Many CSV import failures are caused by opening the file in Excel and saving it again. Excel reinterprets dates, trims leading zeros and rewrites long reference numbers in scientific notation, so the file that goes back to your budgeting app is not the file the bank produced. Edit a copy, check these three things, and import again.
How can opening a CSV break it?
When Excel opens a CSV it guesses a type for every column. A date such as 04/07/2026 may be turned into a number it formats its own way. A reference code like 000482915 loses its zeros. A long card or reference number like 4829155500123456 can become 4.82916E+15 and, once saved, the original digits are gone for good.
So keep one rule: never save over the downloaded file. Copy it, rename the copy, and do your repairs there. If something goes wrong, you start again from the clean original.
Three fixes, in order of how often they matter
Fix 1: format the reference column as text before opening
In Excel, use Data, then From Text or CSV, and in the preview set the reference column's type to Text instead of letting Excel detect it. Google Sheets lets you import with automatic conversion switched off. This way the digits stay as written.
Fix 2: put dates into one order
An Australian export is usually day first. A line dated 07/04/2026 means 7 April, but an importer expecting month-first reads 4 July. Compare a date with a day above 12 to confirm, then convert to YYYY-MM-DD. With true dates in A2, use Format, Custom and type yyyy-mm-dd. With text dates, =DATE(RIGHT(A2,4), MID(A2,4,2), LEFT(A2,2)) rebuilds them.
Fix 3: make the amount a plain signed number
| In the file | Problem | Becomes |
|---|---|---|
| A$1,085.40 | Symbol and thousands comma | -1085.40 or 1085.40, depending on direction |
| 22.50 DR | Suffix instead of sign | -22.50 |
| 1,800.00 CR | Suffix instead of sign | 1800.00 |
| Pending 14.95 | Text inside amount | Remove the word, or skip the row until it posts |
Check by totals. Say your file has three lines: a A$2,150.00 pay credit, a A$1,085.40 rent debit and a A$22.50 coffee debit. They should sum to 2150 - 1085.40 - 22.50 = 1042.10. Use =SUM(C2:C4) on your cleaned column; if it is not 1042.10 for those three, something is still text.
Which column names should I use?
Many importers match columns by name or position. Safe names are Date, Description, Amount, and optionally Type and Currency. Rename odd headers like Transaction Details or Narrative to Description. Remove columns you do not need, such as Serial or a running Balance, if the importer complains, but keep a copy with them.
What do I do with a file that is half broken?
- Make a copy and open it in a text editor to see the raw lines.
- Split it by month if it covers a year; smaller files are easier to diagnose.
- Import one month. If it passes, move on to the next.
- If a month fails, bisect: import the first half of the rows, then the second.
- Fix the offending row and recombine.
- Keep a note of which month and which row caused the trouble.
SpendixAI reads Excel and CSV exports with merchant, type, status and currency columns and presents rows for review before saving, which is a good place to notice a bad amount. It is built around India and INR, so mention it only if you also look after Indian accounts, and no layout is promised to parse perfectly.
Before you hit import
- Original file kept untouched.
- Reference numbers stored as text.
- Dates in YYYY-MM-DD.
- Amounts as signed numbers with no symbols or suffixes.
- Total of the file matches the movement on the statement.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

