Merge CSV Files

Choose how the CSV merger aligns columns

Merging appends records from each selected file. It does not join tables by account ID, update an existing record, or remove duplicates. The alignment setting decides which output column receives each value. Check every source with the CSV validator first, especially when exports come from different systems: merging does not resolve ambiguous headers or malformed rows.

Try the same two files in all three modes

Download merge-east.csv andmerge-west.csv. Select east first, then west, use comma as the delimiter, choose “Use as header,” and leave the source-filename column off to reproduce these outputs. The filename comments below identify the two files and are not part of their CSV content.

# merge-east.csv
account_id,region,total
000184,east,52.20
000219,east,71.00

# merge-west.csv
region,account_id,status
west,000305,active
west,000402,pending

Union: retain every distinct column name

Union is useful when monthly or regional exports share core columns but carry different optional fields. Output headers follow their first appearance across the selected files. Values are matched by header name, so west's reversed first two columns are realigned. A field missing from a source becomes an empty string; the merger does not invent a total or status.

account_id,region,total,status
000184,east,52.20,
000219,east,71.00,
000305,west,,active
000402,west,,pending

Intersection: keep only the common schema

Intersection retains names present in every file, in the first file's order. For these fixtures,total and status disappear, while all four records remain. Use it for a downstream import that deliberately accepts only shared fields. Check what will be discarded before exporting; “common” means common column names, not common customers or duplicate rows.

account_id,region
000184,east
000219,east
000305,west
000402,west

Position: use field order regardless of names

Position uses the first file's headers, extending them with generated labels if another source has more columns. Shorter records receive empty trailing fields. It is appropriate only when each column position has the same meaning in every source, including headerless exports whose layout you know. It cannot recognize a reordered schema. With these two fixtures, the resulting west records are incorrectly labelled for their meaning:

account_id,region,total
000184,east,52.20
000219,east,71.00
west,000305,active
west,000402,pending

Here west sits under account_id, and active under total. Use union or intersection for this pair. A rectangular preview is not proof that the fields match the business meaning of their headers.

Case, spaces, order, and repeated names

Header parsing trims leading and trailing whitespace, so account_id matchesaccount_id. Matching across files is case-sensitive: Account_ID is a separate union column and is not shared with account_id in intersection mode. Internal spaces also remain significant. Reordering is handled by the name-based modes, not position mode.

Standardize spelling and capitalization before merging. Give every field within a source a unique, non-empty name: repeated names are ambiguous, and name-based alignment can retain only the last occurrence's value. When enabled, the source column prepends the filename to each output record; avoid a conflicting source_file input header.

Merge CSV records, not physical lines

A quoted field can contain commas, escaped quotes, and line breaks. Splitting files on newline characters or removing the first physical line from every file can break those records. This merger parses each input before aligning fields and serializes CSV again on export. Compare themultiline fixture with thereordered-header fixture to test both boundaries.

Records retain file-selection order and their original order within each file. The current complete merge limit is 250,000 data records across the output. For the full workflow, follow thedifferent-columns merge guide, then useCSV to Excel if your destination needs a workbook.