Key Takeaways: Capstone Project

This is your reference card for Chapter 35 — the capstone. Use it as a checklist during your project to make sure you're hitting all the required elements.

Key Concepts

• The capstone is integration, not new learning. You already have every skill you need. The capstone tests whether you can combine those skills into a coherent investigation — from question to conclusion, with every step documented and justified.

• A capstone is not a homework assignment. Homework has a predetermined question, provided data, and a known answer. The capstone requires you to make every decision yourself: the question, the data strategy, the analytical approach, the interpretation, and the communication. That's what makes it hard. That's also what makes it valuable.

• Communication is half the project. A capstone that runs perfect models but has no narrative, no interpretation, and no written conclusions is incomplete. The notebook should read like a story, not a code dump.

• Limitations and ethics are required, not optional. The capstone rubric allocates 10 points (out of 24) to critical reflection. This reflects a genuine belief: a data scientist who doesn't think about what their analysis can't tell you, or about the human consequences of their work, is not yet a complete data scientist.

• Done is better than perfect. A completed capstone at 85% quality is infinitely better than an unfinished one at 100%. Set a deadline and meet it.

Three Capstone Options

Required Deliverables

The Notebook (10 sections)

1. Title and Abstract ............... 200-300 words
2. Introduction and Motivation ...... 500-800 words
3. Data Description ................. 400-600 words + code
4. Data Cleaning .................... 600-1000 words + code
5. Exploratory Analysis ............. 800-1200 words + 4-6 charts
6. Statistical Analysis/Modeling .... 800-1200 words + 2-4 charts
7. Findings and Conclusions ......... 500-800 words
8. Limitations and Future Work ...... 300-500 words
9. Ethical Reflection ............... 300-500 words
10. References ...................... Complete citations

The Repository

project-name/
    README.md           # Overview, findings, reproduction instructions
    notebooks/          # The capstone notebook
    data/raw/           # Original data (or download instructions)
    data/processed/     # Cleaned data
    figures/            # Saved key visualizations
    requirements.txt    # Dependencies with versions
    .gitignore          # Exclude checkpoints, caches, large files

The Rubric (24 points total)

Milestone Checklist

Week 1: Foundation

• [ ] Choose project option
• [ ] Write research question (1-2 sentences)
• [ ] Inventory existing work / identify data sources
• [ ] Set up GitHub repository
• [ ] Load and inspect all data
• [ ] Complete data cleaning with documented decisions

Week 2: Analysis

• [ ] Create 4-6 exploratory visualizations with interpretations
• [ ] Run 2+ formal statistical analyses
• [ ] Build 2+ models with proper evaluation
• [ ] Draft findings section

Week 3: Polish

• [ ] Write introduction, abstract, and data description
• [ ] Write conclusions, limitations, and ethical reflection
• [ ] Polish all visualizations
• [ ] Remove debugging cells and clean notebook
• [ ] Kernel > Restart & Run All

Week 4: Finalize

• [ ] Write final README
• [ ] Create requirements.txt
• [ ] Peer review (give and receive)
• [ ] Address feedback
• [ ] Final Restart & Run All
• [ ] Submit and celebrate

Common Pitfalls

Self-Review Checklist

Before you call it done, verify:

• [ ] A non-data-scientist can understand the introduction and conclusions
• [ ] Every chart has a title, labels, and written interpretation
• [ ] At least 3 analytical decisions are documented with reasoning
• [ ] The notebook runs cleanly (Kernel > Restart & Run All)
• [ ] Limitations are specific, not generic
• [ ] The ethical reflection engages with real human dimensions
• [ ] The README is complete and informative
• [ ] You'd be proud to show this to a hiring manager

What You Should Be Able to Do Now

• [ ] Execute a complete data science investigation end-to-end
• [ ] Integrate cleaning, exploration, visualization, statistics, and modeling into one coherent narrative
• [ ] Document analytical decisions with clear reasoning
• [ ] Communicate findings to a non-technical audience
• [ ] Reflect honestly on limitations and ethical dimensions
• [ ] Produce a portfolio-quality Jupyter notebook

If you've completed the capstone, take a moment to appreciate what you've done. You started this book not knowing what data science was. Now you've completed a full investigation — your own investigation, with your own question, your own decisions, and your own conclusions. That's real. Chapter 36 celebrates how far you've come and maps where you can go next.