ع
Learn Tracks Reference Guides Saved
build recipe

Clean up and analyze an Excel file

Hand Claude a messy spreadsheet and get back a clean version plus answers to the questions you actually care about — no formulas, no terminal.

easy ~10 min

you'll end up with A cleaned-up spreadsheet and a plain-English summary of what the numbers say.

The steps

  1. Open the folder with your file in Claude Desktop

    Put your spreadsheet (.csv or .xlsx) into a folder, then open that folder in the Claude Desktop app so Claude can read it.

    paste this to Claude
    There's a file called expenses.csv in this folder. Read it and describe what's in it, including anything that looks inconsistent or broken.

  2. Ask for a clean-up in the chat

    Tell Claude what 'clean' means for you — consistent dates, no duplicates, standardized categories — and to save the result as a new file.

    paste this to Claude
    Clean this up: make all the dates the same format, remove duplicate rows, standardize the category names, and save the result as expenses-clean.csv. Tell me what you changed.

  3. Review the changes before you accept

    Claude shows each edit as a visual diff. Read it, confirm nothing important was dropped, then accept — your original file stays untouched.

    paste this to Claude
    Show me a quick before-and-after for a couple of rows so I can confirm the clean-up looks right.

  4. Ask your real questions

    Now that it's clean, ask the questions you opened the file for in the first place — right in the chat.

    paste this to Claude
    Using expenses-clean.csv: what did we spend the most on this quarter, which category grew the fastest, and is there anything that looks like a mistake?

  5. Get it as a report

    Have Claude write the findings up so you can paste them straight into an email or doc.

    paste this to Claude
    Write a short summary of those findings in plain English — three bullet points a manager would care about.
tips & best practices
  • Always ask 'what looks wrong with this file?' first — clean the data before you trust any total.
  • Save the clean version to a new file (expenses-clean.csv) so your original is never touched.
  • Read the visual diff before you accept — confirm Claude didn't quietly drop a row that mattered.
  • Keep private financial data in a folder you control — don't paste account numbers into untrusted tools.

Questions people ask

Do I need to install anything?
Just the Claude Desktop app. Put your spreadsheet (`.csv` or `.xlsx`) in a folder, open that folder in Claude Desktop, and ask in the chat. There's nothing else to set up.
Will it overwrite my original file?
Not if you ask it to save the clean version to a new file — `expenses-clean.csv` — so your original is never touched. Every edit shows up as a visual diff you accept or reject, and you can ask Claude to say what it changed so a row that mattered isn't quietly dropped.
Is it safe to clean financial data this way?
Keep private financial data in a folder on a device you control, and don't paste account numbers into untrusted tools. Cleaning a file inside your own Claude Desktop workspace is fine — just verify the totals before you act on them.
How long does it really take?
About ten minutes for a typical messy export: a minute to describe what's in it, a couple to clean it, and the rest asking the questions you opened the file for. Bigger or dirtier files take a little longer to verify.