Before analyzing a table, establish what each row represents and what its fields mean. A quick inspection can reveal problems that change the interpretation of later summaries.
Inspect before transforming
Check shape, column names, types, missing values, and representative rows. Validate date formats and numeric ranges. If an identifier should be unique, test that assumption rather than relying on its label.
Document cleaning decisions
Missing values can have different meanings. Decide whether to retain, exclude, or impute them based on the question and explain the choice. Review duplicates and inconsistent labels, then compare row counts before and after each step.
Summarize and reconcile
Use filters and groupings to produce focused summaries. Check whether totals reconcile with the included rows and whether denominators are appropriate for comparisons.
Keep the notebook reproducible
Use descriptive names, small functions for repeated work, and notes for assumptions. Avoid editing source files manually in a way that hides the transformation. Practice with public or synthetic data and follow applicable data permissions.