Every tool, and what it does not do

Diagnose

Workbook health check

Drop a workbook: within seconds, see what weighs it down, what makes it fragile and what it leaks. Every problem comes with its fix, and the fixed file is read back before it is offered.

• The file's weight, part by part: sheets, styles, images, pivot table caches.
• The real end of the data on each sheet, and the phantom rows beyond it.
• Duplicate styles, copied conditional formatting, broken defined names, links to other workbooks.
• Formula errors, hidden sheets, rows and columns, the author and the comments.
• The fixed file is read back and compared with the original, cell by cell, before it is offered.

What it does not do: Excel 2007 and later workbooks (.xlsx, .xlsm). An old .xls file is not read; a password-protected workbook is, once its password is typed, and a workbook whose sheets alone are protected is read directly.

Clean before sending

A workbook carries its author's name, the path of the folder it was saved in, its comments and its hidden sheets. See what it holds, choose what goes, and send a clean file.

• The author, last editor, company, save-folder path and custom properties.
• Comments and notes, with their authors' names.
• Hidden sheets, deleted when nothing else depends on them; formulas that read them become values.
• Hidden rows and columns, unhidden so you can see what you are sending.
• Macros, removed: the workbook becomes an ordinary .xlsx file.

What it does not do: A hidden sheet that feeds a drop-down list, a chart or a pivot table is kept, and the screen says why. Hidden rows and columns are unhidden, not deleted: deleting them would shift the workbook's references.

Inspect a workbook

A hidden sheet with the margins, the author's name in the properties, a forgotten comment, a link to a colleague's drive: it all goes out with the file. Here is the list before you send it, without changing anything.

• Hidden and very hidden sheets (invisible even from Excel's menu), hidden rows and columns.
• Document properties: author, last modified by, company, dates.
• Comments and notes, defined names, data connections and macros.
• The workbooks targeted by external links, with their full path.

What it does not do: The inspection reads and changes nothing. To remove what it finds, the clean-before-sending tool does it in one click.

Formula audit

In the middle of a column of totals, a cell where someone typed 999 instead of the formula. Elsewhere, a formula that multiplies by 1.2 when its neighbours do not. Every formula in the workbook is read and what stands out is listed, cell by cell.

• Formulas that return an error (#N/A, #REF!, #DIV/0!, #VALUE!), with their formula.
• Hard-coded values in the middle of a column of formulas, and formulas that differ from both neighbours.
• Direct circular references, formulas that read another workbook, volatile functions (INDIRECT, OFFSET, NOW) that slow the file down.
• A workbook report: each finding with its sheet, cell and formula.

What it does not do: Formulas are read, not recalculated: the audit points to what deserves a look, it does not say a result is wrong. A circular reference that goes through several cells is not followed.

Dependency map

You inherit a twenty-sheet workbook, want to delete one sheet, move a file on the network, understand where a total comes from. First you need to know what reads what: the map shows it, sheet by sheet, without opening a single formula.

• A diagram of the sheets in tiers: inputs on the left, calculations in the middle, summaries on the right, with the number of formulas on each link.
• The outside workbooks called, the file targeted and an example formula for each.
• The defined names: what they refer to, and the sheets that use them; those no formula uses are flagged.
• Hidden sheets appear too, dashed: they are often where the calculations hide.
• The map can be saved as a workbook, to document the file.

What it does not do: Only cell formulas are read. A reference built by INDIRECT, written in a macro, a data validation or a chart does not appear on the map.

Understand a Power Query query

A Power Query query fetches data from elsewhere and transforms it through a chain of steps, in a language (M) that few people read. You drop the workbook, or paste the M code, and each step is said in one sentence, in order.

• The source, in plain words: the file, folder, website or database the query pulls its data from.
• Numbered steps: keep some rows, remove columns, group, merge with another query, each in one sentence.
• The pitfalls to know: a source that refreshes on its own, credentials kept per machine, data as old as the last refresh.
• Two inputs: drop the workbook, and the tool lists its queries; or paste a query's M code (Excel's Advanced Editor).
• Nothing leaves: the workbook is read in the tab, as everywhere on Klargrid.

What it does not do: The tool explains a query, it does not run it or change it. A rare transformation, written by hand in the M code, gets a generic sentence with the function's name.

