Search for a command to run...
Exploratory Data Analysis is the open-ended phase where you get to know a dataset before committing to a specific claim or chart. It isn't about producing polished output — it's about building an accurate mental model of what the data contains, what's missing, and what's surprising.
Before anything visual, look at counts, ranges, and basic distributions for every column. This catches obvious problems fast — a "percentage" column that goes to 250, a date range that starts in 1970 by mistake — and gives you a baseline to compare deeper findings against.
A histogram or scatter plot reveals structure that a table of numbers hides: bimodal distributions, clusters, and nonlinear relationships all jump out visually in a way that summary statistics alone can miss.
Cross-tabulate categorical columns, correlate numeric ones, and segment metrics by obvious groupings like region or customer type. Relationships between variables are usually where the interesting story is hiding, not in any single column on its own.
Go in with questions, not conclusions. If every chart confirms exactly what you expected, you're probably not looking hard enough — or you're unconsciously steering toward the answer you wanted.
EDA is a mindset as much as a set of techniques: stay curious, look at the data from multiple angles, and resist the urge to jump to a conclusion before you've actually explored what's there.