1. Home
  2. Spreadsheets & Data
  3. Excel & CSV Workbench

Excel & CSV Workbench

Open Excel and CSV files, stack cleanup steps, and export the result in one go.

Runs in your browser · No upload

Operations

No operations yet. Add one below; they run from top to bottom.
Files are processed in your browser and never sent to a server.

How to use

  1. Drop an .xlsx, .csv, .tsv or .txt file under Open file, or switch to Paste and paste cells copied from Excel. CP949 (Korean) CSV files are detected automatically.
  2. If the workbook has several sheets, pick one and check whether the first row is a header. The header guess comes from comparing the first two rows.
  3. Choose operations under Add operation. They run from top to bottom, and enabling, disabling or reordering them recalculates the result right away.
  4. Switch between Original and Result in the preview. When only value-changing operations are used, changed cells are highlighted.
  5. Pick a format (xlsx, csv or json) under Export and download. The file is named like orders_cleaned.xlsx.

Operations

Operation What it does
Remove duplicates Keep one row per key, show duplicates only, or add a “Duplicate (2)” marker column. Optional case and surrounding-space ignoring
Blank rows/columns · trim spaces Delete blank rows and columns, trim ends, collapse repeated spaces, fill down the blanks left by merged cells
Sort · filter · pick columns Multi-key sort, 13 filter conditions, keep columns in a chosen order, transpose, seeded random sample
Find & replace Limit to chosen columns. Supports regex with $1 groups and whole-cell matching
Column stats · validation Leaves the table alone and reports count, blanks, unique values, most common values, mean and more per column, then finds rows that break your rules
Phone number format Normalize mobile, area-code and service numbers to one format such as 010-1234-5678. Restores the 0 Excel dropped, strips +82, keeps extensions, flags unreadable numbers
Address cleanup · split Split addresses into region / district / detail (plus ZIP and road-vs-lot type) and unify names such as 서울시 to 서울특별시
Keep leading zeros · text/number Keep ZIP and ID zeros and 12+ digit numbers as text, pad to a length, or turn 1,234 and ₩500 into numbers
Date format Convert 2026.10.7, 26-10-07, 10/7/2026, Oct 7, 2026 and Excel serials to one format. Unreadable cells stay as they are and are counted
Split · merge columns Split one column by delimiter, fixed width or regex groups, or join several columns with a separator
Mask personal data Mask names 홍*동, phones 010-****-5678, ID numbers, emails, addresses, cards and accounts. Column kinds can be auto-detected
Merge files & sheets Stack several files and sheets vertically. Match columns by name (or position), add a source column, dedupe afterwards, or keep one sheet per source
Split into files Split by N rows, by column value or into N equal parts, saved as ZIP (xlsx/csv) or several sheets. The header repeats in every file, up to 500 results
Compare two tables Cell-by-cell comparison on key columns (same, changed, added, removed, with highlights and a change-list CSV), or what two lists share and what only one has
VLOOKUP match Look up rows in a second table by key and bring columns over. Left or inner join, duplicate-key handling, a report of keys not found
Pivot table 1–3 row levels, a column group, values (sum, count, distinct count, average, min, max, median), totals and date grouping (year, month, day, weekday)
Convert format Sets the export format plus options such as delimiter and encoding

Stacking several steps in one pass

Once a table is open you can stack operations and save once, with no intermediate files. Order matters, and the workbench is built around that.

  • Remove blank rows, trim spaces, remove duplicates, save as xlsx. With rows 홍길동 (trailing space), an empty line, 홍길동, 김영희 and a whitespace-only line, the pipeline removes 2 blank rows, cleans spaces in 2 cells, then removes 1 duplicate row, leaving 홍길동 and 김영희.
  • Sort, filter, save as CSV (CP949). To keep each person’s highest score from A and B with scores 1, 5, 9, 3, sort by score descending, then remove duplicates by name keeping the first row. You get A with 9 and B with 5. Reverse the order and you get each person’s first row instead (A with 1, B with 5).
  • Stats and validation only. Turn on just the column stats operation and the table stays untouched while the per-column report appears.

The Step report lists results such as “Removed 1 duplicate row” next to the row counts, so you can see which step shrank the table.

