Turn rows into columns
Need rows to read as columns? Transposition swaps the table axes and helps reorient small matrices and summary data.

Swap rows and columns to see a table differently. Swap rows and columns with explicit header handling.
Import data, confirm the rules, review the result, and export a new copy.
Drag and drop a CSV, TSV, or delimited text file
UTF-8 · Up to 10 MB per fileExplore the available features and choose what your table needs next.
Exchange rows and columns to produce a new matrix.
Use in the toolChoose whether source headers participate in the transformation.
Use in the toolCheck the resulting row and column counts before exporting.
Use in the toolKeep leading zeros and original characters without automatic type conversion.
Use in the toolReceive explicit feedback when output dimensions exceed supported limits.
Use in the toolSave the reshaped table as CSV or TSV while preserving the source file.
Use in the toolConfirm the input, inspect changes, and save the output you need.
Choose a CSV or TSV file, or paste a table. Check the delimiter, headers, and preview before confirming the full import.
Choose whether headers participate in the transpose and review expected dimensions. Check the resulting shape and first row before exporting a new copy.
Choose the result or report and CSV or TSV format. Review the export summary and protection count, then confirm download or copy.
Transposing a CSV exchanges row and column positions. A value at one row and one column moves to the corresponding column and row in the output. This is useful for small matrices, compact configuration tables, or exported measurements whose orientation needs to change. It is not a pivot operation: the tool does not group categories, sum repeated values, or choose an aggregation. Every participating cell keeps its text value while its matrix position changes.
Start with a CSV, TSV, or pasted table and confirm parsing. Pay particular attention to whether the input has a header. A header is metadata in the normal workbench table model, while transposition operates on a matrix of values. The two header controls make the transition between those representations explicit. Their settings determine whether original header names participate as data and whether the first transposed row is promoted to new column names.
Include original header in the matrix adds the source column names as the first row before transposition. When transposed, those names become the first column of the new matrix. This is appropriate when the labels should remain visible alongside the values after orientation changes. For example, product attributes across the source header can become a column of attribute names in a small comparison table.
Disable that option when you want to transpose only the data values. The source header then remains a label for understanding the input but is not part of the transformed matrix. This distinction is especially important for headerless input, where the workbench supplies generated column labels for navigation. Including those generated labels would add text that was not present as data in the original matrix. Inspect the predicted dimensions before running so that an extra row or column is intentional.
Use the first transposed row as headers promotes that row to destination column names and removes it from the output data rows. The resulting names must be unique and nonempty. If the prospective header contains duplicates or blank cells, the operation stops instead of silently renaming or overwriting fields. Leave the option off when those values should remain ordinary data rather than labels.
Without promotion, the workbench generates distinct column names for previewing the transposed matrix. Those generated names are not additional data values. During export, decide whether you want a header included in the file. If you are preserving a pure headerless matrix, turn header export off. Keeping these choices separate lets you produce either a labeled table or a raw matrix deliberately, rather than treating the first visible row as a header automatically.
Imagine a headerless two-by-three matrix. The first row contains 001, x, and an empty field; the second contains 002, y, and z. Disable original-header participation and new-header promotion. The result is a three-by-two matrix with rows containing 001 and 002, then x and y, then an empty field and z. Neither identifier is converted to a number, and the trailing empty field remains a cell in the matrix.
Apply that result and transpose again with the same two header decisions. The data matrix returns to its original two-by-three shape and values. This reversibility is a useful way to validate a transposition workflow. It assumes that you have not promoted a row into metadata, inserted generated headers as data, or changed the values in an intervening step. If you export between the two operations, use consistent header settings when reading the file again.
A tall table becomes a wide table. Ten thousand records would become roughly ten thousand columns, which is not suitable for this workbench's limit of two hundred columns. The interface shows the projected row and column counts, including the effect of each header setting. The operation blocks an oversized output instead of truncating it. Reducing the input to a meaningful smaller subset is preferable to accepting a partial matrix without knowing which values were lost.
The total cell limit also applies, and long cell text can be expensive even in a modest matrix. The input parser requires a coherent width, with explicit padding available for short records. Missing padded cells become empty fields when exported to ordinary CSV. A destination format cannot necessarily distinguish that missing state from an intentionally empty string, so retain the source if the distinction matters to your analysis.
Transposition fits a direct orientation change where every cell has a clear new coordinate. A compact product comparison with products down rows and attributes across columns can be turned around for another system's expected layout. A small sensor matrix can similarly be reoriented without interpreting measurements. The operation preserves strings such as dates, identifiers, and formula-looking text rather than evaluating them.
Use a pivot or aggregation workflow when you need one output column for each unique category and repeated combinations must be summarized. Use an unpivot workflow when several measurement columns need to become name-and-value rows while identifier columns repeat. Those are different transformations with different record counts and rules. They belong to later capabilities rather than being implied by the word transpose. This page does not claim to perform either one.
Review both the original and result dimensions, inspect corner cells and empty values, and open long cell details when necessary. Apply the result if the next tool should operate on the transposed structure. A later rule may need new field selections because transposition creates a different schema, unlike a simple rename that preserves the same column identity. The workbench should block an invalid old reference rather than applying it to an unrelated new field.
Export the result as CSV or TSV with explicit header and UTF-8 BOM choices. The verification stage rereads the generated text to check its rows and cells before downloading. Spreadsheet protection can prefix formula-like values; review that count if exact text matters to a matrix consumer. A raw export keeps strings but carries the receiving-program interpretation risk. The source file remains untouched, and the memory session can be undone or cleared separately from downloaded copies.
Import the table, decide whether the existing header participates as data, and preview the transposed structure. Check the new field names for blanks or duplicates. This swaps positions without aggregating values. A very tall input becomes very wide, so the output can reach the column limit even when the source file is small.
No. Transpose swaps coordinates in a matrix. A pivot groups records by dimensions and usually needs an aggregation for repeated combinations. This tool does not add, count, or average values. If your task involves grouping repeated keys into summary columns, a transpose can produce the wrong shape even when every cell is preserved.
Check whether original-header participation was enabled. Including the source header adds one row to the matrix, which becomes one additional column after transposition. Generated labels for a headerless source can also be included if you choose that setting. Inspect the dimensions and decide whether those labels should be values in your output.
Promoting a row to headers changes those values into field identifiers. Duplicate or empty identifiers make later column selection ambiguous and can cause receiving software to overwrite values. The workbench therefore requires a clear schema. Leave promotion off when the first transposed row is ordinary data, or prepare unique labels deliberately before trying again.
Only within the output-column and cell limits. A long list can be small in bytes yet create far too many columns. The workbench checks the projected shape and blocks outputs beyond two hundred columns. This is a deliberate limit rather than a promise that every device can render an arbitrarily wide table.
Start with a sample to learn the workflow, then process your own files.