See what a macro does

A macro-enabled workbook can launch a program, download a file or write to disk, right on opening. Before enabling macros, you read what they do: the tool opens the VBA project, lists the modules and points out the calls to watch.

• What runs on its own: the Workbook_Open, Auto_Open… procedures that run on opening or closing.
• Network calls: downloads, HTTP requests, connections to the internet.
• Running programs: Shell, WScript.Shell, system commands.
• The file system: reading, writing and deleting files.
• Each module's code unfolds, to read it in full.

What it does not do: The tool reads the macros' source code, it does not run it or remove it; to remove macros, use Clean up a workbook. Each procedure is described on its own: if one calls another, the chain is not followed.

Explain a pivot table

You inherit a workbook with a pivot table, and you want to know where it gets its figures and what it computes, without clicking all over Excel. The tool reads its definition and turns it into sentences.

• The source: the sheet and range, or the table, the data comes from.
• The fields in rows and columns, and the filters applied.
• The values and their calculation: sum, average, count, and percentages of the total.
• Every pivot table in the workbook, on its sheet and cell.
• Nothing is sent or recomputed: it reads the definition, as Excel saved it.

What it does not do: The tool reads what a pivot table is set to do, it does not run it or show its computed results. A calculated field written by hand is named, without being detailed.

Audit a folder's connections

Before sharing or archiving a batch of workbooks, you want to know which ones reach out for data, and where. Drop them all: the tool lists every data connection, file by file, without changing anything.

• Power Query: the real sources, read from the M code, with their address.
• Legacy data connections (web, database, ODBC) left in the workbook.
• Links to other workbooks.
• A column tells whether the data leaves the machine (a remote source) or stays a local file.
• Nothing is changed: this is a survey. The report downloads as .xlsx; nothing is sent.

What it does not do: The tool lists connections, it does not remove them; to clean a workbook before sharing it, see Remove connections before sharing.

Repair and slim down

Slim down a workbook

A workbook grows without any new data: formatting applied to whole columns, thousands of styles copied from file to file, conditional formatting rules duplicated on every paste. We remove that weight, and nothing else.

• Phantom rows and columns beyond the real end of the data.
• Duplicate cell formats and unused named styles, the cause of the "Too many different cell formats" message.
• Duplicated conditional formatting rules, merged back into one.
• Pivot table data copies, images nothing displays any more, broken defined names.
• The parts we do not touch come out byte for byte: charts, images and macros stay as they are.

What it does not do: Images are not recompressed: an 8 MB photo pasted into a sheet keeps its weight. The health check shows it, so you can replace it.

Remove external links

Every opening asks whether to update links to a file that no longer exists, or sits on a colleague's drive. We break the links the way Excel would: every formula pointing to another workbook is replaced by its last value.

• The list of linked workbooks, with their path and the number of cells that depend on them.
• Formulas pointing to another workbook replaced by their last value; other formulas stay.
• Defined names pointing to another workbook, removed.
• The fixed file is read back: every value is the same as before.

What it does not do: If a drop-down list, a conditional format or a chart points to the other workbook, nothing is touched and the screen says where: breaking the link would leave them pointing nowhere, and Excel would "repair" the file on opening.

Formulas to values

To send a table without its calculations or its links to other files, or so that it stops moving: each formula becomes the value it displayed. It is Excel's Paste Special, Values, on all the chosen sheets at once.

• The number of formulas on each sheet, and the choice of sheets to freeze.
• Each formula replaced by its last calculated value; numbers, text, dates and booleans keep their type.
• Formatting, column widths and the other sheets stay as they are.
• The output file is read back: every value is the same as before.

What it does not do: The value kept is the one Excel saved at the last calculation: if the workbook was saved without recalculating, that is the value. A formula with no saved value becomes an empty cell, and it is counted.

Find and replace

Excel's find and replace knows no regular expressions, does not sweep every sheet at once and shows nothing before acting. Here you search a text or a pattern, see every affected cell, and replace everywhere, in the text as in the formulas.

• A regular expression (\d+, groups, alternatives) or a text taken literally.
• Every sheet at once; match case and whole cell, as you choose.
• In cell text and/or in formulas (rename a range, change a function name).
• A preview of every match, before and after, before saving.
• The workbook is preserved: only the affected cells change, the rest comes out untouched.

