Desktop App Documentation for Windows

Install and First Run

Download the installer from the desktop page and run it. The installer is digitally signed, and the desktop page publishes the certificate and the file size so you can check both before you install anything. Windows shows a warning on first run for any software that is not from the Microsoft Store. That warning is about the publisher, not about the file, and the signature is how you tell the difference between our installer and a copy someone else renamed.

The app opens with the demo dataset already loaded, so you can see what a result looks like before you open anything of your own. The seven modules run down the left side, and they unlock once a licence is entered. Work in the app is saved to a workbook in your Windows user profile as you go, and the app saves again when you close it, so a job interrupted by a restart does not lose everything.

Enter your licence in My Account before the paid modules will run. The licence id is in the purchase email, in the form FLD-XXXX-XXXX-XXXX-XXXX, and it is paired with the email address you bought with.

The Data Grid and Ranges

Open a file with the Open button in the File menu. Commas, tab separated text and Excel workbooks are all read, and a workbook with several sheets imports every sheet with each one named. Files over 256 MB are refused with their size rather than attempting to load them, so split a larger file and open the parts.

Every module takes a range, and you give it one in three ways:

• Click a column header in the data grid to select that whole column
• Type a range into the field, for example A1:B200
• Select cells in the grid, which fills the field in for you

The grid shows the first 300 rows of the active sheet. The row count under the column picker is the count for the whole sheet rather than the part on screen, so it tells you the size of what you are working with. The status bar states what ran and what it found.

The area to the right of the controls has two tabs, Data and Result. Data is the grid. Result is the table a module produces, and it takes the full width of that area rather than sitting inside the form. A profiling table has eight columns and squeezing it into a form column leaves about 38 pixels for each one, which is too narrow to read. Pressing a button that produces a result switches to the Result tab for you, and changing module returns you to Data.

Profile and Insights

Profile reads a range and reports what is in it, so you can decide what to fix before you change anything.

Run profile returns a table with one row per column: the detected type, how many cells are filled and empty, how many distinct values there are, how many duplicates, the average length, and any issues found. Below it you get the row, column and issue counts.

Column groups is the other half of this module. It groups values that are probably the same thing written differently, and for each group gives you the value, how many times it appears, how many case variants it has, and the near matches it found. This is where a column written as ACME LTD, Acme Ltd and Acme Limited stops being three entries and becomes one group with three members.

Clean and Transform

Clean rewrites the values in a range so they are consistent, which is what makes the rest of the app work on them.

The five operations are:

• Remove stop words, using the words in your stop list range
• Remove listed punctuation, using the punctuation in your stop list range
• Remove diacritics, turning Zürich into Zurich
• Extract domain, taking the domain out of an email address or a URL
• Extract path, taking the path out of a URL

Removing diacritics is the one that earns its place most often. It is the difference between Globex and Globex not matching, and it is reversible in the sense that matters: you still have the original column, and you can add a cleaned column beside it rather than overwriting.

Match and Dedupe

This module is the core of the app. It finds records in one range that are the same entity as records in another, on a scale of similarity rather than on an exact key, and then either merges them or removes the duplicates.

Five buttons run it, and they differ in what they change:

• Find duplicates reports the groups and changes nothing. Run this first. It is the only one of the five that cannot damage your data, and it tells you how big the problem is before you commit to a threshold.
• Explain takes one record and shows you its nearest matches with the score for each, so you can see why a threshold behaves the way it does rather than guessing at it.
• Golden records picks the best record from each group and shows you the choice, without writing it yet.
• Merge and purge writes the master values onto your data range and removes the rows that were merged away.
• Remove duplicates collapses the duplicate groups in place, keeping the occurrence you chose.

On the threshold: 0.85 is a reasonable starting point and it is the wrong one to keep for every column. Acme Ltd and Acme Limited should match, and at 0.9 they may not. Run Find duplicates, look at the scores, and lower it until you are seeing the matches you expect. Use Explain on anything surprising before you let the run write.

Data Verification

Verification checks a list against a master list and tells you what is wrong, which is different from cleaning it.

The four entity types are Country, US state, Company suffix and Top level domain. The type matters because it sets how much variation is tolerated. A company name can be written Acme Ltd, Acme Limited or Acme, Ltd. and still be the same company, so matching has to be forgiving. A country cannot, so the same matching applied to a country column quietly accepts errors that look like typos and are not.

Verify and enrich reports what it found and adds the fields you asked for. Cross match compares the two lists against each other without adding anything, which tells you how much of your data is trustworthy before you rely on it.

Automation

Automation runs a job on a schedule without you, on the same computer, against the same workbook.

Each scheduled job in the list has its own controls:

• Pause stops the timer and keeps the job and its run history. Resume starts it again and keeps the remaining wait rather than restarting the interval
• Run now runs it immediately without waiting for the schedule
• Stop removes the job entirely
• Clear all removes every scheduled job

A job that runs while you are asleep is only possible while the app is running, because a scheduled job is a task on your computer rather than a service somewhere. Run on startup if missed is what covers a machine that was off. If you need something to run whether or not your computer is on, that is what the Sheets add-on does, and the automation documentation for the add-on covers it.

Import from a Source

Import from a source pulls a table out of an online system and opens it as a normal table in the workbook. Every other module then works on it exactly as it would on a file you opened, and the table can be saved to a file afterwards like any other.
Three sources are supported: Google Sheets, Salesforce and Airtable.

