
• Explain why predictive importance is not the same as causal impact.

How to teach it

• Turn a regression finding into an experiment students could actually run.
• Use pre-registration style prompts: hypothesis, metric, sample, stopping rule and decision rule.
• Give students an A/B test with peeking and ask how the design was compromised.
• Ask whether an observed effect is large enough to justify operational rollout.

Case-style example

NovaStream tests a new onboarding flow on 20,000 users. Trial-to-paid conversion is 11.8% in control and 12.7% in treatment, but support contacts rise from 6.2% to 8.1%. Students estimate the uplift, consider the support-cost guardrail and decide whether to roll out, redesign or test again. Growth values the conversion gain; Operations worries about service load. Students discover that a valid experiment can still leave a multi-metric managerial decision.

Where students struggle

Students can treat randomisation as proof that any observed difference matters. Keep effect size, uncertainty, practical significance and unintended consequences in the same discussion.

Reading and a quick check

Use an instructor-provided experimentation note and connect back to Camm et al., Chapter 7. Quick check: ask students to name the treatment, counterfactual and guardrail metric in a familiar digital experiment.

Where a simulation fits

A designed experiment or causal critique is the better activity here. The approved simulations are decision environments rather than causal-identification exercises.

Where it leads

Students can now distinguish prediction from causality. The next concept broadens predictive work into classification, segmentation and machine learning.

Central question: How should the course use machine learning to improve decisions without rewarding complexity for its own sake?

Machine learning belongs in a modern Business Analytics course, but usually as structured-data decision support rather than a race for the most complicated algorithm. Students should compare baselines, choose metrics that match business losses and understand threshold trade-offs.

What this concept covers

• Classification versus regression and common business targets.
• Logistic regression, trees and ensemble methods at an interpretable level.
• Segmentation and clustering as exploratory structure, not discovered truth.
• Train, validation and test logic; cross-validation and overfitting.
• Confusion matrices, precision, recall, ROC ideas and threshold choice.

Learning outcomes

• Select a sensible predictive task and baseline.
• Evaluate classification performance using business-relevant metrics.
• Choose and defend a classification threshold.
• Interpret a segmentation without overclaiming its permanence.
• Compare model complexity with transparency, cost and maintainability.

How to teach it

• Start with a naive benchmark and logistic regression before adding tree-based models.
• Give students a confusion matrix with asymmetric costs and ask them to set the decision threshold.
• Have teams explain a cluster solution to a marketing manager and then critique its stability.
• Reward a simpler model when it produces a better decision process.

Case-style example

Meridian Bank screens 50,000 small-business applications. Default prevalence is 6%. At a 0.50 threshold, the model catches 48% of defaults and rejects 7% of non-defaults; at 0.30 it catches 73% of defaults but rejects 16% of non-defaults. Students calculate expected costs using a £18,000 average default loss and £900 contribution from a good account, then recommend a threshold. They discover that model accuracy cannot choose the business policy.

Where students struggle

Students often optimise accuracy on imbalanced outcomes. Ask them to express every metric as an operational trade-off and compare against a simple baseline.

Reading and a quick check

Assign Schmitt (2023) and use Camm et al., Chapters 6, 10 and 11 selectively. Quick check: ask why a 94% accurate default model could still be useless when only 6% of accounts default.

Where a simulation fits

Use a machine-learning model comparison and decision-threshold memo here. Approved simulations become most powerful when prediction feeds a constrained decision in later sessions.

Where it leads

Prediction says what is likely. Prescriptive analytics asks what the organisation should do given objectives and constraints.

Central question: Given forecasts, costs, objectives and constraints, what action should the organisation choose?

Prescriptive analytics is where the course becomes explicitly decision-led. Students translate a managerial choice into decision variables, an objective and constraints, solve the model and then test whether the mathematically optimal answer is managerially credible.

What this concept covers

• Decision variables, objective functions and constraints.
• Linear optimisation and spreadsheet model structure.
• Integer decisions and mutually exclusive projects.
• Resource allocation, capacity, product mix and investment selection.
• Sensitivity, shadow-price intuition and the difference between an optimum and a robust recommendation.