Working with several tables

  • Opening several files at once treats each file and sheet as a source and puts a “Merge files & sheets” step first. Tables with columns name, age and age, city, name merge into name, age, city, and missing cells stay blank.
  • Steps that need a second table (compare, VLOOKUP) take a file or pasted cells inside the step card, with their own sheet, encoding, delimiter and header choices. The table you picked stays when the card is collapsed.
  • In Compare, 10 and 10.0 are the same value, but a code with leading zeros such as 007 is not treated as a number, so it differs from 7.
  • VLOOKUP reports how many rows were not found and the keys not found (first 50) in the step report. When the lookup table repeats a key, choose between adding rows (all) or using one match (first/last).
  • Pivot leaves non-numeric values out of sum, average, min, max and median and tells you how many. In total rows and columns the average is over all values, not an average of averages.
  • When Split or Merge by sheet is the last step, export becomes a ZIP or a multi-sheet xlsx; adding another step after it turns the several results back into one table.

Supported formats and reading rules

Format Supported Notes
.xlsx Yes Sheet picker; formulas are read as calculated values
.csv · .tsv · .txt Yes UTF-8, CP949 and other encodings auto-detected; delimiter auto or manual
.xls (legacy) · .ods No In Excel or LibreOffice use Save As → .xlsx first
Password-protected files No Remove the password, then save as .xlsx
  • CSV values are never converted to numbers. 00123, 010-0000 and 1e3 are read as text, so leading zeros survive. Sorting and stats treat a column as numeric only when every non-blank value looks like a number.
  • Formulas become values. Formula cells in xlsx are read as their last calculated value, so the saved file holds values, not formulas.
  • Merged cells are read with the value in the first cell and blanks after it. Use “fill down” in the blank rows/columns operation to turn Sales, , , Dev, into Sales, Sales, Sales, Dev, Dev.
  • Tables beyond 200,000 rows, 200 columns or 5 million cells are loaded only in part, and you are told. An xlsx over 50 MB uses a lot of browser memory.

Saving as CSV

  • Formula-injection guard (on by default). Excel can run text beginning with =, +, - or @ as a formula when it opens a CSV. The workbench puts a ' in front of such text cells: =SUM(A1) becomes '=SUM(A1) and @cmd becomes '@cmd. A phone number like +82-10-1234-5678 gets the same prefix for the same reason, while a real number such as -12.5 is left alone. Turn the option off if the ' gets in the way of another program.
  • Encoding. If Korean text looks garbled in Excel, choose UTF-8 with BOM (the default). CP949 is available for old programs; characters it cannot represent become ? and you are told how many. For encoding conversion alone there is also the Text Encoding Converter.
  • Line breaks and quotes. CRLF is the default for Excel compatibility, and you can quote every field. Cells containing commas or line breaks are quoted automatically.
  • Saving as xlsx makes the header bold, sizes columns to their content (Korean characters count double, 8 to 60 characters), and stores text cells as text so leading zeros stay. The option that saves number-like text as number cells skips values with leading zeros.

Sorting and stats rules

  • Sorting is stable and blank values always go last, ascending or descending. A column that is entirely numeric is compared as numbers, giving 2, 9, 10, 100 (10 after 9). Text uses Korean collation, so Hangul sorts before Latin letters.
  • Filter conditions greater than, at least, less than and at most compare numbers when both sides are numbers, dates when both are dates, and text otherwise; blank cells never match. Equals, contains, starts with, ends with and regex ignore case.
  • Standard deviation in column stats is the sample one (n-1). For 1, 2, 3, 4 the sum is 10, the mean and median are 2.5, and the standard deviation is about 1.2910. With a single value no standard deviation is shown.
  • Validation rules can be set per column: required, type (number, date, TRUE/FALSE), pattern (email, phone number, digits only) and unique. A unique violation lists every row carrying the repeated value. The list holds up to 1,000 rows and the total count is reported separately.
  • Operations with regex, and tables above 500,000 cells, run in a separate worker so the page stays responsive; a regex that runs longer than 10 seconds is stopped.

Files are processed in your browser and never sent to a server.

Can't find the tool you need?

Tell us what you want to do. We build the most requested tools first.