What it does not do: The replacement covers text and formulas, not the number format nor the numeric values themselves. A changed formula has its value recalculated by Excel on open.

Unmerge cells

"To do this, all the merged cells need to be the same size": merges prevent sorting, filtering and pivot tables. Unmerging in Excel leaves empty cells; here the merged value is copied into each one.

• The number of merged ranges per sheet, and the choice of sheets to process.
• Each merge removed, and its value copied into every cell it covered.
• The original cell's formatting carried over, so dates stay dates.
• The output file is read back: no existing value has changed.

What it does not do: The value is copied, never the formula. A merge of more than 10,000 cells is unmerged without being filled, so the file does not balloon.

Recover a damaged file

An interrupted download, a USB stick pulled out too early, a cut-off attachment: Excel refuses to open the file, or "repairs" it by removing half the content. The file is read part by part, even without its table of contents, and whatever can be saved is saved.

• Part-by-part reading, including a truncated file whose end is missing.
• Unreadable parts set aside, broken references removed, a minimal style sheet if the real one is lost.
• A repaired workbook, read back and checked before it is handed to you.
• Always, in addition, the recovered values in a new workbook, sheet by sheet.

What it does not do: What is no longer in the file cannot be rebuilt: a sheet cut off midway gives back its first rows. The old .xls format is not supported; a password-protected workbook is, once its password is typed.

Batch-process workbooks

Every branch's workbook to turn into values before sending, the month's reports to rid of their links and comments: tick the fixes once, drop all the workbooks, and get an archive back.

• The fixes of the single-file tools, to combine: formulas to values, external links broken, cells unmerged, workbook slimmed down, properties, comments, printer settings and macros removed.
• Each workbook is read back and compared with the original after the fix, as in the single-file tool; one that fails the check is not delivered.
• A summary in the archive: for each workbook, processed, nothing to fix, protected, unreadable or set aside, with its size before and after.
• The ticked fixes are saved as a routine, to open again next month.

What it does not do: The settings apply to every sheet of every workbook: to convert only some sheets, use the Formulas to values tool, one workbook at a time. Hidden sheets are neither deleted nor shown.

Remove sheet protection

Last year's budget model, protected by someone who has left; a locked sheet whose password nobody remembers. Sheet protection is not encryption: it is a tag the workbook carries, and it can be removed without looking for the password.

• Sheet protection (locked cells) and the allowed ranges that come with it.
• Structure protection: sheets can be added, renamed, hidden and unhidden again.
• The read-only recommendation and its password to modify.
• The workbook comes out with its data, formulas, formatting and macros, read back and compared with the original.

What it does not do: For your own workbooks, or those entrusted to you with the agreement of whoever protected them. A workbook with an opening password is encrypted: this tool does not open it without its password, and no tool should.

Remove connections before sharing

A workbook you send can, on the recipient's machine, try to reconnect to a website, a database or a network file, and show link errors. This tool removes those connections; the figures already computed stay, frozen.

• Power Query: the queries that fetch data from elsewhere.
• Data connections (web, database, ODBC) and query tables.
• Links to other workbooks.
• The values already in the cells stay: they just no longer refresh.
• The output is an .xlsx, ready to share; nothing is sent.

What it does not do: The tool removes connections, it does not recompute and removes neither macros nor formulas; for macros, see See what a macro does, and to break links in formulas, Remove external links.

Consolidate workbooks

Each subsidiary filled in the same budget template; the group total is needed. Rather than thirty copy-pastes or an endless formula, drop the workbooks: each input cell becomes the sum of all of them, in the template itself.

• The template keeps its formatting, its formulas (subtotals, ratios) and its sheets; Excel recalculates the formulas on opening.
• Inputs in every sheet of the same name are added up, address by address; a date is never added.
• What does not match the template is reported: a missing or extra sheet, text or a formula instead of a number.
• Each sum is read back in the output workbook before it is offered to you.
• Password-protected workbooks open with their password, and everything stays in your browser.

What it does not do: The workbooks must follow the same template: same sheets, same cells. A row inserted in only one workbook shifts everything below it, and the addition then happens in the wrong place: the list of differences helps spot it, not fix it.

Clean data

Fix a CSV file

Drop the CSV: the tool recognises its encoding and delimiter, repairs damaged characters, and gives you a file Excel opens right the first time, as CSV or XLSX.

