New
Storytelling with Data
A Practical Guide to Communicating Effectively with Data Visualizations and Charts
2025
View →
Pages
257
Published
2023
Data Analytics
✨ New
Data Science: The Hard Parts
Techniques for Thinking Analytically and Solving Real Data Problems
Build the analytical reasoning and problem-solving instincts that separate competent data scientists from indispensable ones.
Most data science books teach you tools. This one teaches you how to think. Daniel Vaughan's Data Science: The Hard Parts focuses on the analytical judgment, communication habits, and problem-framing skills that intermediate practitioners rarely get from tutorials. Across 257 pages, you'll learn to decompose ambiguous problems, avoid common reasoning traps, and deliver results that actually land with stakeholders.
About this book
You already know how to run a regression. You can write a Pandas pipeline and explain a confusion matrix. But somewhere between your notebook and the boardroom, the analysis loses its impact, the recommendation gets ignored, or the project quietly dies. That gap is not a tooling gap. It is a thinking gap.
Data Science: The Hard Parts by Daniel Vaughan targets exactly that gap. Rather than covering another library or another algorithm, it focuses on the analytical habits and reasoning skills that experienced practitioners build slowly, through costly mistakes, and that most books never address directly.
Vaughan draws on years of applied work to lay out a structured way to approach the problems that trip up intermediate data scientists: decomposing vague business questions into answerable ones, choosing the right metric for a situation without over-optimizing it, and communicating uncertainty in a way that earns rather than erodes trust.
The book covers topics that sit in the uncomfortable middle ground between statistics and strategy:
- Metric design and the traps that come from measuring the wrong thing
- Decomposition frameworks for breaking down complex analytical questions
- Storytelling with data in ways that drive decisions, not just nods
- Thinking about causality carefully without a full causal inference toolkit
- Avoiding the subtle biases that invalidate otherwise clean analyses
- Translating stakeholder requests into well-scoped, answerable problems
This is not a reference book. It is a thinking companion for the practitioner who has the technical foundation and wants to become more rigorous, more persuasive, and more useful inside a real organization. Every chapter is grounded in realistic scenarios, not toy datasets or contrived examples.
At 257 pages, the book respects your time. It is dense with practical reasoning rather than padded with background theory you already know. Read it cover to cover or treat individual chapters as targeted coaching when a specific situation arises.
🎯 What you'll learn
- Decompose ambiguous business questions into precise, answerable analytical sub-problems
- Design metrics that measure what actually matters and resist gaming or over-optimization
- Identify reasoning traps and cognitive biases that silently corrupt otherwise sound analyses
- Communicate uncertainty and caveats in ways that build credibility rather than confusion
- Apply lightweight causal thinking to observational data without overstating conclusions
- Structure analytical narratives so that recommendations reach decisions, not just slide decks
- Scope data projects realistically so that they deliver value within organizational constraints
👤 Who is this book for?
- Data scientists with one to four years of experience who have the technical basics down and want to sharpen their analytical judgment
- Analytics engineers or BI developers who regularly translate business questions into data problems and want a more structured approach
- Data analysts moving into senior or lead roles who need to communicate findings more persuasively to non-technical stakeholders
- Applied ML practitioners who notice their models land well technically but struggle to generate real organizational action
- Product or growth analysts who work closely with decision-makers and want to improve how they frame, scope, and present their work
Table of contents
-
01
The Thinking Gap
Vaughan sets up the central argument: technical skill is necessary but not sufficient, and most data scientists hit a ceiling not because of what they cannot compute but because of how they frame and communicate problems.
-
02
Decomposing Hard Problems
You practice a structured decomposition method for breaking vague business questions into specific, measurable sub-questions that your data can actually answer.
-
03
Metric Design and Its Traps
You learn how to choose and define metrics carefully, and how poorly designed metrics create perverse incentives, misleading conclusions, and organizational dysfunction.
-
04
Causality Without a Causal Toolkit
This chapter gives you a practical vocabulary and set of checks for reasoning about cause and effect in observational data, without requiring a formal econometrics background.
-
05
Bias and the Validity of Your Analysis
You identify the most common analytical biases, from selection bias to survivorship bias, and work through realistic scenarios where each one silently invalidates a conclusion.
-
06
Communicating Uncertainty
You develop techniques for presenting probabilistic and uncertain findings to non-technical audiences in ways that inform rather than paralyze decision-making.
-
07
Storytelling That Drives Decisions
Vaughan walks through the structure of analytical narratives that actually change behavior, contrasting them with the presentation patterns that produce polite applause and no follow-through.
-
08
Working Inside Organizations
You apply the book's frameworks to the political and social realities of working inside a company, including scoping projects realistically, managing stakeholder expectations, and knowing when to push back.
Frequently asked questions
Do I need a strong statistics background to get value from this book?
No. The book assumes you are a working data practitioner with basic statistical literacy, but it does not require graduate-level theory. The focus is on applied reasoning, not mathematical proofs.
Is this book heavy on code and technical examples?
No. This is primarily a conceptual and reasoning-focused book. Code snippets appear where they support a point, but the core value is analytical frameworks and mental models, not syntax.
How current is the material for 2024 and beyond?
The book was published in November 2023. Because it focuses on durable analytical thinking skills rather than specific libraries or platforms, the material ages well and remains relevant to current practice.
Who is this book not suitable for?
If you are just starting out in data science and still building foundational Python or statistics skills, this book will be more useful once you have some practical experience to anchor it to. It is written for intermediate practitioners.
Is there companion code or a GitHub repository?
Check the publisher's page on O'Reilly Media for any supplementary materials associated with the book. The content is primarily conceptual, so the book itself is the main resource.
You might also like
New
New
Fundamentals of Data Engineering
A practical guide to the complete data engineering lifecycle, from ingestion to serving
by Joe Reis, Matt Housley
2022
View →
New
Data Analysis with Python and PySpark
A hands-on guide to scalable data analytics using Python and PySpark
2022
View →
New