A file that cannot be used is refused on this step with one of these messages: {name} is not a CSV file, The file is larger than 10 MB, The file could not be read, The file is damaged: a quote opened on line N is never closed., The file has no transaction rows, or The file has N rows (more than 3,000). Each is explained in Prepare your CSV file.

Download template gives you a CSV named import-template-<account>.csv with the export's main columns — Date, Description, Amount, Currency, Type, Category, Counterparty — followed by your workspace's attributes, and 14 example rows that already use your own categories, counterparties and options, dated backwards from today so none of them is planned. Next to it, How to prepare a file opens the formatting guide, and Need help opens the support chat.

To add an attribute that does not exist yet, choose Create new attribute at the bottom of the Import as list — this needs the Data schema changes permission; without it the row reads Creating attributes needs the Data schema changes permission. The attribute window opens with the file column's header already filled in as the name, and Income and Expense ticked — the import never creates transfers. When you save, the column is mapped to the new attribute, and the attribute is made visible in your transactions table.

Check the values — the list on the left of the window, or a row of sections you scroll sideways on a phone — shows what needs a look: the date, actual date, amount and direction columns, each category, counterparty and select-attribute column, Split transactions when the file has any, attribute columns with values that cannot be read, Invalid rows and Summary. Each item names the Vault field and, when it differs, the file's own column header underneath. A number badge shows how many values or rows need attention.

Choose the Date format read from your file: ISO (2026-09-24), European (24.09.2026) or American (09/24/2026). The one marked Suggested was detected. When every date in the column reads both ways, nothing is preselected and the panel says Every date in this column reads both ways. Confirm the format before continuing. The table shows each value In the file and how it is Read as; a value that cannot be read shows Not a date.

A date may carry a time. Read as then shows the day and the time, for example 2026-09-24 10:15. A time without a time zone is taken as wall-clock time in your workspace's time zone. A time with a time zone, such as Z or +02:00, is converted to your workspace's time zone, and the row takes the day it lands on there: 2026-09-24T23:30Z in a Europe/Vienna workspace becomes 2026-09-25 01:30. Such rows are marked from {value in the file} next to the result, so a shifted day is never applied silently. A date without a time is stored at noon in your workspace's time zone.

Shown only when a column is mapped as Actual date: A single date or a period, «from - to», read with the date format chosen for the Date column. A date without a time is taken at noon. A period reads as start - end, each end with its time. A value that is neither a date nor a period shows Cannot be read as a date or a period; such rows are still imported with the actual date left empty, and the panel says so: N actual dates could not be read and will be left empty, with Examples.

A period written with its start after its end is listed under N rows have a reversed actual period: the start is after the end. One choice applies to all of them — under What to do with these rows pick:

Choose the Decimal separator, Dot (1,234.50) or Comma (1.234,50), again with a Suggested choice and a table of how each amount is read. When the column reads both ways the panel says The amounts in this column read both ways. Confirm the decimal separator before continuing.

With Signed amount and no negative value anywhere, the panel warns: No amount in this column is negative — all N rows would become income. Is the direction in another column?

With Signed amount and a Type column in the file, a row whose sign disagrees with its Type word — a minus on a row that says Income, or no minus on a row that says Expense — is a conflict. The panel lists them under N rows have a sign that contradicts their Type, with the note Choose one rule for all of them. The import waits until you do. Under Which is right? pick one:

Until you choose, the panel says Pick a rule to continue. and Import stays off.

Shown only with Amount and direction. Direction values lists each distinct value of your direction column with its Rows and what it Means: Income or Expense. Common words are assigned for you; any other value shows Choose an option and Not mapped — its rows are invalid until you assign it. A signed amount that disagrees with the direction column is listed under N rows have a sign that contradicts the direction — They will be imported by the direction column. Review them before continuing.