• The encoding recognised: UTF-8, Windows-1252, Mac, DOS code page, UTF-16.
• Double-encoded characters ("é", "’") repaired one by one, with examples on screen.
• The delimiter detected from how regular the columns are, not from how many commas there are.
• Quoted fields, line breaks inside a cell, quotes that were never closed.
• In XLSX, numbers and dates recognised column by column, and codes with leading zeros kept as text.

What it does not do: A file that is not text (an old .xls renamed .csv, for example) is refused with an explanation. A number a spreadsheet already truncated at its 16th digit cannot be repaired: the lost digits are no longer in the file.

Remove duplicates

Choose the columns that make one row the same as another. The tool shows the duplicates before removing them, and gives you the file without them, with the list of what was removed.

• The columns that matter, your choice: all of them, just the email address, or the name and the city.
• Case, extra spaces and accents ignored on request: "SMITH LTD" and "Smith Ltd" become the same value.
• The first or the last occurrence kept.
• A "Removed duplicates" sheet in the output file, so you can check.

What it does not do: These are exact duplicates once case, spaces and accents are set aside. "Smith Ltd" and "Smith Limited" are not recognised as the same company: fuzzy matching is a different job, with a row-by-row review.

Near-duplicates

Before a CRM import, the same company appears under three spellings, a customer with a typo. Exact deduplication cannot see them: here, rows that look alike are grouped, and you decide group by group.

• Spellings reduced to the essentials: case, accents, punctuation, dotted abbreviations, legal forms (Ltd, Inc, LLC, GmbH, SA…), "&" and "and".
• Typos and words in a different order brought together, with an adjustable sensitivity.
• The least certain groups first; rows identical once cleaned are accepted in advance.
• Your choice: remove the duplicates and keep the first row, or number the groups in a column and remove nothing.

What it does not do: Similarity is not proof: "Martin Ltd" in Leeds and "Martin Inc" in Boston may be two companies. Add the city or the postcode to the compared columns, and review the groups.

Clean text

"Smith" and "Smith " are not the same value to a spreadsheet, and a VLOOKUP fails without a word. Choose the columns: extra spaces, non-breaking spaces, line breaks and invisible characters go, and the case is made consistent if you want.

• Leading, trailing and double spaces, and non-breaking spaces from a web page or a PDF.
• Line breaks inside cells, replaced by a space.
• Invisible characters: zero-width spaces, soft hyphens, control codes.
• Case: UPPERCASE, lowercase, Title Case, or Sentence case.
• The preview highlights every changed cell before you save.

What it does not do: Only the ticked columns are touched, and only their text cells: numbers and dates stay as they are. Typos are not corrected.

Convert text to numbers

A number stored as text does not add up, does not sort and breaks lookups. The tool reads numbers with a decimal point or a decimal comma, with their currency symbols, thousands separators and accounting negatives, and says which ones it could not read.

• The decimal point or comma, detected column by column, or set by you.
• Thousands separators, €, $ and £ symbols, percentages.
• Negatives in brackets (accounting) and with a trailing minus (SAP exports).
• Codes with leading zeros and numbers longer than fifteen digits stay text.
• Cells that could not be read are counted and shown, and left as they are.

What it does not do: A number a spreadsheet already rounded (beyond fifteen digits) does not get its lost digits back. An ambiguous value such as "1.234" is read according to the column's convention: check the preview.

Fix dates

A date stored as text does not sort and cannot be calculated with. The tool reads European, US and ISO dates, decides the day and month order from the evidence in the values (a day above 12), and refuses dates that do not exist.

• DD/MM/YYYY, MM/DD/YYYY, YYYY-MM-DD, with or without a time, and the compact YYYYMMDD form.
• The day and month order decided per column, never cell by cell.
• Serial numbers (a date turned into a number, such as 46023) on request.
• Impossible dates (February 31st, February 29th of a common year) flagged, not invented.

What it does not do: A column where nothing goes above 12 (03/04/2026) keeps the order of your language, unless you set it. Dates written in words ("3 April 2026") are not read yet.

Leading zeros and long numbers

A spreadsheet reads "01000" as the number 1000, and "3012345678912" as 3.01E+12. The tool pads codes back to the column's original length and writes long numbers out in full, as text, so they are not lost again.

