BADM 557 - Business Intelligence
Program-level details: See program/curriculum.md
| Credits: 4 | Term: Summer 2027 (Weeks 1-8) | Instructor: Gautam Pant |
Course Vision
This course teaches students to build BI systems by reasoning forward from a business framework to a defensible decision input, with AI methods deployed at the points where they add measurement power.
Core insight: Most BI work fails because the metrics on dashboards do not actually measure the constructs that matter to the strategic question at hand — the failure is upstream of both dashboard design and model quality. This course addresses the upstream problem.
Pedagogical flow: Course material maps to one or more stages of:
Strategic / Operational Framework → Constructs → Data → Measurement → Validation → Decision Inputs
Recording timeline: May 2026 and July 2026 (Gautam developing and delivering solo).
Key assumptions about students: By Summer 2027, students will have completed BADM 554, BDI 513, and FIN 550 — they are fluent in GitHub, Python, VS Code + Copilot, and Colab. No tool onboarding needed.
Learning Outcomes (L-C-E Framework)
Literacy (Foundational Awareness)
- L1: Explain why BI fails upstream of dashboards and models — at the construct definition stage
- L2: Identify the distinction between constructs (latent concepts) and proxies (measurable stand-ins)
- L3: Recognize AI-specific validity threats: prompt sensitivity, distribution shift, hallucination, calibration drift
Competency (Applied Skills)
- C1: Operationalize a management framework (Porter’s Five Forces, value chain) into measurable constructs using WRDS and SEC data
- C2: Apply text-based measurement techniques: TF-IDF, Loughran-McDonald dictionaries, LLM embeddings
- C3: Build graph-based constructs from corporate network data (BoardEx centrality measures)
- C4: Link data across SEC/EDGAR, BoardEx, and Compustat with provenance tracking
Expertise (Advanced Application)
- E1: Build a deployed web application that measures, validates, and presents BI constructs as decision inputs
- E2: Design and execute a validity stack: content validity, construct validity, and robustness checks
- E3: Translate validated measurements into decision inputs under uncertainty for a real firm
Week-by-Week Breakdown
| Week | Live Session (Content) | Tutorial (Project-Based) |
|---|---|---|
| 1 | Framing BI as Measurement: Constructs and Proxies | WRDS orientation; firm and peer-set selection |
| 2 | Constructs I: Measuring Value Chain Activities with Dictionaries and Embeddings | 10-K text extraction; TF-IDF and Loughran-McDonald scoring |
| 3 | Constructs II: Graph Thinking and Centrality Measures for Five Forces | BoardEx network construction; centrality as construct |
| 4 | Data: Sourcing, Linking, and the Provenance Problem | Linking across SEC, BoardEx, Compustat; first web app deployment |
| 5 | Measurement I: Turning Text into Numbers | Three measurements compared; build comparison view in app |
| 6 | Measurement II: Networks, Relationships, and Multi-Source Constructs | Composite text+network construct; build construct dashboard in app |
| 7 | Validation: The Validity Stack and Robustness Checks for AI Measurements | Robustness check matrix; build validation page in app |
| 8 | Decision Inputs: From Validated Measurements to Decisions Under Uncertainty | Final app deployment |
Team Project: BI Measurement Web App (Team of 3, spans Weeks 1-8)
Problem Statement: Teams select a real firm and a strategic question in Week 1. Over 8 weeks, they build a deployed web application that traces a management framework through to validated construct measurements and decision inputs. The app is built incrementally: first deployment Week 4, construct dashboard Week 6, validation page Week 7, final decision-inputs view Week 8.
Data sources used: WRDS (firm-level), SEC/EDGAR (10-K filings), BoardEx (governance and network), Compustat (financial). Teams link across these sources with documented provenance.
