Excel & CSV Data Cleaner

Try it free: click Start without uploading a file to run a free demo on sample data.

Request: "Remove duplicates by email, convert the dates to YYYY-MM-DD, remove extra spaces and capitalize names and cities."

Before	After
" ana LOPEZ ", seattle, 19/04/2025	Ana Lopez, Seattle, 2025-04-19
Ben Carter, "AUSTIN ", 2025-01-28	Ben Carter, Austin, 2025-01-28
chloe diaz, Denver, 13/10/2025	Chloe Diaz, Denver, 2025-10-13
Chloe Diaz, denver, 2025-10-13 (same email)	removed: duplicate
Daniel Kim, CHICAGO, 26/07/2025	Daniel Kim, Chicago, 2025-07-26

Clean Excel data and CSV files: remove duplicates, fix date formats, convert numbers stored as text and standardize data. Upload one Excel (.xlsx) or CSV file and say in plain words what you want fixed. You get back the cleaned file, a short note on what was done, and the detail of every automatic check run on the result.

You pay USD 1.00 per verified file. If the result is not verified, the file is rejected, or the run fails, you are not charged.

What this Excel data cleaner does

The Actor applies a fixed set of operations. Each one is ordinary code, written and tested in advance, and nothing is improvised for your file:

• Remove duplicates: removes rows that repeat on the columns you name, and keeps the first one.
• Fix date formats: converts a date column to YYYY-MM-DD. Dates that can be read two ways (03/04/2025) or that do not exist (31/02/2025) are left as written and flagged in a separate column.
• Numbers stored as text: turns them into real numbers and works out whether the decimal separator is a comma or a dot.
• Values that look wrong: flags numbers in a column that cannot be read, are negative, are zero, or are far from the rest.
• Standardize data: removes extra and invisible spaces. Words are capitalized only if you ask for it.
• Headers: finds the real header row when there is a title above it, and renames headers (for example, from two languages to one).
• A calculated column: the product, sum or difference of two columns.
• One sheet per value: splits the table into sheets by the values of one column (Excel output only).

When a CSV is delivered as .xlsx, columns that only hold plain numbers (such as 2444 or 212.51) become real numbers, so Excel can add them up. Codes with leading zeros, like 007, stay as text.

Requests that work

Please remove duplicate rows based on the email column and standardize the name and city columns (proper case, no extra spaces).

This file has dates in different formats. Convert everything to YYYY-MM-DD and flag anything you cannot resolve.

The amount column is stored as text. Convert it to numbers and tell me if any value looks wrong.

Normalize the headers to English, fix the decimal separator and calculate the total amount column. Deliver as xlsx.

Clean up this shipment report and split it into three sheets by port. Keep all the columns.

Name the columns and say exactly what should change. A generic "clean up" only removes extra and invisible spaces. The note tells you what was left as it came, so you can run the Actor again and ask for it.

Try it without a file

Run the Actor with no file and it cleans a small sample file (sample_customers.csv, made-up data) with the example request. The demo is free: you get the normal output and are not charged. To clean your own data, upload your file.

Files for the CSV cleaner

• One file per run: .xlsx or .csv, up to 20 MB, with one table.
• The format is detected from the file's content. A CSV can be comma, semicolon or tab separated, in UTF-8 or Windows (cp1252) encoding.

How it is verified

After the cleaning, automatic checks written in code run on the output file. They cover, for example:

• the file opens;
• no column or row disappeared without a reason;
• every date in a converted column is YYYY-MM-DD or flagged;
• every converted number matches the original;
• columns you did not ask to change are identical to the input.

You are charged only if every check passes.

An AI review also compares the result with your request. If it finds something you asked for that no operation covered, the file is not charged. It can never make a file pass a check.

What you get

Output	What it is
The cleaned file	<your file name>_cleaned.xlsx (or .csv), in the run's key-value store, linked from the dataset row and the Output tab
NOTE.txt	What was done, how each part of the request was read, and what was left as it came
VERIFICATION.json	Every check run on the result, with its outcome
WARNINGS.txt	Only when the result did not pass verification: what did not pass

Each run adds one row to the dataset with status (verified, not verified, not supported, blocked, invalid input, max charge too low or free plan limit), charged, and links to the file, note, verification and warnings.

If the result could not be verified, the file is still delivered as NOT_VERIFIED_<your file name>.xlsx, WARNINGS.txt says why, and you are not charged.

Pricing

• USD 1.00 per verified file, charged once per run.
• No charge when the result is not verified, the request is outside what the Actor does, the file is not a readable .xlsx or .csv, or the run fails.
• Set the maximum cost per run to USD 1.00 or more. With a lower limit, the run stops before doing any work and nothing is charged.
• Apify Free plan: up to 3 verified files per month. After that, the run stops before processing, nothing is charged, and the note explains that a paid Apify plan is needed to continue.

What it does not do

• PDF, Word, images, old .xls files or more than one file per run.
• Requests outside the operations above (for example, summaries, pivot tables, charts or merging files). You get a note saying so, your file is not changed, and you are not charged.

Privacy

• Your file is processed in a temporary container that is deleted when the run ends. The input and the cleaned file stay only in your own run storage.
• To read your request and review the result, the column names and a sample of up to 10 rows are sent to Anthropic's Claude API. Before sending, emails, phone numbers, IBANs, card numbers, tax IDs and IP addresses are replaced with placeholders. Names are not.
• We do not keep your file or its values. We keep only anonymous metrics:
  • the shape of the table (column types and patterns) and the plan that worked, with column names stored as salted hashes;
  • for requests the Actor could not handle, a short description with the column names hashed;
  • for rejected files, only the file type and the date;
  • run results and costs.