Learning outcomes

• Formulate a business problem as an optimisation model.
• Solve and interpret a constrained resource-allocation problem.
• Check model feasibility and boundary assumptions.
• Perform sensitivity analysis around key constraints.
• Defend when a near-optimal solution is preferable to the mathematical optimum.

How to teach it

• Require students to write the decision variables in words before equations.
• Ask which constraint would be worth relaxing and what it would cost.
• Compare the optimised result with a managerially simpler alternative.
• Use project allocation and managerial resource decisions as applied contexts.

Case-style example

Apex Manufacturing has a £10 million capital budget and five projects requiring £2m, £3m, £4m, £5m and £6m. Forecast NPVs are £0.7m, £1.0m, £1.8m, £2.1m and £2.5m, but projects 3 and 5 cannot both be selected and at least one safety project must be funded. Students formulate the binary allocation model, solve it and compare the optimum with management's preferred portfolio. They discover that constraints, not project rankings alone, determine the decision.

Where students struggle

Students can present Solver output as if it were a recommendation. Ask them to identify which assumptions, constraints or stakeholder priorities could invalidate the optimum.

Reading and a quick check

Assign Bertsimas and Kallus (2020) for the predictive-to-prescriptive bridge and use Camm et al., Chapters 12, 14 and 15. Quick check: ask students to identify the variable, objective and two constraints in a resource-allocation brief.

Where a simulation fits

The Capital Budgeting simulation fits directly after this concept: students act as CFO, apply NPV, IRR, profitability index and payback, then allocate a fixed $10 million budget across a portfolio of projects. The Managerial Accounting simulation also fits because students use CVP, break-even, ROI and NPV before allocating $100 million across competing opportunities.

Where it leads

Optimisation needs uncertainty. The next concept adds scenarios, simulation and portfolio trade-offs so students can stress-test decisions rather than trust point estimates.

Central question: How should an analytics recommendation change when inputs, returns or outcomes are uncertain?

A strong course should make uncertainty explicit before the capstone. Students move from deterministic optimisation to scenarios, Monte Carlo logic and risk-return trade-offs, learning that a recommendation should be robust to plausible changes rather than optimised for a single forecast.

What this concept covers

• Scenario design and sensitivity analysis.
• Monte Carlo simulation logic, distributions and correlated inputs.
• Expected value, downside risk and risk appetite.
• Portfolio diversification, covariance, CAPM and mean-variance intuition.
• Stress tests and robust decision criteria.

Learning outcomes

• Build and interpret a simple scenario or simulation model.
• Distinguish parameter uncertainty from model uncertainty.
• Explain diversification and correlation in portfolio risk.
• Compare decisions using expected return and downside measures.
• Stress-test a recommendation against plausible adverse scenarios.

How to teach it

• Start from the Session 9 optimum and perturb the inputs.
• Use tornado charts or scenario tables to expose the assumptions that matter most.
• Show why independent simulations can understate risk when drivers are correlated.
• Ask students to defend risk appetite, not simply calculate volatility.

Case-style example

Harbor Investments allocates £25 million across four asset groups with expected returns of 5%, 7%, 9% and 12%. The highest-return pair has correlation 0.82, while the lower-return assets provide diversification. Students compare a return-maximising allocation with a diversified mean-variance portfolio, then stress equity returns down 20%. One team prioritises expected return; another prioritises downside resilience. Students discover why covariance changes the portfolio decision.

Where students struggle

Students can turn risk analysis into more charts without changing the recommendation. Require them to state which scenario would trigger a different action.

Reading and a quick check

Use Camm et al., Chapters 13 and 16. Quick check: ask why two individually risky assets can reduce total portfolio risk when combined.

Where a simulation fits

The Portfolio Management simulation is the strongest fit here. Teams of 3-5 act as Portfolio Managers under Hedge Fund or Pension Fund mandates, use CAPM and mean-variance optimisation across 25 global companies, then construct and rebalance portfolios through four quarters as news and prices change.

Where it leads

Students can now make decisions under uncertainty. Session 11 asks how analytics supports ongoing performance management and scarce-resource allocation inside an operating business.