• The original length detected per column (the longest one that comes up often), or set by you.
• Scientific notation undone when the file still holds every digit.
• Numbers that have already lost digits are flagged, never padded at random.
• The result is written as text: Excel can no longer strip the zeros.

What it does not do: A number shown as "3.01234567891E+12" only has twelve significant digits: the thirteenth is lost and no tool can recover it. Take it again from its source.

Clean a software export

"Printed" exports from ERPs, banks and payroll software repeat the header on every page, slip subtotals between the data and write a group label only once. We turn them into a table you can filter and sort.

• Headers repeated through the file, removed.
• Subtotal and total rows ("Subtotal", "Grand total", "Carried forward"), removed and shown.
• Fully blank rows, removed.
• A header spread over two or three rows, flattened into one.
• Group labels filled down in the columns you tick.

What it does not do: A subtotal is recognised by its label: a total row without a label stays, and shows in the preview. Merged cells in a workbook are handled by the unmerge tool.

Split or merge columns

A full name into first and last name, an address into street, postcode and city, a column cut on a delimiter: the new columns take the old one's place. Or the reverse, several columns joined into one.

• First and last name, in the usual orders: "Marie Curie", "CURIE Marie", "Curie, Marie".
• Street, postcode and city, from a one-line address.
• On a delimiter (space, comma, semicolon, hyphen, bar), into as many columns as needed, or at the first one only.
• Merge several columns with the delimiter of your choice.
• The original column kept or replaced, your choice.

What it does not do: Compound names without a hyphen ("Mary Ann Smith") are split at the first space, unless the surname is in capitals. The preview shows every split before you save.

Standardise formats

Before importing into a CRM or payroll software, each column has to speak one language: phone numbers in one format, postcodes complete, emails in lowercase. Values that cannot be read are shown, not guessed.

• French phone numbers as "06 12 34 56 78" or international "+33612345678"; other international numbers kept.
• French postcodes on five digits, Corsica included (2A, 2B).
• Emails in lowercase, without spaces or "mailto:", invalid addresses flagged.
• SIRET and SIREN numbers grouped by three digits, proper names in title case.

What it does not do: Phone number formatting follows French conventions; foreign numbers are kept as they are when already in international format, and flagged otherwise. An email's validity is only checked on its form.

Chain cleaning steps

Every month the same export arrives with the same defects, and the same ten minutes of manual fixes. Chain the steps here once, watching the result build up: the routine is saved, and next month it applies in one click, to one file or to every file you drop.

• Eighteen steps to combine: clean the export, clean text, convert numbers and dates, restore leading zeros, standardise formats, split or merge columns, remove duplicates, filter and sort rows, validate IBANs, company numbers and emails, look up a reference file, unpivot or pivot, build a pivot table, delete or rename columns.
• An assistant that composes the routine from a sentence ("keep rows where Status is paid, sort by Client"): it shows the steps, you check them, nothing is guessed in silence.
• Ready-made routines for the files found everywhere (software export, bank statement, contacts for a CRM, file to import, codes and identifiers), set up from your file's columns.
• The lookup brings in, like a VLOOKUP, the columns of a reference file (the customer list, the chart of accounts); the routine remembers its name, and you just drop it along with the month's files.
• Each step applies to the result of the previous one; the preview and each step's summary update with every setting.
• The routine is kept in your browser and opens again in one click; it can also be downloaded as a small file, to share with a colleague.
• From any cleaning tool, "Continue in a routine" takes the file and the current setting along as the first step.
• Several files: the routine applies to each, and you get an archive with a summary saying which ones to check, or a single file combining all the results, or one file per branch, per sales rep, per value of a column.
• A whole folder, in Chrome and Edge: every file in the folder is processed, and the results are written to a subfolder, next to the originals; next month, the same folder is offered in one click.
• Columns are named, not numbered: a file whose columns have moved is still processed.

What it does not do: A column renamed in the source file is not guessed: the step looking for it is flagged and the file marked "to check" in the summary. The routine holds settings only, never data.

Anonymise a file

A payroll export for a provider, a customer base for a demo, a file to attach to a support ticket: the data is needed, not the people. Personal columns are recognised, and each value is replaced by a made-up one, always the same for the same person.

