Pre-Build Spec — BAN2100 Module 6
Status: Approved — retrospective as-built documentation
Generated by: Nexus (Claude Code)
Approved by: Fadl Altarzi (LXD) — 2026-04-28
Build version: v2.6
Date: 2026-04-28
1. Assessment Identity
- Course: BAN2100 — Data Analytics for Business
- Module: 6 — Introducing Prescriptive Analytics
- Company: Tesco (TSCO)
- Version: v2.6 (last updated 2026-04-27)
- Route:
/assessments/ban2100/module-6
2. Step Structure
| # |
Step ID |
Label |
Type |
Lock dependency |
showScenario |
| 1 |
scenario |
Dashboard Data |
Read-only overview (3-panel dashboard) |
None |
no |
| 2 |
classify |
Classify Types |
MCQ (4 questions) |
Step 1 complete |
yes |
| 3 |
patterns |
Interpret Patterns |
MCQ (3 questions) |
Step 2 complete |
yes |
| 4 |
error |
Analyst QA |
Written response (Spot the Error) |
Step 3 complete |
no |
| 5 |
recommendation |
Advisory Brief |
Mixed: MCQ selection + written justification |
Step 4 complete |
yes |
| 6 |
tools |
Tool Selection |
MCQ (4 questions, incl. AI question) |
Step 5 complete |
no |
3. MCQ Inventory
Step 2 — Classify Types (AnalyticsClassifier)
- Questions: 4
- Categories: Analytics type classification — one per panel plus the recommendation
- CLO targets: CLO-1 (analytics continuum definitions), CLO-4 (understanding analytics type → decision context)
- Panel A question (qa): Identify Panel A (descriptive — revenue, transactions, return rate, category mix) → correct: "Descriptive"
- Panel B question (qb): Identify Panel B (diagnostic — regional revenue vs marketing spend, correlation) → correct: "Diagnostic"
- Panel C question (qc): Identify Panel C (predictive — ARIMA forecast, Q1 FY2024 by category) → correct: "Predictive"
- Recommendation question (qd): Classify the final action recommendation → correct: "Prescriptive"
Step 3 — Interpret Patterns (PatternInterpreter)
- Questions: 3
- Categories: Identifying the most significant management-decision-driving finding in each panel
- CLO targets: CLO-3 (analyze patterns and trends), CLO-4 (impact on business decisions)
- Panel A question (pa): Most significant flag in descriptive panel → correct: return rate +33% relative increase (5.1% → 6.8%)
- Panel B question (pb): Most significant finding in diagnostic panel → correct: South (9.4×) and West (6.1×) structural underperformance vs North/East (12.8×) despite equal budgets
- Panel C question (pc): Highest combined risk category → correct: Sports & Outdoors (−18% steepest decline on smallest revenue base)
Step 5 — Advisory Brief — MCQ portion (AdvisoryBrief)
- Questions: 2 (embedded in the Advisory Brief component alongside the written justification)
- Categories: Recommended action selection; analytics type identification of the recommendation
- CLO targets: CLO-1 (prescriptive analytics), CLO-4 (data-driven decision)
- Action question: Select the strongest prescriptive recommendation → correct: "Investigate return rate root cause; reduce Sports & Outdoors Q1 inventory 15–20%; reallocate South/West marketing budget to North/East"
- Analytics type question: Classify the recommendation → correct: "Prescriptive"
Step 6 — Tool Selection (ToolSelector)
- Questions: 4
- Categories: Analytics tool selection for specific workflow tasks; ML readiness (AI question)
- CLO targets: CLO-6 (apply analytics tools to real-world scenarios)
- t1: Interactive dashboard for regional managers → correct: Tableau or Power BI
- t2: Role of BAN2100 tools in a forecasting workflow using ARIMA → correct: Tableau/Power BI visualize output; ARIMA runs in Python/R
- t3: Drafting a 200-word executive narrative summary → correct: ChatGPT
- t4 (AI question): Prerequisite for replacing ARIMA with deep learning demand forecasting → correct: labeled outcome records (decision-outcome feedback loop) required for supervised ML
4. AI Question
Category label: "Technology Strategy — ML Readiness" (teal badge)
Step: Step 6 (Tool Selection), question t4
Question angle: Tesco's head of analytics wants to replace the Panel C ARIMA model with a deep learning demand forecasting system using 3 years of transaction data. Learners must identify which prerequisite must be confirmed before commissioning — the correct answer is that labeled outcome records (e.g. whether prior inventory decisions led to stockouts or markdowns) are required because deep learning demand forecasting is supervised ML and needs a decision-outcome feedback loop, not just transaction volume history.
Why appropriate for discipline: Analytics courses must include at least one question requiring classification or evaluation of an analytical approach, not just interpretation. An ML readiness question directly tests whether students understand the data prerequisites that distinguish supervised ML from traditional statistical models — a core data analytics literacy skill at BBA level.
LD approval: Approved 2026-04-24.
