Data Cleaning with Power BI (Gus Frazer)（Z-Library）

【One-Line Pitch】 A practical, process-first guide to cleaning and preparing data inside Power BI, aimed at analysts and BI developers who already know the tool but want disciplined, repeatable ways to turn messy sources into trustworthy models. Read it if you want to move beyond ad-hoc fixes toward standards, Power Query/M fluency, and profiling-driven quality. 【Book Arc】 - **Opening (~0%–11%)**: Frames data quality as an organizational concern — integrity, ownership, accountability, and the roles (stewards, BI managers, leadership) who keep data trustworthy — before any tooling appears. - **Early (~11%–29%)**: Establishes data cleaning fundamentals and principles, then moves into hands-on Power BI work: removing duplicates, handling gaps, and reading column quality/distribution in Power Query Editor. - **Early–Middle (~29%–39%)**: Deepens the mechanics — applied-step naming and hygiene, DAX versus M roles, and M-language transformations (type conversion, multi-column transforms, parameters for deployment and conditional sources). - **Middle (~39%–50%)**: Extends into dataflows, folder/helper-query automation, and exploratory profiling, including fill down/up, fuzzy matching, and Python-script steps for custom transformations. - **Late (~50%+)**: Covers advanced and AI-assisted cleaning — dataflow actions (incremental refresh, ML model application), model training reports, and AI Insights via cognitive services — positioning automation as the endgame. 【Key Takeaways】 - **Data cleaning is a governance problem before it is a tooling problem** (Opening): standards, ownership, and leadership alignment determine whether cleaning sticks or becomes perpetual rework. - **Seven planning principles precede any transformation** (Early): the book stresses documenting intent and impact so cleaning is deliberate rather than reactive. - **Power Query Editor is the workhorse** (Early): duplicate removal, gap handling, and column quality/distribution checks are the everyday moves that fix completeness and accuracy. - **Applied steps deserve real names** (Early–Middle): descriptive, consistent, concise step labels make transformation logic auditable and reusable by others. - **Know when to use M versus DAX** (Middle): M handles extraction and shaping; DAX handles model-level calculations, measures, and business logic — mixing them up causes design debt. - **Parameters turn one-off queries into deployable assets** (Middle): parameterized sources and conditional logic avoid manual edits when promoting to production workspaces. - **Star schemas pay off at scale** (Early): the excerpts cite a flat-table query at ~29 seconds versus ~7 seconds in a star schema on comparable data, plus faster refreshes. - **Automation and AI extend the cleaner's reach** (Late): dataflows, incremental refresh, ML models, and AI Insights reduce manual preparation for large or complex datasets. 【Reading Tips】 - Skim the governance and role chapters if you already own data quality; deep-read the Power Query and M chapters, where the practical value concentrates. - Treat the M-language chapter as the hardest section — work the syntax, `let` blocks, and type-conversion examples alongside the book rather than reading passively. - Use the profiling/EDA material as a checklist: run column quality and distribution on your own datasets before and after cleaning to see the difference. - Note the star-schema versus flat-table performance discussion; it is the clearest argument in the excerpts for modeling discipline. - The AI/ML and dataflow sections are the most advanced; read them after you are comfortable with Power Query basics. 【Coverage Limits】 This guide is based on stratified excerpts covering roughly the first half of the book plus late-chapter material on dataflows and AI Insights; the excerpts do not cover the full text of the custom-functions, advanced-techniques, or later chapters in detail, so chapter-level specifics beyond those noted may be incomplete.