8.1 EDA is an investigation

Exploratory data analysis is the disciplined process of learning what the data contain, how they were produced, which problems they have, and which relationships deserve further analysis. It is not a gallery containing one default graph for every variable.

A useful EDA has two audiences. The analytical team needs detailed evidence about quality, definitions, distributions, and feasibility. Decision makers need a clear account of which patterns and limitations affect the problem. The EDA report should serve both purposes without confusing exploration with final explanation.

EDA should help the team:

• verify structure, scope, and coverage;
• detect errors and limitations;
• understand important variables and subgroups;
• identify time, location, and relationship patterns;
• test whether proposed methods are feasible;
• refine the question map; and
• decide what should not be analyzed.