One supplied CSV. Agreed cleanup rules.
Know exactly
which rows
were removed.
A cleaned CSV, the original kept separately, and a removal log you can check. Start with a small file and a clear definition of a duplicate.
Synthetic evidence
The difference is in the details.
Matching an ID alone is not enough. In this example, every source field must match before a later row is removed.
One repeated row
| Source record | SKU | Item | Quantity |
|---|---|---|---|
| 2 | 001 | Workshop mug | 8 |
| 4 | 001 | Workshop mug | 8 |
Keep record 2. Remove record 4. All five source fields match, including the category and synthetic label not displayed here.
One ID, two different records
| Source record | SKU | Item | Quantity |
|---|---|---|---|
| 3 | 002 | Desk tray | 4 |
| 7 | 002 | Desk tray | 5 |
Keep both. The quantities differ. The cleanup does not decide which business record is correct.
The ZIP includes QA notes and checksums. Source numbering includes the header as record 1. Import SKU as Text in spreadsheet software to display leading zeros.
The first service
A small job,
with a defined finish.
One supplied, non-sensitive UTF-8 CSV, using commas as separators. Up to 5,000 data records and 30 columns, subject to sample review.
What you receive
- A cleaned CSV in the agreed column layout.
- A removal/change log and QA counts.
- The original file preserved separately.
- One correction within the agreed scope.
What we agree first
- Full-row matching after agreed normalization, keeping the first occurrence.
- Whitespace, header and any field-conversion rules.
- How to handle missing or conflicting values.
- Output layout, acceptance checks, fee and deadline.
Rules before removals
The processor normalizes headers and trims surrounding whitespace before comparing fields. Formula-like text is escaped for spreadsheet use. We review these behaviors against your file before work begins. The sample already uses clean headers and values; its five-column output is a checked projection of the fuller audit export.
Fuzzy matching, conflicting-record merges, PDF/OCR, workbook formulas, macros and connected systems require a separate feasibility review. Intake limits do not promise a processing speed or delivery time.
Start with the shape of the file
Describe the task.
Then agree the quote.
Include the file count, approximate rows and columns, the desired result and preferred timing. Begin with generic requirements. A redacted or fictional sample can follow through an agreed channel.
Do not send passwords or private customer records in an initial enquiry. No order, price or deadline is accepted by this page.