5. Media Elements
- ScenarioOverview — Panel A table: Descriptive metrics table with year-over-year comparison and flag indicators (Return Rate flagged)
- ScenarioOverview — Panel B table + Recharts bar chart: Regional revenue vs marketing spend; horizontal bar chart (Chart B.1) showing revenue-to-spend ratio by region, with South and West flagged red; average reference line overlaid
- ScenarioOverview — Panel C table + Recharts grouped bar chart: Q1 FY2024 category forecast vs Q4 FY2023 actual; grouped bar chart (Chart C.1) with Sports & Outdoors flagged; ARIMA model note
- SpotTheError — Recharts scatter chart: Deliberately flawed scatter plot presenting r = 0.42 as a "strong correlation" between marketing spend and regional revenue; trend line overlaid; learner must identify that r = 0.42 is weak-to-moderate and that the data contradicts a marketing-driven explanation (similar budgets, revenue ranges $1.7M–$4.1M)
6. Industry Context Panel
4-item 2×2 grid in ScenarioOverview, placed between Panel A (descriptive metrics table) and Panel B (first Recharts chart) per Standards §required components. Items are present-tense pressures shaping retail analytics decisions:
- AI-Driven Demand Forecasting — labelled-outcome data prerequisite for replacing ARIMA with deep learning (frames the AI question in Step 6)
- Returns Margin Erosion — return-rate flag in Panel A is a P&L issue not just operational
- Marketing Attribution Fragmentation — iOS / cookie deprecation degrading attribution; frames the Panel B diagnostic interpretation
- Regional Cost Pressures — UK regional cost inflation asymmetry; frames why South/West underperform Panel B despite similar marketing spend
Updating any panel item must not change the correct answer to any MCQ (CLAUDE.md invariance rule). Company identifier (Tesco ticker pill, sector, description) is rendered in the scenario brief at the top of the same component.
7. Written Brief
Advisory Brief Justification (Step 5)
- Prompt: "In 30+ words, explain which data evidence (from which panel) most strongly supports your recommended action and why it is more appropriate than the alternatives."
- Minimum words: 30
- CLO mapping: CLO-3 (pattern analysis applied to evidence selection), CLO-4 (data-to-decision reasoning)
Analyst QA Response (Step 4)
- Prompt: "Your analysis — what is wrong, and why does it matter?" (with guidance to identify the error in the chart title and explain what a decision-maker might wrongly conclude)
- Minimum words: 20
- CLO mapping: CLO-3 (data interpretation, recognizing analytical error)
8. Company Profile
- Name: Tesco
- Ticker: TSCO
- Sector: Large-format Grocery & General Merchandise Retail
- Description (as rendered): "Tesco operates retail stores across four UK trading regions (North, East, South, West), offering groceries, Home & Kitchen, Electronics, Apparel, and Sports & Outdoors goods." — 48 locations, Q4 FY2023 reporting period
- Source:
docs/simulation-data/global_simulation_company_database.json (entry C0434 — all dashboard figures are plausible fiction, not real Tesco data)
9. Grading Configuration
- MCQ weight: 0.80
- Written weight: 0.20
- Degree level: Undergrad (BBA)
- Mastery threshold: ≥90% (Mastery), 70–89% (Developing), <70% (Insufficient)
- CLO demonstrated threshold: ≥70%
- Pillar weights: Domain 55%, Reasoning 25%, Contribution 20%
10. Compliance Checklist
- ✅ Step locking (isAccessible) — all 6 steps locked behind prior completion; QA mode bypasses
- ✅ QA bypass (?qa=nxf-qa-8f2k) — implemented via useEffect on URLSearchParams
- ✅ Ticker pill in header —
{company.ticker} rendered as monospace pill in h1 breadcrumb
- ✅ US English spelling — no British spellings observed in component copy
- ✅ Warm learner-facing tone — feedback uses "Not quite!", "You got X of Y — review the explanations and give it another go!", "✓ Correct" patterns throughout
- ✅ Version string in breadcrumb —
v2.0 rendered in font-mono in breadcrumb line
- ✅ No external logos (no Clearbit) — no company logo used; ticker pill used instead (compliant)
- ✅ AI question included — t4 in ToolSelector (Step 6): "Technology Strategy — ML Readiness" (teal badge); supervised ML prerequisite angle; approved by LD 2026-04-24
- ✅ Recharts chart in ScenarioOverview — Panel B horizontal BarChart and Panel C grouped BarChart both use ResponsiveContainer
- ✅ SpotTheError component — present as Step 4 (Analyst QA); scatter chart shows r=0.42 mislabeled as "strong correlation"
- ✅ Analyst QA step — present (Step 4, id: "error")
- ✅ ← Back button on every step — Back buttons added to all steps (classify→scenario, patterns→classify, error→patterns, recommendation→error, tools→recommendation) — fixed 2026-04-24
Approval notes
- The advisory brief completion logic requires both the selected action AND the analytics type to be correct before
onComplete() fires — this is stricter than a pure MCQ pass and effectively combines two MCQ checks with a written component in one step. An LXD reviewer would have wanted this interaction model approved upfront.
- Written justification minimum is 30 words (advisory brief) and 20 words (analyst QA) — below the typical written section threshold. LD sign-off on these minimums was not captured pre-build.
- The Analyst QA written response is graded together with the advisory brief as part of the written component (20% weight).
- AI question (t4) added to ToolSelector 2026-04-24 — "Technology Strategy — ML Readiness" category, teal badge, supervised ML prerequisite angle. Approved by LD same date.