MGT 816 · Fall 2026

Problem Sets

Two graded problem sets, done in groups. One submission per group, due on Canvas at 1:00 pm — the start of the first section — regardless of which section you attend.

Problem Set 1: Building the Cap Table

Follow Houzz from formation through a priced Series B: journal entries, cap tables round by round, an option-pool negotiation, and a recommendation to the founders.

Out Wednesday, September 9 (Session 2)
Due Wednesday, September 23 at 1:00 pm
Groups 2–4 students; one submission per group
Points 40 — Part I 18 · Part II 10 · Part III 8 · Part IV 4

How to do it

Work in the PS1 digital workbench — the full assignment is there: the case narrative, the pre-built cap-table grids and journal-entry widgets, and the submission generator.

  1. Open the workbench and enter your group (2–4 names + NetIDs).
  2. Work through Parts I–IV. Progress autosaves to this browser only — use Export progress / Import progress to hand off within your group.
  3. When every part is complete, Generate submission sheet downloads one HTML file — upload that file to Canvas by the deadline.

The workbench handles structure and arithmetic hygiene — ownership percentages, balance warnings — and does not check answers. Every number entered is your group’s own work; AI may be used as a study aid, but you must be able to reproduce and defend every figure.

Downloads

  • Assignment (PDF) — the same case and questions in print form, with exhibits; use it to read offline or as the traditional submission path if you can’t use the workbench
  • Excel scratch template (optional) — blank tabs mirroring the four parts, for those who prefer to rough out the math in a spreadsheet first

Problem Set 2: AI-Assisted Forecasting

Build a driver-based FY2027 operating forecast for Redline Labs, an AI contract-review startup, with AI tools doing the analyst work — and you certifying every number against the model and the verification log. Part A is a unit-economics warm-up (churn, retention, NRR, and spotting the AI’s errors); Part B is the Excel build; Part C is scenarios, sensitivity, and a one-page board memo.

Out Monday, September 28 (Session 7)
Due Monday, October 12 at 1:00 pm
Groups Same groups as PS1; one submission per group
Points 40 — Part A 8 · Part B 18 · Part C 10 · Verification Log 4

How to do it

You hand in three files: your Excel model, a one-page board memo, and your workbench submission sheet.

  1. Build the model in Excel, starting from starter.xlsx in the Model Pack below. Part B — the model itself — stays in Excel; clean model hygiene in Excel is the point.
  2. Complete Part A, the five checksum certifications, the scenario and sensitivity readouts, and the Verification Log in the PS2 digital workbench. Progress autosaves to this browser only — use Export progress / Import progress to hand off within your group. When it is complete, Generate submission sheet downloads one HTML file.
  3. Write the memo one-pager from one_pager_template.pptx and export it to a one-page PDF.
  4. Put every group member’s name on all three files and upload all three to Canvas by the deadline.

The workbench formats and flags (percentages, identity warnings, completeness); it does not check answers. AI is required on this problem set, and every number in your workbook and memo must be human-certified: you must be able to reproduce and defend each figure without the AI. Any number in your memo that cannot be traced to a cell in your model loses 2 points, each.

TipThe AI workflow, in order (prompts P1–P6 on the prompt card)
  1. Extract — P1, Analyst. Paste case_facts.txt; take back the driver table and driver tree. Anything not in the case text goes under OPEN QUESTIONS, not into the model.
  2. Benchmark — P2, Skeptic. Classify each assumption aggressive / typical / conservative. Every external number traces to the benchmark sheet or a checkable source, or it gets stamped UNVERIFIED and never becomes a driver.
  3. Draft schedules — P3, Builder. Take formula logic in plain words and build every cell yourself — the AI never touches the .xlsx. Hand-recompute one month, then confirm beginning + adds − churn = ending across all 12.
  4. Scenarios — P4. Vary only the four named drivers. The AI proposes values; your own Excel data table decides which driver moves December cash the most.
  5. Red-team — P5, Board Red-Team. Paste your one-pager numbers; take the five hardest board questions; recompute December top to bottom.
  6. Narrate — P6, Drafter. The memo is written from verified numbers only — a bracketed placeholder beats an invented figure (that is the −2 rule).

Log as you go: every AI claim you rely on gets a Verification Log row the moment you check it, not reconstructed at the end. The full rules are in ground_rules.pdf; the exact prompts are on the prompt card.

Downloads

Solutions and grading guides are on Canvas only — they are never posted on this site.