• Names, emails, phones, IBANs, identifiers (social security, company number, staff number), addresses: recognised from the name and content of the columns.
• Consistent replacements: the same person becomes "Person 12" everywhere, in every column; the file can be counted, looked up and deduplicated like the original.
• Obviously made-up values, but well formed: an import check accepts the IBANs and emails produced.
• The key table, only if you ask for it, in a separate file.
• The file never leaves your computer: that is the whole point, since it holds what needs protecting.

What it does not do: With the key table, this is pseudonymisation under the GDPR, not anonymisation. The columns left as they are (dates of birth, postcodes, rare amounts) can still make someone recognisable: read them over.

Check

Validate data

A mistyped IBAN, a company number one digit short, a social security number whose key does not add up: see it before the bank or the payroll software rejects the import. Each column gets its check, suggested from its name.

• IBAN (mod 97 and length per country), BIC, SIREN and SIRET (Luhn, with the La Poste exception), EU VAT numbers.
• French social security numbers, Corsica included, EAN codes, emails, phone numbers, postcodes, impossible dates, required fields.
• The failing cells listed, with the reason.
• An annotated workbook: a "Check" column per row, failing cells highlighted, a sheet of issues.

What it does not do: These are form checks: a valid SIRET does not prove the company is trading, nor a valid IBAN that the account is open. Nothing is looked up online.

Column profile

Before analysing an unfamiliar file, look at what it really holds: how many blanks, distinct values, numbers stored as text, values out of the ordinary. One card per column, and a report to keep.

• Filled and empty cells, distinct and repeated values, and the columns that could serve as an identifier.
• Mixed types and numbers stored as text, stray spaces.
• Minimum, maximum, mean and median; outliers by Tukey's rule.
• The most frequent values, the period covered by dates.

What it does not do: An "outlier" is out of the ordinary, not necessarily wrong: a large customer is still a customer. The profile reads the whole file into memory.

Pre-import check

The import into payroll, the ERP or the CRM stops at row 4,212 with an unreadable message. Here you describe what the software expects, and every row it will reject is listed before the import, with the reason.

• The format inferred from a template file, or from the file itself, then corrected on screen.
• Per column: the type, required or not, the maximum length, the allowed values, uniqueness, and a format check (IBAN, email…).
• Missing, extra or out-of-order columns, matched by name.
• The format is saved in a small file, to drop again with next week's file.

What it does not do: The check covers what can be seen in the file: it does not know that a customer code must exist in the software. To check one list against another, look up the two files.

Combine and compare

Merge Excel files

Branch reports, monthly exports, the sheets of one workbook: drop them in, check how the columns match up, and get a single table.

• XLSX, CSV and ODS files mixed, with all their sheets or only the ones you tick.
• Columns matched by name, ignoring case, accents and spaces; the matching is shown and can be corrected.
• A "Source" column that tells which file and sheet each row came from.
• The result as XLSX (dates and numbers kept) or CSV.

What it does not do: Values are carried over, not formulas or formatting: the merged table is a data table. Beyond 1,048,576 rows, the XLSX continues on a second sheet; CSV has no limit.

Look up between two files

Add each customer's name and city from the customer file to the orders file: no formula, no copying both files into one workbook, and knowing exactly which keys found nothing.

• The key suggested automatically when a column has the same name on both sides.
• The columns to bring over, your choice, added to the right of the main file.
• The usual causes of failure neutralised: extra spaces, case, a number stored as text on one side.
• Keys that found nothing counted and shown; duplicate keys in the reference flagged.

What it does not do: A duplicate key in the reference returns its first occurrence, as VLOOKUP does. Fuzzy matches ("Smith Ltd" and "Smith Limited") are not made.

Reconcile two lists

A bank statement and the general ledger, a customer file and the CRM export: each row looks for its row in the other list, by a shared reference or, failing that, by amount and the closest date. What remains are the differences, to review one by one.

• By reference (invoice or document number) or by amount and date, with a tolerance in days.
• An amount tolerance, and reversed signs from one list to the other (debit and credit).
• Each row is used only once: two payments of the same amount do not match the same receipt.
• Both lists' totals and their difference; a workbook with one sheet per category.

What it does not do: Matching by amount does not know that one 300 payment covers three of 100: groupings are still done by hand. Differences are shown, not corrected.

