Data Cleaning Checklist

A dataset-specific cleaning checklist that catches structural errors, missing data, and outliers before they corrupt your analysis.

Best For

•A data analyst inheriting a messy CSV export from a legacy CRM and needing a documented cleaning process before running any analysis on it
•A data engineer setting missing-data thresholds and imputation rules for a customer table feeding a churn model
•A freelance analyst producing a client-facing audit trail showing exactly what was changed and why during data prep

Prompt Template

Generate a comprehensive data cleaning checklist for the dataset described as: [DATASET_DESCRIPTION]. The dataset contains [NUMBER_OF_ROWS] rows and [NUMBER_OF_COLUMNS] columns. It will be used for: [INTENDED_ANALYSIS]. Known data quality issues: [KNOWN_ISSUES]. The checklist should cover: (1) Structural checks, column names (standardized, no spaces, consistent case), data types (verify each column's type matches its content), duplicate rows (detection and removal criteria), (2) Missing data, for each column, specify: what % missing is acceptable, and the imputation or removal strategy, (3) Outlier detection, which columns to check for outliers, the method to use (IQR, Z-score, domain-specific rules), and the threshold for flagging vs. removing, (4) Consistency checks, cross-column validation rules (e.g., "end_date must be after start_date"), (5) Domain-specific checks, [DOMAIN_SPECIFIC_RULES, e.g., email format, phone number format, valid country codes], (6) Documentation, what to log for each cleaning action (original value, new value, reason), (7) Final validation, 3 sanity checks to run after cleaning to confirm the dataset is ready. Format as a step-by-step checklist with checkboxes.

Pro Tip

Set the acceptable missing-data percentage per column before writing any cleaning code, since a blanket 'drop rows with nulls' rule often deletes far more usable data than a column-specific threshold would.

Example Output

A sample of what this prompt produces once you fill in the placeholders.

Dataset: 84,000 rows, 22 columns of e-commerce order data. Email column: 3% missing, drop rows. Discount_code column: 61% missing, treat as 'no code used' rather than drop. Outlier check: order_total flagged using IQR method, values above $4,200 reviewed manually before removal. Final validation: row count matches source system count within 0.5%.