Weekly Assignments (35%)
Individual assignments that build foundational skills and feed into the team project.
| Week | Assignment | Format |
|---|---|---|
| 1 | Construct framing memo: choose a firm and strategic question; define 2-3 constructs and their proxies | Written memo (1-2 pages) |
| 2 | Text scoring exercise: run TF-IDF and Loughran-McDonald scoring on one 10-K; interpret business implications | Notebook + short memo |
| 3 | Network analysis memo: build a BoardEx centrality measure for one firm; interpret as a Five Forces construct | Notebook + short memo |
| 4 | Data linking report: document how you joined SEC, BoardEx, and Compustat for your team’s firm; flag provenance issues | Written memo (1-2 pages) |
| 5 | Measurement comparison: apply three text-based measurement approaches to the same construct; compare results | Notebook + comparison memo |
| 6 | Composite construct memo: design and implement a composite text+network construct; interpret business meaning | Notebook + memo (2 pages) |
| 7 | Validation memo: apply robustness checks to one team construct; document failure modes found | Written memo (2 pages) |
Rubric (4 dimensions):
| Dimension | Excellent (A) | Proficient (B) | Developing (C) |
|---|---|---|---|
| Business Understanding | Deep understanding of context + constraints | Good understanding | Surface-level reading |
| Analytical Rigor | Systematic exploration, clear insights | Adequate exploration | Shallow analysis |
| Communication | Professional, clear, executive-ready | Adequate write-up | Unclear or incomplete |
| AI Attribution | Proper disclosure of AI use with reflection | AI use noted | Missing or vague attribution |
Project Milestones (25%)
Team deliverables that build toward the final project.
| Week | Milestone | Deliverable |
|---|---|---|
| 2 | Project proposal | 1-page proposal: firm selected, strategic question, 2-3 constructs, data sources, team roles |
| 4 | First app deployment | Working web app: linked data (SEC + BoardEx + Compustat), at least one construct view, provenance documentation |
| 6 | Construct dashboard | App updated with: composite text+network construct, measurement comparison view, construct dashboard page |
Rubric (4 dimensions):
| Dimension | Excellent (A) | Proficient (B) | Developing (C) |
|---|---|---|---|
| Progress | On track, clear trajectory toward final deliverable | Adequate progress | Behind or unfocused |
| Technical Quality | Rigorous approach, appropriate methods | Functional work | Technical issues |
| Team Coordination | Clear evidence of shared work and planning | Adequate collaboration | Uneven contribution |
| Incorporation of Feedback | Substantively addresses prior feedback | Some adjustments | Ignores feedback |
Final Project Deliverable (15%)
The culminating team submission in Week 8.
Deliverables:
- Deployed web application with: construct measurement view, measurement comparison, construct dashboard, validation/robustness page, decision inputs view
- Executive brief (1 page: strategic question, 3 key validated findings, decision recommendation)
- Technical documentation (5-7 pages):
- Firm and strategic question framing
- Construct definitions and proxy choices
- Data sourcing and provenance log
- Measurement approach comparison
- Validity stack and robustness check results
- Decision inputs under uncertainty + limitations
- Peer evaluation of team contributions
- GitHub repo with all code + app
Rubric (5 dimensions):
| Dimension | Excellent (A) | Proficient (B) | Developing (C) |
|---|---|---|---|
| Dashboard | Executive-ready, clear story, interactive | Functional, mostly clear | Needs polish |
| Analysis | Deep insights, data-driven recommendations | Good analysis | Shallow or missing insights |
| Business Acumen | Understands constraints, realistic recommendations | Shows business sense | Generic or impractical |
| Documentation | Clear guide for stakeholders, professional formatting | Adequate explanation | Minimal docs |
| Technical Rigor | Sophisticated visualization, robust clustering, clean code | Standard features used well | Basic functionality |
Oral Defense (20%)
Team oral defense: 15-min live presentation to mock leadership team + Q&A. Each team member must answer questions individually.
