A CSV cleanup can look correct while silently losing an identifier such as 0012. Here is a small, original fictional example you can run locally, plus a selectable-PDF table example to check against the source.

Make the rules explicit

Our sample starts with four inventory rows. We agree to trim the product column and remove only exact duplicates after that trim. The result has three rows; SKU 0012 stays a string. No missing values are guessed and no other columns are normalized.

sku,product,quantity 0012, Desk tray ,12 0013,Pen holder,8 0013,Pen holder,8 0014,Notebook,25

The free example repository includes the Python script, before/after CSV files and a JSON change report. Python 3 is the only dependency for the CSV example. It makes no network or API calls.

python3 clean_csv.py messy-inventory.csv cleaned-inventory.csv --trim product --drop-exact

The helper preserves the source, refuses uneven row widths and existing output files, and records source line numbers for changes. The example was run: input 4 rows, output 3 rows, one trim and one exact duplicate removed. This is deliberately a small helper, not an automatic decision about which business records are duplicates.

Verify the output, not just the command

  1. Confirm the row count and review every deletion.
  2. Compare identifier strings, including leading zeros. CSV cannot enforce Excel types: import identifier columns as Text.
  3. Check totals against the source when totals are meaningful.
  4. Review untrusted formula-like cell values before opening a CSV in a spreadsheet. This helper is not a general CSV security sanitizer.

PDF tables need a source check too

The repository also contains an original fictional selectable-text PDF and an Excel workbook showing its extracted table, plus a separate raw/cleaned/change-log example. Compare each row, column, SKU and quantity. Being able to select PDF text does not guarantee a clean table layout. Scans, handwriting and complex merges need different handling. Flag unclear cells instead of inventing a value.

Optional done-for-you pilots

Disclosure: I operate the following Fiverr services. The examples above are free and do not require an order.

  • Excel/CSV cleanup: $20, 2 days, one flat sheet up to 100 data rows and 10 columns, agreed cleanup and change log, one revision.
  • Selectable PDF tables to Excel/CSV: $20, 2 days, one PDF up to 5 pages, 500 table rows and 10 columns, one revision. Send a de-identified sample on Fiverr first so I can confirm extractability.

Only send authorized material with personal and confidential information removed. All communication, payment and delivery stay on Fiverr. AI assisted this article and the example code; the sample output was checked locally. These are fictional demonstrations, not client results.