Importing Transactions

Rather than looking for a specific bank, Cadence inspects the header row and matches your columns using flexible heuristics. As long as your file is a standard CSV with one transaction per row, it can almost always read it, even when the column order is unusual.

Moving from another budgeting app is a first-class case. Files from apps not on the auto-detect list - including EveryDollar - still import fine; you may just need to confirm a column or two in the mapping step.

1. Open the + menu and choose Import Transactions.
2. Upload your CSV from your device.
3. Enter a Source Name for these transactions, or map a Source column from your file, and pick the source type.
4. Review the column mapping. In most cases Cadence has detected everything correctly.
5. Preview the parsed transactions and check the dates, amounts and merchant names.
6. Confirm to import.

Nothing is written until you confirm, so it is safe to back out if something looks off. Once confirmed, transactions are added in chunks.

Cadence tries to match each column to one of its standard fields. Only three are required:

If your file uses non-standard headers, or no headers at all, map each column by hand with the dropdowns in the mapping step. Unmapped columns are ignored, so a file with extra fields you do not care about is fine.

Credit card statements often format charges as positive numbers and payments as negatives, the opposite of a chequing-style export. On the upload step, set the source type to Credit Card to flip the signs, so charges appear as expenses and payments as income.

If your preview looks inverted - income where expenses should be - go back, switch the source type, and preview again.

After you confirm, Cadence checks your imported rows against your existing synced and manual transactions to catch double-counts. A row is treated as a duplicate when it matches a transaction from a different source on date (within three days), amount (within one percent) and sign, with a similar name.

Detection runs server-side once the import is committed. There is no per-row duplicate flag in the preview; instead Cadence hides the duplicates it finds and reports how many were skipped on the results screen.

Anything Cadence hides stays reviewable under Settings → Data & Privacy → Review Duplicates, where you can confirm the match or undo it and bring the hidden transaction back.

Imported transactions go through the same categorisation as synced ones, and it runs automatically. There is no category-mapping step. Cadence works through, in order: what it has learned from how you categorise that merchant, its own merchant and keyword patterns, and finally the category column from your file if you mapped one. It then links each transaction to a budget based on where it landed.

There are no import rules to write. Cadence has no rules engine, so you cannot author a “if the name contains X, use category Y” rule. Instead it learns: correct a merchant a couple of times and it applies that category from then on, including on future imports. Every change you make on the preview step feeds that learning.

To start over, Settings → Suggestions has a Clear Learning action. More in Smart Features.

A few patterns trip up imports more than others. If the preview looks wrong, check these first:

• Currency symbols or thousands separators in the amount column. Clean them out before re-uploading.
• Day-first dates. Slash and dash formats are read month-first, so reformat the date column to YYYY-MM-DD if your export is day-first.
• Blank or summary rows at the top or bottom of the file. Delete them in your spreadsheet before exporting.
• Mixed currencies. Import them as separate files, one per currency.
• Statements combining several accounts. Split them apart so each import covers one account, and give each its own Source Name.