Kumu / Labs
The Labs
Interactive versions of each course's core model. Every lab mirrors the real project artifact — same named ranges, same conventions, same validation checks — so what you learn playing here transfers directly to the graded work. Below them, the guides: the how-to pages the stages assume you have read.
Labs — models you can move
Ratio Lab
Two years of financial statements in, all six ratio categories out — with the balance-sheet
check, startYear_* averaging, and the Du Pont decomposition validated against direct
ROE. Mirrors performance-ratios-template.xlsx.
FX Hedging Lab
Forward hedge, three-step money-market hedge, covered-interest-parity check, put-option floor, and the ±5% sensitivity grid with payoff chart. Mirrors the FX hedging workbook, named ranges and all — and doubles as the Stage 4 check-figure oracle.
Econ Policy Lab
A live supply-and-demand market: impose taxes, subsidies, ceilings, and floors, and watch consumer surplus, producer surplus, government revenue, and deadweight loss respond — the Policy Shock project's diagram, touchable.
Farm Profit Lab
A price-taking market garden: plant beds of three crops, watch marginal cost climb with
diminishing returns, find P = MC per crop, and race the optimizer to the best mix. Mirrors
farm-profit-optimizer-template.xlsx, Solver and all.
AI Tools Lab
A guided field guide, not a slider playground: chat vs "code" tools and the harness idea, web vs desktop vs CLI, and why giving the AI your repo is the single biggest upgrade. Read it before your first AI build session — every project's build stage assumes it.
Guides — the how-to pages
A lab is a model you can move. A guide is a page you read once and come back to. These are the same pages the Reference menu lists — they are repeated here because this is the directory, and a directory is the one place where duplication is a feature.
Portfolio repo standard
The one public repository every engagement commits into: the folder tree explained, the four starter files, naming rules, and a copy-paste prompt that builds the skeleton for you.
AI conventions
What goes in AGENTS.md — and why CLAUDE.md is one line pointing at
it. A baseline to copy, how to keep it current, and what must never be committed.
Deliverable templates
The engagement brief, the spec, the decision memo, and the prompt log — structure and an annotated example for each, so no deliverable starts from a blank page.
Git mechanics
Local vs remote, add → commit → push in GitHub Desktop and on the command line, creating
folders on github.com, .gitignore, and the mistakes that cost people a grade.
Excel Solver
Install it on Windows or Mac — and find out now that Excel for the web has none. The three inputs, integer decisions, and what GRG Nonlinear is really doing when it walks uphill from wherever you left the numbers.
How to use a lab well
- Predict, then check. Before moving a slider or loading data, say what should happen. The gap between your prediction and the screen is where the learning is.
- Break things. Every lab has validation checks — trip them on purpose so you recognize the failure modes in your own graded models.
- Bring your own numbers. Sample data is for exploring; your assigned scenario or chosen company is for rehearsing.
- Interrogate Kumu. The ✳ tutor on each lab knows the current model and will quiz you, or be quizzed.