Each distinct value in the file is a row of the table: Value in the file, how many Rows use it, its Type — Income or Expense, because the same name can exist as an income option and as an expense option — and what it is Imported as. On a phone, Imported as comes second, right after the value, so the picker is in view without scrolling sideways.

The Imported as cell is a picker. Its list holds the existing options valid for that side — each with its colour, a child option shown as Parent / Child — and then the actions: Create «{value}» for a value nothing matches, or Restore «{value}» for a value that matches an archived option, and Skip. The cell shows the current state:

A value written as Parent / Child is matched as the child option under that parent. If it has to be created, it is created as a child under the existing parent of that name; when no such parent exists, the parent is created too.

Create missing options at the top of the panel is on by default when you may create options; the note under it counts the values with no match. Switch it off to decide value by value. A value that matches an archived option is never created a second time. Without the Data schema changes permission the panel says Creating options needs the Data schema changes permission. Map or skip the value instead. When the list has nothing to offer, it says so — No income categories yet, No options yet and so on.

An attribute value on a row of a side the attribute does not apply to is left empty, and the panel says so: N values in {column} will be left empty: {attribute} does not apply to income rows (or expense rows).

Shown when the file carries split cells — the form a Vault export writes: Food (400.00); Insurance (600.00). The parts are restored with their categories, counterparties and attribute values, and each part's label is decided on its column's own page like any other value.

The panel lists the restored splits under Restored: N split transactions, N parts — one line per part with its Amount and Part values. Each value shows what the part becomes: the option it maps to (Fuel → Rent when it differs from the file's label), + Create «{value}», Restore «{value}», Empty, or Choose an option while undecided.

A split that cannot be restored is imported as one transaction without the split. The panel says N split transactions cannot be restored and will be imported without the split and lists each row with its Reason: The columns list different numbers of parts, The part amounts differ between the columns, The parts do not sum to the amount, A part cannot be read as the column's value, or An option name in this workspace contains the separator. Under Value for these rows there is one line per split column and side; its picker offers Empty (preselected) and the existing options valid for those rows. Nothing is created from the cell.

A number, checkbox or timestamp attribute column with values its type cannot read gets its own item: N values in {column} could not be read as {type} and will be left empty, with Examples. The rows themselves are still imported.

Rows that will be skipped, each with its Row number, its Reason and the row's own values: These rows are skipped. Fix them in the file and import it again to include them. A row with more or fewer fields than the header is listed with the reason N fields, the header has N and its raw text, because its cells cannot be lined up with the columns. The full list of reasons is in Prepare your CSV file.

N rows will be imported, then the skipped rows with each reason and its count, the Income and Expense totals with their row counts, and the notices that apply:

Two things make an imported row a planned transaction:

Every planned row is created as a one-off planned transaction, never as a repeating series — a recurring series you exported comes back as the individual occurrences that were in the file. Planned rows do not change the balance.

When the account already has a balance other than zero, the summary shows Account balance with two options and the note Choose one to import. Import stays off until you choose.

Each option shows the balance you will have after the import. Rows dated in the future and rows marked Planned do not move the balance in either case.

Worked example. The account has a balance of $1,000.00. The file has 6 rows to import: 2 income rows totalling +$6,000.00 and 4 expense rows totalling −$1,714.36, one of which (−$320.00) is dated in the future.

Undo lives only on that message. It stays on screen until you use it or close it. Once the message is gone there is no other place to undo the import — you would remove the transactions by hand.

Undo import removes the transactions the import created, together with their split parts, and takes their effect out of the account balance, including the adjustment made by Keep current balance. Transactions you added after the import are not touched; imported rows you edited are still removed. Categories, counterparties or attribute options created or restored for the import stay in the workspace. When it is done you see Import undone.

If the import stops with an error, Try again repeats the same attempt — options already created by the first attempt are reused, not created again.

If you import a file that was already imported into the same account, the import stops and asks: This file was already imported — It went into {account} on {date}. Import it again? Choose Cancel to stop, or Import again to import it once more. An import that was undone does not trigger this question.