Compare two CSV files

How to compare two CSV files

1. Drop both files. The older export on the left, the newer one on the right. Both are parsed in this tab — the delimiter is sniffed and the encoding is detected before anything is compared.
2. Choose the key column. Pick the column that identifies a row: an ID, an email, an order number. Rows are paired on that value rather than on their line number.
3. Read the diff. Changed cells show the old value struck through next to the new one. Rows present in only one file are colored and counted separately.
4. Export the differences. Download just the changes as CSV or XLSX — one row per difference, with the file each value came from.

Why a text diff is the wrong tool for a CSV

Most tools that compare CSV files are text diffs underneath: they line up file A line 1 with file B line 1 and walk down. That is exactly right for source code and quietly wrong for exported data, because two things happen to exports that do not happen to code:

• The order changes. A report re-run tomorrow can come back sorted differently, or with one row inserted near the top. A line-based diff then reports every following line as changed, and a two-row difference arrives as four thousand.
• The columns move. Somebody adds a column in the middle and every field shifts one to the right. Character by character, nothing lines up; row by row, almost nothing changed.

Matching by a key column removes both problems. The pairing is on the value in that column, so order is irrelevant, and the comparison is per named column, so an inserted column shows up as one new column rather than as a rewritten file.

Delimiters and encodings, before the comparison starts

A CSV is not one format. The separator may be a comma, a semicolon (normal in locales where the comma is the decimal separator), a tab or a pipe; the file may be UTF-8 with or without a byte order mark, UTF-16, Shift-JIS, Big5, GBK, EUC-KR or Windows-1252. Both are detected per file, so two exports that disagree about either still compare correctly — and a file that arrives as mojibake is decoded rather than diffed as garbage. If you only need that half of the job, the encoding fixer does it on its own.

What counts as a difference

Values are compared as text after trimming, which is what makes a diff of exported data usable: 0012 stays 0012 rather than becoming the number 12, and a trailing space introduced by one system does not turn every row red. Quoting is normalized away — a field quoted in one file and bare in the other is the same value, because the quotes are CSV syntax, not data. Line endings, likewise: a file saved on Windows and the same file saved on a Mac are not different files.

Bigger than the browser likes

Files in the tens of megabytes parse and compare in this tab. Past that, the limit you meet is the memory of the tab rather than the tool. Two ways through: split the file first with the CSV splitter and compare the parts, or drop the pair into the TableDI desktop app, which reads the file off disk instead of into a page.

Nothing leaves this tab

CSV exports are usually the whole customer table, the whole order history, or a payroll run. Both files are read by JavaScript running on your machine and the comparison happens there. There is no request to send, so nothing to intercept, log or retain — watch the Network tab while you drop a file and you will see it.