Merge CSV files

Merge CSV files is the tool for stacking datasets vertically. It creates one table containing the records from the chosen sources, in the order you specify. This is useful for supplier catalogs, monthly exports, or registration batches with related structures. It is a different operation from matching identifiers and adding columns from another table. If your question is “Which price belongs to this SKU?”, use Join CSV files. If it is “Put all supplier records into one catalog”, start here.

Choose multiple files or add them one at a time. Each file receives its own parsing confirmation, because a comma-delimited export and a tab-delimited export can belong to the same task. Confirm whether the first record is a header, inspect any initial records being skipped, and verify quoted cells before accepting the import. A field containing a newline remains one cell in one logical record. The tool does not stack physical text lines or remove arbitrary lines just because they resemble a header.

Same columns, in any order requires each file to have the same set of mapped column names. Their positions may differ. The result follows the first selected table's column order, and values from later files are placed under the corresponding names. This strategy is useful when all exports should conform to one schema and you want an unexpected extra or missing field to stop the operation rather than pass unnoticed.

All columns preserves every distinct mapped name encountered across the inputs. Names from the first table appear first, followed by newly encountered names. A record from a file that lacks one of those columns receives an empty field in that position. That empty output does not prove the source contained an empty value: it may mean the source had no such column. Keep the original files or source report when that distinction matters for your review.

Only shared columns keeps names that appear in every input. Before allowing a run that would discard fields, the workbench reports which columns would be lost and requires explicit confirmation. This mode can be useful for a small common export format, but it is a deliberate information reduction. If you need the extra data later, choose All columns or download a separate copy. A result containing no shared columns is rejected rather than presented as a successful empty structure.

Open the per-file header mapping controls when two sources use different labels for the same concept. You can map Product ID in one supplier file to SKU in another. This is a statement you make about meaning; the tool does not infer it from a similar-looking name. A field called Product ID could instead be a supplier's internal identifier. Check documentation or representative records before mapping it to your catalog identifier.

Within each input, mapped destination names must be unique and nonempty. Mapping two source columns to the same target would lose a distinction, so it is blocked. Duplicate original headers also need resolution during import. You can preserve a source provenance column in the merged output. Its generated name avoids colliding with an existing user column, and each value identifies the contributing file and logical record position. File names themselves may be sensitive when the exported table is shared.

Supplier A contains SKU, Product, and Price, with records for 001 and 002. Supplier B contains Price, Product ID, and Product, with records for 001 and 003. Supplier C contains Product, SKU, and Price, with records for 004 and another 003. Confirm each file, map Supplier B's Product ID to SKU, and select Same columns. Although all three layouts differ, the merged table has six data records under one coherent header.

The repeated 001 and 003 records remain. Appending does not deduplicate or decide which supplier is authoritative. Add source provenance if you need to inspect the competing values later. Apply the merged result, move to Remove CSV duplicates, and group by SKU using Mark only or a chosen retention policy. The differences between duplicate rows remain visible there. Once the catalog is ready, compare it with a previous snapshot to examine added, removed, and modified records.

The primary-table selector chooses the first input, including the current workflow data when you are continuing a task. Additional files can be selected and moved upward in the append order. Records within each file retain their order, and files are appended in the visible sequence. That order can affect a later Keep first or Keep last deduplication step. Decide whether the sequence represents a priority before relying on it as a retention policy.

The input file report lists each contributing file and its record count. The merged output count should equal the sum of those counts, because this operation does not delete records. Header records are not included as data when correctly identified during import. A genuine data row whose values happen to spell the header names is still data and remains in the result. Inspect the first and last records around file boundaries when reviewing a merge for the first time.

Run and review creates the full merged result. Compare its column order with your destination requirements, look at empty fields introduced by the selected strategy, and check the total record count. Exporting from this page lets you choose the table itself or supporting reports. View searching does not restrict the export. CSV and TSV are both available with explicit header and UTF-8 BOM choices. Protection for formula-like values is visible and has a count, so it is not a hidden mutation.

For a multi-step task, apply the result before selecting another tool. This keeps the data in the memory session and records the merge as an ordered workflow step. You can undo the step without altering the source files. Refreshing or closing the page clears the session, so export required results before leaving. Saved rules describe configuration but do not contain the source records, and they are not a backup of the merged dataset.