Central question: How should managers monitor performance, diagnose operating drivers and allocate resources when metrics compete?

Business Analytics is not only a one-off model. Organisations need repeatable measures, driver trees and decision routines that connect financial outcomes with operational levers. This concept integrates financial, managerial and working-capital analytics before the final governance session.

What this concept covers

• KPI hierarchies, driver trees, targets and variance analysis.
• Financial and operating measures, including profitability, liquidity and efficiency.
• Contribution margin, break-even, ROI and investment trade-offs.
• Working capital metrics including DSO, DIO, DPO and cash conversion cycle.
• Resource allocation, corrective action and the danger of metric gaming.

Learning outcomes

• Design a small KPI set linked to a decision objective.
• Trace a financial outcome to operating drivers.
• Use variance and ratio evidence to prioritise corrective action.
• Evaluate resource-allocation trade-offs across functions.
• Identify where a metric could create unintended behaviour.

How to teach it

• Limit teams to five executive KPIs and force them to justify every inclusion.
• Use a driver tree that links revenue, margin, working capital and cash.
• Ask what behaviour a target could unintentionally incentivise.
• Combine financial statements with operating events so students diagnose causes rather than recite ratios.

Case-style example

Terra Consumer has £180 million revenue, 31% gross margin and £14 million cash. DSO has risen from 42 to 57 days, DIO from 61 to 78 days and marketing proposes a £12 million growth programme. Students quantify the working-capital cash tied up, compare the growth proposal with inventory and receivables interventions and allocate a £20 million improvement budget. Finance and Marketing reach different priorities. Students discover that KPI optimisation can create cross-functional trade-offs.

Where students struggle

Students often create too many metrics and treat each function separately. Make them connect measures to one enterprise decision and show second-order effects.

Reading and a quick check

Use Orjatsalo, Hussinki and Stoklasa (2025) with relevant Camm chapters. Quick check: ask students which five metrics they would put on a CEO page and which they would deliberately leave off.

Where a simulation fits

The Working Capital Management simulation fits here: students act as CFO of Northwood Furnishings through a 12-month scenario, manage receivables, inventory and payables, track DSO, DIO, DPO and the cash conversion cycle, and finish with an expansion recommendation. The Managerial Accounting simulation also fits for cross-functional resource allocation.

Where it leads

The final concept asks whether the whole analytics process is explainable, governable and defensible, including when AI is used.

Central question: How should an analyst defend a recommendation when models are imperfect, AI is involved and stakeholders need to trust the process?

The last concept turns technical work into accountable management practice. Students need to communicate uncertainty, document assumptions, respect privacy and fairness, control model risk and state where AI contributed. The capstone should reward the quality of the decision process more than polished output.

What this concept covers

• Decision memos, executive visuals and oral defence.
• Data governance, ownership, access, lineage and reproducibility.
• Privacy, bias, fairness and model-risk controls.
• Human oversight, monitoring, drift and escalation.
• Permitted use of generative AI, disclosure and verification responsibilities.

Learning outcomes

• Communicate an analytics recommendation to a non-technical decision maker.
• Document material assumptions, limitations and uncertainty.
• Identify ethical, privacy and governance risks in a business model.
• Design basic monitoring and escalation controls.
• Defend how AI-assisted work was verified and where human judgement remained decisive.

How to teach it

• Require a one-page executive recommendation and a five-minute oral defence.
• Ask students to separate facts, assumptions, model outputs and judgement.
• Give a model that performs well overall but poorly for one subgroup.
• Have peers challenge evidence quality and ask what would change the recommendation.

Case-style example

Helix Mobility uses an AI-assisted model to prioritise 8,000 driver applicants. Overall validation accuracy is 84%, but the false-negative rate is 9% for one region and 17% for another because historical data is sparse. Students decide whether to deploy, retrain, limit use or add human review, then write the governance controls and disclosure. Operations wants speed; Risk wants a higher evidence threshold. Students discover that technical performance does not settle accountability.

Where students struggle

Students can treat ethics as a final checklist or AI disclosure as a footnote. Bring governance into the recommendation by asking who owns the model, who can override it and what happens when performance drifts.

