Synthetic CSV quality audit: missing, duplicate and invalid data

CSV quality audit — synthetic demonstration

Fictional data, not client work. AI-assisted preparation with checked arithmetic. This demonstrates a review method, not customer results or production throughput.

Input rows (ID / quantity): 001 / 10; 002 / 7; 002 / 8; blank / 4; 004 / oops; 005 / -2.

Agreed rules: exact text IDs must be present and unique; quantities must be integers. Negative quantities are flagged for owner review, not declared wrong. No values are changed.

Rows 2 and 3 repeat key 002 with different quantities. This is a key collision, not an exact duplicate. Row 4 lacks an ID. Row 5 has an invalid integer. Row 6 has a negative quantity requiring a business-rule decision.

Five distinct rows are flagged. Five numeric quantities total 27: 10 + 7 + 8 + 4 - 2. The invalid value is excluded and listed explicitly; 27 is a numeric subtotal, not a complete business total. Leading zeros are preserved.

The deliverable is a row-level issue register, column checks, a subtotal with exclusions, and questions for the data owner. No guessed repairs or silent source changes. Checks verified both totals and the exact flagged row identities 2, 3, 4, 5 and 6.