I will do data cleaning, deduplication and formatting on messy data files

Yes, and it is most of the work on a typical order. Different column names, different date formats, overlapping rows and one file with an extra column nobody documented. I map them to a single schema and tell you which rows came from where.

No. Above about 50,000 rows the work moves out of a spreadsheet and into Python and Pandas, which is faster and repeatable rather than slower. Half a million rows is an ordinary job. Send a sample and I will confirm the tier before you order.

You tell me, and that is the part worth agreeing before I start. Exact matches are easy. The useful work is fuzzy matching: Acme Ltd against ACME Limited, the same address written four ways, two records that differ only in whitespace. I show you the matches before merging.

Yes, every order comes with a changelog: rows in, rows out, what was merged, what was dropped and why, and which fields were standardised. You should not have to trust a clean file. You should be able to check it.

Included in Standard and Premium, in Python, commented, so you can run it on next month's export without ordering again. Basic delivers the cleaned file only, which is the right choice if this is genuinely a one-off.

Yes, and it is a common one. Mojibake from the wrong encoding, HTML entities left in text fields, stray whitespace and newlines inside cells, numbers stored as text with currency symbols attached. Send a sample and I will tell you what is recoverable.

Summary tables, pivots and charts are included in Standard and Premium, and they are useful for sanity-checking the clean. If you need real statistical work or modelling, that is a different job and I would rather point you elsewhere than do it adequately.

Yes, as a scheduled pipeline rather than a repeated order: the file lands, the script runs, the clean version appears where you need it. That is my data pipeline gig, not this one. Message me and I will send the link rather than charging you twelve times.

Tell me before you send it. I can work on it, and I will delete the source files after delivery on request, but I would rather know upfront than find names and addresses in a file I did not expect them in. Pseudonymised or sampled data is fine for scoping.

Send the file first, or a sample of it. Some jobs look simple and need fuzzy matching across three sources; others look terrifying and are one find-and-replace. I will tell you which tier it is, or quote a custom offer, before you spend anything.