What to remove before you send a customer file to software support: how to make a small file that still shows the problem

A support team often asks for the file that fails. This guide shows how to make a smaller file that still reproduces the problem, what to remove or replace first, how to test the result, and what to ask the vendor before you attach it.

An import fails, a report shows the wrong total, or a program stops on one file. You open a ticket with the software company, and the first reply asks you to attach "the file that causes the error."

The file is a customer export. It holds names, addresses and order histories, and the support team does not need any of them to find the fault. This guide shows how to send them a file that still shows the problem and holds almost nothing about real people.

Why does support ask for the real file?

Because most faults depend on the shape of the data and not on who the data is about. A date in an unexpected format, a comma inside a name, an empty cell where the program expects a value, a number with too many digits, a letter with an accent: any of these can stop an import. The person at the other end cannot guess which one it is, so the quickest request is "send us the file."

Sending the whole file is not the only way to answer. Usually the problem sits in one row, one column, or a pattern that repeats, and the rest of the file adds risk without adding information.

What does the support team really need?

Three things, in most cases.

1. A file that fails the same way. It does not need to be the full size. One failing row and a few ordinary rows around it are often enough.
2. The same structure. The column names, their order, the file type, and the way dates and numbers are written must match the original, because the fault often depends on them.
3. Values that behave like the real ones. A name with an apostrophe still has to have an apostrophe. A postal code with a leading zero still has to have one.

They do not need the real names, the real contact details, or the customers that were not involved.

How do you make a smaller file that still fails?

Work on a copy, never on the original.

1. Cut the rows. Keep the row that fails, if you know which one it is, and about ten ordinary rows. If you do not know which row fails, remove half the rows and try again. Repeat until the fault disappears, then put back the last half you removed. This usually finds the failing row in a few rounds.
2. Cut the columns. Remove every column the error message and the failing step do not mention. If you are not sure, remove them in groups and test after each group.
3. Deal with what is left. Look at the columns that remain. The next section explains what to do with each kind.
4. Run the failing step on the small file. If the same error appears, you have a reproduction file. If it does not, the removed part mattered. Put it back and try a smaller cut.

What do you do with the columns that remain?

Decide for each one whether the problem depends on the real value or only on its shape.

Free text deserves a close look, because a notes column can hold names and numbers inside ordinary sentences. The guide to removing personal details from a spreadsheet explains why. If the fault does not depend on the notes, take the column out.

What does this look like with an invented example?

A shop exports its orders and the import into its accounting program stops with the message "Invalid date in row 3,418." The export has 5,000 rows and 22 columns. Everything below is invented.

The first rounds of cutting show that the row is the one with the order number 100417, and that only four columns are involved: the order number, the customer number, the order date and the amount.

The customer's name, address and email address are gone, along with the 4,990 other rows. The customer number is replaced with a stand-in, and it is the same everywhere in the file. The date is kept as it is, because the fault is that 31 April does not exist, and an invented date that does exist would make the error vanish. The support team sees the failure at once.

What should you check before you attach the file?

What should you ask the vendor?

A short ticket is enough. Ask where the file will be stored and who can open it, how long it will be kept, and how to ask for it to be deleted when the case is closed. Many vendors give this in their terms or in the ticket form. If the answers do not exist, that is a reason to send less, and to send nothing that identifies a person.

Regulators ask the same kind of question when an organization hands data to a supplier. The Canadian privacy commissioner issued guidance on assessing third-party service providers in September 2026. It advises an organization to map how personal information flows between it and the provider, and to confirm each purpose for which the provider will use it. It also advises using contractual or other means to keep the protection comparable. Your own rules and contract may say more. Ask your data protection contact before you send anything that identifies a person.

How can DataAttendant help?

DataAttendant is a tool that runs in your web browser, on your own computer. You open a CSV or Excel file, and it suggests what to do with the names, email addresses, phone numbers, card numbers and other personal details it finds. You choose every treatment. Before you download the protected copy, you see a preview and a list of what remains. The copy is the same type of file, with the same rows and the same column names in the same order. A column you set to Remove is left out, or in a workbook stays empty when other parts of the file point at it. The same value gets the same stand-in everywhere in the file, and in later files for the same recipient, so rows that belong together stay linked.

In a free-text column, it looks for parts such as email addresses, phone numbers and card numbers inside the sentences, and it can change only those parts. It can miss a name in a sentence, so read the column, or take it out. It does not remove rows, and it does not decide which rows or columns the fault depends on. You still make the file small first, and you still test it. It cannot find every personal detail, and it does not decide whether you may send a file to a supplier. Your file is processed in your browser and is not sent to DataAttendant. How DataAttendant works shows the steps.

Sources

• Office of the Privacy Commissioner of Canada, "Guidance on assessing third-party service providers", page dated 10 September 2026 (open for comment until 4 December 2026). Read 2 November 2026.
• Related guides on this site: removing personal details from a spreadsheet, replacing real names with consistent stand-ins, hidden data in Excel files.
• Which law applies to your organization, and what your contract with the supplier requires, are not covered here. Your data protection contact decides.