Rubric (4 dimensions):
| Dimension | Excellent (A) | Proficient (B) | Developing (C) |
|---|---|---|---|
| Presentation Clarity | Confident, clear narrative, well-structured | Adequate delivery | Unclear or disorganized |
| Q&A Handling | Handles tough questions well, articulates trade-offs | Reasonable responses | Unprepared or evasive |
| Individual Contribution | Can speak to any part of the project with depth | Knows own section | Limited understanding |
| Business Judgment | Realistic recommendations, acknowledges limitations | Shows business sense | Generic or impractical |
AI Role in the Course
AI methods are deployed at specific stages of the Framework → Constructs → Data → Measurement → Validation → Decision Inputs flow:
Constructs stage (Weeks 2-3): LLMs and embeddings measure latent constructs from unstructured text (10-K filings) and networks (BoardEx) — constructs previously inaccessible without AI. AI is used throughout as a coding assistant.
Measurement stage (Weeks 5-6): AI models become the measurement instrument, inheriting all validity concerns of any measurement instrument. Students must treat AI-generated measurements with the same skepticism as any instrument.
Validation stage (Week 7): AI-specific failure modes are diagnosed alongside classical validity threats: prompt sensitivity, distribution shift, hallucination, calibration drift.
Studio Session Topics:
- Week 1: Constructs vs. proxies — the upstream failure in BI
- Week 2: Text as data — how dictionaries and embeddings operationalize constructs
- Week 3: Networks as constructs — centrality measures and competitive position
- Week 4: Data provenance — what linking across sources requires
- Week 5: Measurement instruments — what it means for AI to be the instrument
- Week 6: Composite constructs — combining text and network signals
- Week 7: Validity stacks — what “checking your work” means for AI measurements
- Week 8: From measurement to decision — how to present uncertainty to decision-makers
Assessment Summary
| Component | Weight | Notes |
|---|---|---|
| Weekly assignments | 35% | Individual; memos and notebooks on constructs, text/network measurement, linking, validation |
| Project milestones | 25% | Team; proposal (Wk 2), first app deployment (Wk 4), construct dashboard (Wk 6) |
| Final project deliverable | 15% | Team; deployed app + executive brief + technical documentation |
| Oral defense | 20% | Team; 15-min presentation + individual Q&A |
| Studio participation | 5% | Weekly attendance + peer discussion |
No traditional exam. One major team project built incrementally through weekly milestones, with individual assignments reinforcing each week’s concept.
AI Usage Levels (AIAS)
| Assessment | AIAS Level | AI Permitted |
|---|---|---|
| Weekly assignments | 2 | AI as coding assistant and writing aid — with attribution; students must understand and explain all output |
| Project milestones | 2 | AI for data exploration, app scaffolding, construct narrative drafting — with attribution |
| Final project deliverable | 3 | AI as collaborator for executive brief and measurement interpretation — with full disclosure |
| Oral defense | 0 | No AI |
| Studio participation | 1 | AI for exploration during exercises |
Technology Stack
- Data: WRDS (firm-level), SEC/EDGAR (10-K text), BoardEx (governance/networks), Compustat (financial)
- Text analysis: Python — TF-IDF (scikit-learn), Loughran-McDonald dictionaries, LLM embeddings
- Network analysis: Python — graph construction and centrality (networkx or similar)
- Web app: Python (Streamlit or similar) — deployed progressively Weeks 4-8
- IDE: VS Code with GitHub Copilot; Google Colab (browser alternative)
- Version Control: GitHub
- AI assistant: Claude/ChatGPT as coding assistant throughout
Prerequisites
Pre-Course Setup (Week 0)
Before Week 1, students must:
- Activate WRDS account (program-provided access — see program/tools.md)
- Confirm VS Code + Copilot setup (or Colab as fallback)
- Clone the course GitHub repository
WRDS orientation is covered in Week 1 tutorial; no prior WRDS experience required.
| Course Sequence: ← Agentic AI for Analytics | Next: BADM 576 — Data Science and Analytics → |