Google Sheets

Enter the spreadsheet id from the spreadsheet's url, and a range in A1 notation. The default range is A1:Z10000, which reads the used area of the first sheet. The sign-in asks for read-only access to spreadsheets, so the token this creates cannot change anything in your Google account.

Google requires every application that signs in to be registered before it can do so, which is why a client id is needed in Settings. Create a Google Cloud project, enable the Google Sheets API, set up the OAuth consent screen and create an OAuth client of type Desktop app, then paste its client id into Settings. Until a client id is entered the connector reports that it is not configured. The app does not ship one, and no client secret is used or stored.

Salesforce

Enter a SOQL query, for example SELECT Id, Name, BillingCity FROM Account. Each page holds at most 2000 records and the remaining pages are fetched automatically. The sign-in needs a connected app with the API and refresh token scopes enabled, whose consumer key goes into Settings. A sandbox org is enough to try this; no external review is needed.

Airtable

Paste a personal access token with the data.records:read scope, the base id which begins app, and the table name or id. Pages hold up to 100 records and are fetched in turn. Airtable omits empty fields from its responses, so those cells arrive empty.

Where the sign-in is stored: tokens are written to your Windows user profile alongside the rest of the app's data, and never leave the machine.
Nothing is sent to us. If you would rather not keep a sign-in, press Disconnect, which removes it.

Writing back

Write back sends your cleaned values to the source again.
It updates records that already exist and nothing else: no record is created and none is deleted. You name the column that identifies a record, usually an id or an email address, and tick the columns to write. The app then matches each local row to a record on that column and writes only the rows that matched.

Because this changes data outside the app, the first press is always Preview write back. It reports how many rows matched, how many were skipped, and shows the first ten changes without sending anything. Apply write back is only live after a preview, and the button then reads Confirm write back, so a write cannot happen by one press or by accident. Every write is recorded on the activity list in My Account.

Undo keeps the real previous values, captured while the records were being read, rather than reconstructing them later. A field that was empty before the write is restored as empty. An undo that has been applied is dropped from the list, so the same write cannot be rolled back twice.

A write back counts only the records the service confirmed. If Airtable or Salesforce reports a record as rejected, that record is named in the result and it is left out of the undo record, because it never changed. A write that stops part way through still records an undo for the records that did change, so those can be rolled back on their own.

The write back addresses the cells or fields the pulled table actually occupies. A Google Sheets table pulled from a range that does not begin at column A, or from a named sheet, is written back to that sheet and those columns rather than to the top left of the first sheet, and columns that are not next to each other are written as separate requests so a column between two written ones is never overwritten.

Each pull remembers where its table came from. When you open a pulled table and go back to Import, the panel says which source it belongs to and fills the spreadsheet id, the range, the object or the base and table in for you, so a write back does not need an id typed a second time. Only the locator is kept, never a token.

Permissions differ between the three. Airtable needs a token with the data.records:write scope. Salesforce writes through the same sign-in, which already carries the API scope. Google Sheets needs a broader permission than reading alone: its values.update resource accepts only the full spreadsheet scope, so a Google sign-in created for write back can change any spreadsheet that account can open. Reconnect Google Sheets after this release, because a token issued earlier for reading cannot write.

My Account

My Account holds the licence, the plans, the update check and the list of computers the licence is active on. It is grouped into Status, Activate, Plans, Updates and This device.

The version number and the integrity result are both in this panel, and both are worth including if you contact support. The integrity check is the one that matters: it compares the engine files against the hashes the app was built with, so a clean result is arithmetic rather than a promise.

To move to a new computer, use Release device on the old one first. One licence covers one seat on one computer, so the seat has to be free before the new machine can take it. Releasing does not cancel the subscription, it only detaches the machine.

Licences Are Bound to Your Device

A desktop licence is tied to the machine rather than to your account. The binding combines the machine GUID, the disk volume serial, the processor identifier, your Windows account name and your profile path, hashed together. A licence id that leaks therefore cannot be used on someone else's computer, and a second Windows account on the same machine does not inherit the licence.

The cost of that choice is that a licence stops matching if you reinstall Windows or rename your account, because both change the binding. You enter the licence id again in My Account and you are working on the same licence the same day. Your other devices are unaffected, and a reinstallation is not a cancellation.
We would rather you know this before you rebuild a machine than discover it afterwards.

The licence is verified on your own machine, so nothing about it depends on a server answering at the moment you need it. It does need to hear from us once every seven days, and the app checks by itself about once a day, so that happens without you doing anything. If the app cannot reach us for seven days running, the paid modules pause and scheduled jobs are skipped, and they resume on their normal schedule once it can check again. My Account shows the days remaining.
Your work is never deleted and no file of yours is held back because a server is unreachable.

Reporting a Problem

If something behaves in a way you cannot account for, tell us through the contact page. Three things make a report answerable rather than a guessing game:

• The version number, shown in My Account under Updates
• The result of Check integrity in the same panel
• What you expected and what happened instead, with a row or two of the data that caused it if you can share it

The app writes a log file at %APPDATA%\flookup-desktop\logs\app.log, capped at 1 MB with the oldest lines dropped. It records what the app did and never what your data contained, so it is safe to attach. Type %APPDATA%\flookup-desktop into Explorer to reach it.

If a signature check fails, or a hash does not match, do not run the file. That is not a licensing problem and support cannot help with it, because the file is not the one we published. The desktop page has the certificate and the size to compare against.

Frequently Asked Questions

You Might Also Like

• Download the Desktop App
• Pricing and Plans
• Google Sheets Add-on Documentation
• Data Cleaning Automation
• How to Remove Duplicates in Google Sheets