Instructor Guide
The course in one paragraph
Twelve lessons in four modules — data preparation, modeling & DAX, visualization, deployment & AI — taught over one fictional retailer (Lantern & Pine) whose every figure is generated by seeded scripts and therefore verifiable by construction. Assessment is built grader-first: each assignment is a named contract (exact measure names, verified values, required model shapes) that machines can check, plus judgment components (memos, data logs, titles) a human grades against short rubrics. The through-line students should be able to state by Week 15: tools produce confident output whether or not it is right; the analyst's defense is possession of expectations.
The 15-week map
| Week | Session | Due |
|---|---|---|
| 1 | L01 From Data to Decision | — |
| 2 | L02 Power Query I | A01, Quiz 1 |
| 3 | L03 Power Query II | A02, Quiz 2 |
| 4 | L04 The Data Model | A03, Quiz 3 |
| 5 | L05 DAX I | A04, Quiz 4 |
| 6 | L06 DAX II | A05, Quiz 5 |
| 7 | Midterm (concept quiz + Harbor Lane practical) · Project 1 launched | A06, Quiz 6 |
| 8 | L07 Choosing the Right Chart | — |
| 9 | L08 Beyond the Bar Chart | A07, Quiz 7 |
| 10 | L09 Interactivity | A08, Quiz 8, Project 1 due |
| 11 | L10 Report Design | A09, Quiz 9 |
| 12 | L11 The Service · Capstone launched | A10, Quiz 10 |
| 13 | L12 Analytics & AI | A11, Quiz 11 |
| 14 | Capstone studio (milestone checks) | A12, Quiz 12 |
| 15 | Final concept exam · Capstone due | — |
Pacing variants
- 2 credits / 10 weeks: spine = L01–L07 + L09 + L10 + capstone-lite (model + Briefing page + memo). Cut L08, L11, L12 and Project 1; the midterm practical becomes the only exam.
- 8-week accelerated: two lessons/week; capstone only; midterm becomes a take-home checkpoint after Week 3.
- Cert-prep bolt-on: the PL-300 domain mapping (COURSE-PLAN §3) plus two timed runs of the exam banks. The course covers Prepare/Model/Visualize thoroughly and Deploy at concept level; supplement deployment drills if certification is a stated goal.
- The capstone is never cut.
The two questions every instructor asks first
1 · Mac students
Power BI Desktop is Windows-only. In descending order of preference: campus lab machines; an institution-provided virtual desktop (Azure Lab Services and equivalents work well and are the cleanest at scale); student-managed Windows VMs (Parallels/UTM — workable, support-heavy). Browser-based authoring in the Service covers exploration but not the graded Desktop workflow. Decide the answer before Week 1 and put it in the syllabus; the L01 topic page tells students to ask you, so have the answer ready.
2 · Licensing lane
| Your institution has… | What runs live | What becomes concept-only |
|---|---|---|
| Nothing (students use free accounts) | All of L01–L10, all assignments, both projects — the entire graded core | L11's shared-workspace/app/refresh exercises (its assignment A11 is deliberately Desktop-checkable) |
| Microsoft 365 with Pro | Everything above + live workspaces, apps, RLS assignment | Capacity-scale topics |
| Fabric capacity | Everything | — |
Every assignment grades from the .pbix in every lane — this was a design constraint, not luck. Verify current Microsoft licensing terms each term; the tables in L11 describe the shape, not the prices.
Datasets and the verification standard
All data is generated: build-tools/generate_universe.py (Lantern & Pine),
generate_secondary.py (midterm clinic, Project 1 garden centers, capstone
roasters). Regenerate any time — output is deterministic, and every figure cited anywhere in
the course traces to a manifest (datasets/*.json). If you modify a generator,
re-run the compute scripts and expect content edits: the manifests are the single
source of truth, and the course's own history includes catching its author asserting figures
the data contradicted. The planted defects (documented in each manifest) are
pedagogy; do not "clean" the raw files.
Grading architecture
- Rubric JSONs (
assessments/rubrics/) define machine checks: visual contracts, data-state checks, model checks (incl. relationship-absence), measure checks (names, values, composition, required functions), filtered-value checks, RLS checks, and report checks (alt-text coverage, tab order, page names). They are the integration contract for automated grading; graded manually, they are simply precise checklists. - Instructor-graded components are deliberately few and always short-rubric'd: titles (A07), design log (A10), deployment plan (A11), influencer finding + DAX review (A12), the data log (P1), and the insight memo (capstone, 25 pts — the single largest human-graded item, by design).
- Quiz banks ship as D2L-format CSV (25+ questions, deliver 15,
randomized);
check_quiz_bank.pygates structure. Points suggestion: quizzes 15% (drop 2) · assignments 30% · midterm 15% · Project 1 10% · capstone 20% · final 10%.
Session notes, in one line each
| Lesson | The moment to land |
|---|---|
| L01 | May beats December — let the room predict wrong first |
| L02 | Profile-first; the name-only dedup deletes 77 real customers |
| L03 | The inner join silently keeps 6%; the 23 "duplicates" are real sales |
| L04 | The M:M dialog warns and permits — decline it on camera |
| L05 | Two averages, 2× apart, both correct — resolve in words first |
| L06 | Targets joins the star; 107% attainment has two honest readings |
| L07 | The chart is a claim; build the three lies and un-build them |
| L08 | The map silently drops 44.7% — the caption is the map |
| L09 | $24.8M on the card: the invisible filter, produced live |
| L10 | Count the left edges before and after; alt text as the finding |
| L11 | RLS rides the star; the leak IS a filter-flow gap |
| L12 | The generated formula renders cleanly and is wrong — the thesis lands |
Materials inventory
content/— all student pages (HTML, self-contained) + the twelve decks (.pptx, regenerable frombuild-tools/build_deck_*.py)assessments/— 12 quiz banks + 2 exam banks + all rubricsdemos/specs/— 24 narrated-demo specifications (render-ready when the video pipeline is provisioned; usable as live-demo scripts meanwhile)screenshots/inventory.json— every screenshot slot with capture specstheme/iris-theme.json— the course report theme; also the accessibility worked-example- Solutions ship for every practice tier; assignments and exams have rubric answer keys only