The Well-Grounded Data Analyst Solve messy data problems like a pro (David Asboth)（Z-Library）

# The Well-Grounded Data Analyst ## 【One-Line Pitch】 A practical, project-driven guide for junior data analysts who want to move beyond textbook examples and tackle the messy, ambiguous problems that actually show up in industry—complete with a repeatable process for learning new techniques fast. If you've finished an intro course or bootcamp and feel underprepared for real-world requests, this book is your bridge. ## 【Book Arc】 - **Opening (~0%–9%)**: Introduces the "results-driven approach," a seven-step framework that starts with defining a minimum viable answer before touching data. The author argues that most analysts jump too quickly into coding, and that spending more time on problem definition yields higher ROI. - **Early (~9%–19%)**: Walks through the first full project—extracting city-level customer spending from messy address data. Demonstrates the framework in action: handling missing values, cleaning addresses, creating derived columns, and presenting findings with clear caveats. - **Early (~25%–34%)**: Tackles a customer identification problem with three overlapping data sources (e-commerce database, CRM, and raw transactions). Covers merging datasets, handling guest checkouts, and building a unified customer data model. - **Middle (~38%–44%)**: Continues the customer project with deduplication using the Python Record Linkage Toolkit. Introduces data modeling concepts like event-based thinking and discusses when to use external libraries versus building from scratch. - **Middle (~44%–47%)**: Summarizes data modeling skills that transfer across projects—reshaping datasets, joining tables, cross-referencing, and deduplication—and begins a new project on identifying best-performing products, emphasizing the importance of strictly defining metrics before calculating them. ## 【Key Takeaways】 - **Start at the end** (Early): Define your minimum viable answer before exploring data. This prevents aimless coding and ensures your work directly addresses stakeholder needs. The author explicitly encourages spending more time on problem definition than feels instinctive. - **The seven-step framework is the backbone** (Early): Steps include establishing the problem, defining the minimum viable answer, identifying data, obtaining data, doing the work, presenting, and iterating. This process is applied consistently across all projects, making it a transferable skill. - **Messy data requires explicit assumptions** (Early): When extracting city names from addresses, the author chose a substring-matching approach and categorized unmatched records as "OTHER." Documenting these choices is critical for transparency and reproducibility. - **Data modeling is a core analyst skill** (Middle): Creating clean, deduplicated, restructured datasets from raw sources makes even basic tasks like counting easier. The author recommends thinking in terms of business events (e.g., "customer buys product") rather than just tables. - **Use existing tools when appropriate** (Middle): For record deduplication, the author uses the Python Record Linkage Toolkit rather than building algorithms from scratch. You don't need deep algorithm knowledge if you understand expected outputs and can debug issues. - **Present early, iterate often** (Early): Bring findings to stakeholders as soon as you have a minimum viable answer, even if the analysis isn't complete. This allows for course correction and avoids wasted effort on unnecessary precision. - **Metrics need strict definitions** (Middle): Words like "volume" or "revenue" sound self-explanatory but require precise definitions before calculation. The choice of metric fundamentally shapes what "best" means in any analysis. ## 【Reading Tips】 - **Read the opening chapters carefully** (~0%–9%): The seven-step framework is the intellectual core of the book. Internalize it before moving to projects, as every subsequent chapter references it. - **Skim the code, focus on decisions** (Early–Middle): The Python code is illustrative, but the real value is in the reasoning behind each step—why the author chose one approach over alternatives, and what trade-offs were made. - **Attempt projects before reading solutions**: The author explicitly encourages this. Each project has multiple valid approaches, and comparing your path to the author's reveals your own decision-making patterns. - **Use as a reference, not just a cover-to-cover read**: The book is designed to be opened at the chapter most relevant to your current challenge. The framework chapters and project chapters can be read independently. - **Pay attention to the "alternative decisions" notes**: These highlight where another analyst might have diverged, which is invaluable for understanding the range of acceptable approaches in real-world analysis. ## 【Coverage Limits】 This guide covers the book's core framework and the first several projects (address parsing, customer identification, and product performance). Excerpts do not cover later projects involving PDF data extraction, vague metrics, or inherited projects, though these are mentioned in the book's promotional material. ##