Data Quality Checklist: 7 Checks for Any CSV (2026)

Every wrong dashboard number and every confused model traces back to the same root cause: nobody checked the data first. Data health is simply whether a dataset is complete, consistent, and shaped the way you expect — and you can check it in minutes.

Use this 7-point checklist on any CSV before you analyze, import, or hand it to a stakeholder.

1. Completeness — how much is missing?

Start by counting nulls and blanks per column. A column that is 40% empty can't carry a conclusion. Look for whole rows that are mostly blank, too — they're often broken exports.

2. Uniqueness — are there duplicates?

Duplicate rows inflate totals and double-count customers. Check for fully duplicated rows and for repeated values in columns that should be unique, like IDs or emails.

3. Validity — do values match their type?

A 'price' column with the word 'free' in it, or a date stored as text, will break math and sorting. Confirm each column holds a single, expected data type.

4. Range & outliers — is anything impossible?

Negative ages, quantities in the millions, or a timestamp from 1900 are red flags. Quick statistics — min, max, mean, standard deviation — surface these instantly.

5. Consistency — is formatting uniform?

• Dates in mixed formats (01/02/2026 vs 2026-02-01).
• Phone numbers with and without country codes.
• Categories like 'US', 'USA', and 'United States' meaning the same thing.

6. Structure — are the rows well-formed?

Ragged rows (too many or too few columns), stray quotes, and encoding artifacts will derail any import. A structure check catches them before your database does.

7. A single health score

Rather than run six tools by hand, let a dataset health analyzer score completeness, duplication, type consistency, and outliers in one pass and tell you exactly which columns need attention.