Reading and a quick check

Assign Papagiannidis, Mikalef and Conboy (2025). Quick check: ask students to identify one risk that requires technical control, one that requires organisational control and one that requires managerial judgement.

Where a simulation fits

Use the final capstone to synthesise evidence from any approved simulation already completed. The platform evidence should support academic judgement, not substitute for a lecturer's evaluation or establish which individual student made which argument in a team activity.

Where it leads

This closes the course: students should be able to move from an ambiguous decision to credible data, a defensible model, an action, a control plan and a clear managerial defence.

10-week course structure

A 10-week version should compress adjacent methods rather than delete the decision lifecycle. Combine descriptive analytics with the first applied financial dataset, combine inference with regression, and integrate governance into the capstone.

14-week course structure

A 14-week version creates room to separate data visualisation from KPI application, classification from broader machine learning and prescriptive optimisation from risk. Use the extra time for model defence, simulation reflection and assessment integration rather than more disconnected techniques.

The Financial Statement Analysis Simulation places each student in the role of a Financial Analyst. Students work through an income statement, balance sheet and cash flow statement, calculate and interpret profitability, liquidity, efficiency and solvency ratios, and update their view as the company moves through five reporting periods. The activity culminates in a supported share-price prediction, which makes it useful for teaching how descriptive evidence becomes a forward-looking judgement.

Why it fits here

By the end of Session 3, students can calculate descriptive measures and explain why a KPI matters. The simulation makes them use those skills on an integrated company record rather than isolated exercises. It is especially useful before regression and forecasting because it establishes the habit of connecting numerical changes to business events.

What students do

• Review the income statement, balance sheet and cash flow statement as one connected evidence set.
• Calculate and interpret profitability, liquidity, efficiency and solvency ratios.
• Track how ratios and financial outcomes change across five reporting periods.
• Relate operating events to changes in the statements rather than treating ratios as formula exercises.
• Form a supported view of company performance and make a share-price prediction using the evidence available.

Recommended course placement

• At the end of Session 3 after descriptive analytics, KPIs and visualisation.
• As a pre-class or live individual activity before a seminar on inference and forecasting.
• As a formative diagnostic for students who need more practice linking financial statements to business performance.

Knowledge students should have before the simulation

• Basic purpose of the income statement, balance sheet and cash flow statement.
• Meaning of common profitability, liquidity, efficiency and solvency ratios.
• Ability to explain a trend or comparison in business language.

Delivery format

Suggested debrief questions

• Which ratio changed your view of the company most, and why?
• Which financial-statement movement looked important but turned out to be secondary?
• Where could two analysts reach different conclusions from the same evidence?
• What additional data would you want before making a stronger forward-looking claim?
• How would you distinguish a descriptive conclusion from a predictive one?

The Portfolio Management Simulation is a team-based portfolio decision environment in which students act as Portfolio Managers under either Hedge Fund or Pension Fund mandates. Teams work with 25 global companies, use CAPM and mean-variance optimisation, construct portfolios and rebalance through four quarters as price history, company information and market news evolve. Hedge Fund teams target Alpha while Pension Fund teams focus on Sharpe Ratio, creating different defensible portfolio objectives from the same market environment.

Why it fits here

Session 10 is where the course brings prediction, optimisation and uncertainty together. Students already know that the mathematically best point estimate may be fragile. The simulation makes risk appetite, diversification, mandate and rebalancing visible across sequential decisions rather than in a single static optimisation problem.

What students do

• Work in teams of 3-5 as Portfolio Managers under a Hedge Fund or Pension Fund mandate.
• Review historical price information and the supplied analytical materials for 25 global companies.
• Use CAPM and mean-variance optimisation to support portfolio construction.
• Choose holdings consistent with the mandate, including the different emphasis on Alpha or Sharpe Ratio.
• Rebalance across four quarters as new market and company information arrives.
• Compare portfolio outcomes with other teams and explain which decisions, assumptions and risk choices drove the result.