Split a file

Choose the column that divides (branch, sales rep, customer) or a batch size: every piece keeps the header, and they all go into a ZIP archive, or as sheets of a single workbook.

• By the value of a column: one file per value, named after it.
• By number of rows: regular batches, for an import that refuses large files.
• Each piece as XLSX or CSV in a ZIP archive, or a single workbook with one sheet per piece.
• A preview of the pieces and their row counts before splitting.

What it does not do: A workbook with sheets cannot hold more than 255 pieces: beyond that, choose separate files. Formulas and formatting are not carried over.

Compare two Excel files

Drop the old version, then the new one. Rows are matched by a key (an employee number, a reference) or by position, and you get an annotated workbook: green for what arrived, red for what left, yellow for what changed.

• The key suggested automatically when a column identifies every row; several columns allowed.
• Columns matched by name, even if their order changed; added or removed columns flagged.
• 12 and "12" are different, 0.1 + 0.2 and 0.3 are equal: we compare the way a spreadsheet calculates.
• An annotated result workbook, and a sheet listing every changed cell, before and after.

What it does not do: Values are compared, not formulas or formatting. Without a key, a row inserted in the middle makes everything after it look changed: the screen says so, and suggests a key when it finds one.

Differences between two lists

Two exports of the same customer file, yesterday's list and today's: what was added, what disappeared? Choose one column on each side, and the tool sorts every value by whether it is in one, in the other, or in both.

• What is only in the first list, only in the second, and what is common.
• Optionally ignore case, extra spaces and accents, to bring close spellings together.
• The result, one value per row with its side, saves to Excel, CSV or ODS.
• Two files, or two sheets of the same workbook.
• Everything happens in your browser, nothing is sent.

What it does not do: The tool compares one column against one column (a value is present or not); to match rows and bring back columns, see Reconcile two lists or VLOOKUP between two files.

Analyse and convert

Open a large file

Excel stops at 1,048,576 rows. Here, the file is read from your disk in chunks: browse, filter, sort, and export what you kept as CSV, or as XLSX if it fits.

• Files of several gigabytes, read in chunks: nothing is loaded whole.
• The exact number of rows, with irregular rows flagged.
• Combined filters (contains, equals, greater than, empty…), with numbers and dates read in your format.
• Sorting by column, and a profile of each column: most frequent values, minimum, maximum, sum.
• The result exported as CSV with no size limit, or as XLSX up to 1,048,575 rows.

What it does not do: Text files only (CSV, TSV, TXT), in UTF-8 or a Western code page; an XLSX workbook is limited to 1,048,576 rows by Excel anyway. Sorting keeps one key per row in memory: beyond three million rows of text, filter first.

Automatic pivot table

Revenue by sales rep and by month, number of orders by region: the tool recognises what can be pivoted, suggests useful pivots in one click, and calculates the table with its totals.

• Dimensions (categories, dates) and measures (amounts, quantities) recognised automatically.
• One or two dimensions in rows, one in columns; sum, count, average, minimum, maximum, distinct values.
• Dates grouped by day, month, quarter or year; percentages of the total; each period compared with the previous one, with its change as a percentage.
• The table and a bar chart on screen; in the workbook, the calculated table and its Excel chart, a real PivotTable that refreshes in Excel, and the data.

What it does not do: The calculated table is saved as values, to open anywhere; the PivotTable is the one that refreshes, in Excel (it builds itself on opening). The "distinct values" calculation does not exist in a classic PivotTable: it is delivered as values only.

Summarise a column

How many orders per region, how often each code comes back, the total per customer: instead of building a pivot table, choose a column and read its distinct values straight away, counted and sorted.

• The distinct values of a column, counted and sorted from the most frequent to the least.
• If needed, a total, average, minimum or maximum of another column per value.
• Empty values grouped separately.
• The summary saves to Excel, CSV or ODS.
• Everything happens in your browser, nothing is sent.

What it does not do: The tool groups on a single column. To cross two dimensions, or to get a chart and a native pivot table, see Automatic pivot table.

SQL queries

You speak SQL, your data is in three Excel files. Drop them: each becomes a table, and the query is written as on a database, joins included, without installing or importing anything anywhere.

