Part I
Language of Data: Reading the Business in Rows and Columns
Before any model, read the table
This part builds the literacy a manager needs before running any model, because most analytical disasters are settled at the level of the table, not the model. Chapter 1 covers how to read a dataset the way you would read a financial statement — recognizing how a row's grain, a table's shape, and a column's measurement type cap what any analysis can honestly claim. Chapter 2 puts that reading into motion, translating familiar spreadsheet moves into the GROUP BYs, JOINs, and reshapes that scale, while exposing where business reality leaks in: joins that double revenue, averages that secretly weight every store equally, deletes that fabricate a trend. Together they move from passively receiving a dataset to actively governing one, and leave a reusable data-and-metric brief for the rest of the book to build on.
2 chapters · 9 articles
What you’ll learn
- Diagnose a table's grain, shape, and column types before trusting any number it produces
- Translate familiar spreadsheet operations into SQL, dplyr, and pandas that run identically and scale
- Catch the conceptual errors that survive clean syntax — exploding joins, average-of-ratios, survivorship bias from hard deletes
- Convert raw columns into governed metrics with definitions that hold up under scrutiny
- Assemble a reusable row-contract, metric-definition, and join-map brief that anchors the rest of the book
Chapters in this part
Before you ask what model to run, ask what one row means, what shape the table is in, and what kind of column you are looking at.
Spreadsheet operations, written down as reproducible code — and the conceptual errors that hide inside correct-looking queries.