Recommended course placement

• At the end of Session 10 after scenario analysis, CAPM and mean-variance optimisation.
• Across Sessions 10 and 11 if the course uses the activity as a substantial applied workshop.
• Before a capstone memo so teams can convert performance outcomes into a defended recommendation and monitoring plan.

Knowledge students should have before the simulation

• Expected return and risk.
• Basic CAPM.
• Sharpe Ratio.
• Diversification, covariance and correlation.
• Mean-variance optimisation and the idea of a portfolio constraint or mandate.

Delivery format

Suggested debrief questions

• Which assumption most influenced your initial allocation?
• How did the Hedge Fund and Pension Fund mandates change what counted as a good decision?
• Which rebalance improved the portfolio, and which was hindsight rather than a process you could defend ex ante?
• Where did diversification protect the team, and where did correlation surprise you?
• Would you keep the same risk budget if the next quarter began today?
• What individual evidence would you add if the simulation formed part of summative assessment?

Grading with the Financial Statement Analysis Simulation

The Financial Statement Analysis Simulation is individual and places students in the role of a Financial Analyst working across the three financial statements, ratios, reporting periods and a supported share-price prediction.

• Accuracy and interpretation of selected profitability, liquidity, efficiency and solvency measures.
• Ability to connect ratio movements to business events and statement relationships.
• Quality of the final evidence-based performance view and prediction rationale.
• Recognition of uncertainty, missing evidence and alternative interpretations.

Grading with the Managerial Accounting Simulation

The Managerial Accounting Simulation uses individual submissions. Students first work with contribution margin, break-even, CVP, ROI and NPV across three products, then allocate $100 million across five opportunities while considering CFO, COO, CMO and CEO perspectives.

• Correct use and interpretation of cost-volume-profit and investment metrics.
• Quality of resource-allocation reasoning across competing opportunities.
• Ability to explain trade-offs between functional perspectives rather than optimising one metric.
• Individual reflection on which assumption or management priority changed the allocation.

Grading with the Capital Budgeting Simulation

The Capital Budgeting Simulation is an individual CFO exercise. Students evaluate projects with NPV, IRR, profitability index and payback before allocating a fixed $10 million capital budget.

• Correct interpretation of investment appraisal metrics and their limitations.
• Consistency between the stated decision rule and the selected portfolio.
• Recognition of budget constraints and project interactions.
• Sensitivity or reflection on how changed assumptions would alter the allocation.

Grading with the Portfolio Management Simulation

The Portfolio Management Simulation is team-based. Students work as Portfolio Managers under Hedge Fund or Pension Fund mandates, use CAPM and mean-variance optimisation across 25 global companies and rebalance over four quarters.

• Portfolio logic and consistency with the assigned mandate.
• Use of risk, return, diversification and rebalancing evidence.
• Quality of team decision documentation and post-simulation explanation.
• An individual memo, viva or reflection that creates attributable evidence because the platform outcome is team-level.

Grading with the Working Capital Management Simulation

The Working Capital Management Simulation is single-player. Students act as CFO of Northwood Furnishings in a 12-month scenario, manage receivables, inventory and payables, monitor DSO, DIO, DPO and the cash conversion cycle, and make a final expansion recommendation.

• Interpretation of working-capital measures and cash implications.
• Consistency and sequencing of receivables, inventory and payables decisions.
• Quality of the final expand, expand with caution or delay recommendation.
• Use of simulation calculations, decisions and written rationales as supporting evidence rather than automatic grading.

The platform records decisions and comparative outcomes, and some individual simulations also retain individual performance evidence. That evidence supports your academic judgement; it does not replace it. For a team activity, it does not establish which individual student made which argument, so add an individual memo, viva or reflection when the mark must be attributable.

Moderation, individual evidence and free-riding

• Publish criteria before the activity so students know the simulation score is not the grade.
• Retain the brief, rubric, submitted memo or model, and the relevant simulation evidence used in judgement.
• For team activities, add individual evidence such as a 400-word assumptions note, five-minute viva or structured reflection.
• Moderate borderline or unusually high/low cases using the same artefacts and criteria as other coursework.
• If a student did not participate meaningfully, use local contribution and academic-regulation procedures rather than inferring participation from the team result.