• Each file (or each sheet) becomes a table; its columns are inserted with a click.
• SELECT, DISTINCT, inner and outer joins, WHERE, GROUP BY, HAVING, ORDER BY, LIMIT, CASE, LIKE, IN, BETWEEN.
• The common functions: COUNT, SUM, AVG, MIN, MAX, GROUP_CONCAT, UPPER, SUBSTR, REPLACE, ROUND, COALESCE, YEAR, MONTH.
• Examples written with your real column names, and the result as XLSX, CSV, ODS or JSON.

What it does not do: This is the part of SQL that is useful on tables: no subqueries, UNION or window functions, and nothing that modifies the files. Tables stay in the tab's memory: a few hundred thousand rows, not tens of millions.

Unpivot or pivot

A table with one column per month reads well, but cannot be filtered or pivoted. Unpivoting turns it into a list, one row per product and month; pivoting goes the other way.

• The columns to keep as they are, your choice; the others become rows.
• The names of the two new columns, your choice ("Month", "Amount").
• Empty cells skipped or kept.
• Pivot: the values of one column become columns, with duplicates flagged.

What it does not do: When pivoting, the same key with the same attribute keeps only its first value; the number of values ignored is shown.

Convert a spreadsheet

Drop one or more files and choose the target format. Each sheet becomes a CSV, a workbook or a JSON table.

• Input: XLSX and XLSM, old XLS (Excel 97-2003), ODS, CSV (any encoding), JSON.
• Output: XLSX, CSV for Excel (comma), CSV for European Excel (semicolon), ODS, JSON.
• Dates stay dates and numbers stay numbers, when the target format allows it.
• A multi-sheet workbook to CSV: one file per sheet, in a ZIP archive.

What it does not do: Values are converted, not formulas, charts or formatting. Old .xls files (Excel 97-2003) are read for their values, password included: you are asked for it when you drop the file.

PDF to spreadsheet

A statement, an invoice, a report exported as a PDF: the figures are there, but impossible to rework. The tool reads the PDF's text and its layout to rebuild the tables, page by page, in your browser.

• The tables recovered page by page, aligned into rows and columns.
• Numbers read as numbers (comma or dot decimal, parentheses for negatives).
• Export to Excel (one sheet per page), CSV or ODS.
• Everything happens in the tab: the PDF never leaves your device.
• For digital PDFs, produced by software. A scanned PDF is an image, with no text to extract.

What it does not do: The tool does not read scanned PDFs (images): there is no character recognition. Column rebuilding follows the text alignment; a very irregular table may need a pass of “Clean a report export” afterwards.

Understand formulas

Explain a formula

A three-line formula inherited from a colleague, a VLOOKUP returning something odd: paste it, each function is spelled out, from the outside in, and whatever is likely to go wrong is flagged.

• Formulas from an English or a French Excel, recognised automatically (, or ; separators).
• Each function in one sentence, its arguments named, nested functions numbered.
• The classic pitfalls: approximate match, range not locked with $, hard-coded column number, IFERROR too broad, date joined with &, volatile functions, division by zero.
• The formula in the other Excel language, ready to copy.

What it does not do: The formula is read, not calculated: without the workbook, the cell contents are unknown. A whole workbook's formulas are reviewed with the formula audit.

Translate a formula

A formula from a French colleague does not work in your English Excel: function names change, and semicolons replace commas. Everything is translated at once, in both directions.

• More than three hundred functions, from SUM to XLOOKUP, in both directions.
• Argument separators, decimals, array constants {1,2;3,4} and error values (#VALUE!, #NAME?).
• TRUE and FALSE, structured references ([#This Row]); sheet names and texts left untouched.
• The formula's spaces and line breaks kept.

What it does not do: TEXT format codes (dd/mm/yyyy, jj/mm/aaaa) are not translated: this is flagged. An unknown function is left as it is, and named.

Find the right formula

You know what you want, not which function does it. Choose the intention, give your cells: the formula is written in your Excel's language, and explained step by step.

• About twenty common intentions: look up, count and add up by criteria, flag duplicates, extract a first name, date gaps, percentage changes.
• Ranges locked with $ when the formula is meant to be copied down.
• The XLOOKUP version for Excel 2021 and Microsoft 365, or INDEX and MATCH for every version.
• An empty cell rather than an error, in one click, when the calculation can fail.

What it does not do: The intentions are the most common ones; a rarer need still means combining functions yourself. No artificial intelligence: formulas are written by rules, always the same.