## 📦 Challenge 01 — Deliverables

**From Messy Files to a Reproducible Data Foundation**

This challenge emphasizes **clarity, reproducibility, and communication** — the three pillars of trustworthy data science.
By the end, participants will produce a fully documented workflow showing how disorganized spreadsheets become a structured, analytics-ready relational database — plus a simple baseline model to validate its usefulness.

---

### 🧩 Required Deliverables

#### 📄 Problem Statement & Notebook Stub

A concise description of the business context, data goals, and success criteria.
Include a Jupyter notebook skeleton (`Setup`, `Ingest`, `EDA`, `Modeling`, `Results`) with brief markdown prompts outlining what belongs in each step.

#### 📚 Data Dictionary

A clear, human-readable table listing **every column** across all provided CSVs.
Include:

* Data type and valid range
* Example values
* Notes on missingness or inconsistent formats
* Business meaning (what each field actually represents)

#### 🗺️ Entity–Relationship Diagram (ERD)

A conceptual visualization showing entities, keys, and relationships.
Highlight normalization decisions — e.g., why `Products` and `LineItems` were separated — and justify any compromises between strict 3NF and analytical efficiency.

#### 🧱 Normalized Schema + DDL Scripts

SQL scripts (`ddl.sql`, `indexes.sql`) defining the normalized schema in PostgreSQL.
Document naming conventions, key constraints, and indexes optimized for analytical joins (e.g., date, store, product).

#### ⚙️ ETL Pipeline

Python scripts or notebooks that:

* Parse, clean, and load data into PostgreSQL
* Enforce type coercion and referential integrity
* Log row counts before/after loads
* Capture rejected or anomalous rows for audit

#### 📊 Exploratory Data Analysis (EDA)

A notebook visualizing key business dimensions such as category mix, store performance, or temporal trends.
Include at least 3–5 visuals (histograms, trends, correlations) and a short written interpretation that connects structure to insight.

#### 🤖 Baseline Model

A simple, defensible predictive model demonstrating analytic readiness — e.g., a regression or time-based forecast of daily store sales.
Report at least one metric (MAE, MAPE, RMSE) and note any assumptions or data limitations.

#### 🖼️ Presentation Deck

A 6–10 slide summary (PowerPoint or PDF) covering:

* Data pipeline overview
* ERD and schema design
* EDA highlights and baseline model results
* Key takeaways and next steps for **Challenge 02 (Errors → Integrity)**

---

### 🗂️ File Structure Checklist

| Folder       | Files / Outputs                            |
| :----------- | :----------------------------------------- |
| `docs/`      | `architecture.md`, ERD image, design notes |
| `sql/`       | `ddl.sql`, `indexes.sql`                   |
| `scripts/`   | `load_data.py`, ETL logs                   |
| `notebooks/` | `01_eda.ipynb`, `03_model_baseline.ipynb`  |
| `reports/`   | `executive_summary.pdf`                    |
| `slides/`    | `presentation_deck.pptx`                   |

---

### ✅ Outcome

A complete, validated, and **well-documented relational data foundation** — the backbone for all future stages:
data-quality testing, imputation, forecasting, and ROI modeling.