Clean up a CSV file

Two things happen to your file: values that are the same category written differently (Toyota / toyota / TOYOTA) get merged, and missing values get filled in with a method matched to each column type. Nothing is changed without your approval - the next screen shows exactly what the tool proposes.

How it works

1. You upload a file. It is read, and each column is classified - text category, ordered scale, whole number, measurement, identifier, date. What happens next depends on that type.
2. You see the plan and change it. Which columns to fill, which to use only as context, which to leave alone. Which spelling variants to merge. Whether a suspicious number like 999999 is really a missing-data code. Nothing has been altered at this point.
3. You get the cleaned file, plus a note on what was left alone and why.

What it is careful about

• It refuses columns it cannot learn from. A column that is mostly empty is skipped rather than filled - inventing a value for 90% of rows from the remaining 10% is not cleaning.
• It declines columns it cannot predict. Where a column is too weakly related to the rest of your data for the model to have anything to go on, it inserts one sensible constant instead of a guess. You are told which columns those were, and how weak the relationship was.
• It leaves identifiers, dates and anything resembling a credential untouched, and returns them exactly as they came in.
• It does not silently reshape your file. Duplicate rows stay unless you ask for removal; dates keep their format unless you choose one.

What it will not do

Filled-in values are estimates, not recovered facts. Where a value was missing, the tool produces the most plausible one given the rest of the data - and it can be wrong. Check the result before acting on it, keep your original, and do not use a filled value as the sole basis for a decision about a person. The safeguards above reduce that risk; they do not remove it.

Your file

It is deleted 24 hours after the upload, along with the cleaned result. It is never used to train anything and never shared. The page you are reading loads nothing from third-party servers; there is no third-party analytics and no tracking. We count how many times each page was opened, per day, and that count holds nothing that could identify you. Details: Privacy Policy.