In most bank statements you can tell a UPI debit from a debit card swipe or an ATM withdrawal by the first few characters of the narration. UPI rows often begin with the word UPI, shop-machine card payments often carry POS, and cash withdrawals mention ATM. Banks word these differently, so treat the prefixes below as clues and confirm them against your own statement.
What do the common prefixes look like?
| Narration clue | Usually means | How to tag it |
|---|---|---|
| Starts with UPI, followed by a handle or name | A UPI payment or transfer | UPI |
| POS followed by a merchant name or a masked card number | Debit or credit card used at a shop machine | Card |
| ATM or a cash withdrawal code | Cash taken out at a machine | Cash withdrawal, not spending |
| NEFT or IMPS followed by a name | A bank transfer | Transfer or bill payment |
| Name of an online merchant with a card number tail | Card used online | Card |
| Words like charges, fee or GST | A bank fee | Bank charges |
These are patterns, not rules. Open your own statement, pick ten rows of each kind you remember making, and write down how your bank spells them.
How do you tag hundreds of rows without opening each one?
Put the narration in column B and use a formula in column E that looks for the clues. In Excel or Google Sheets:
=IF(ISNUMBER(SEARCH("UPI", B2)), "UPI", IF(ISNUMBER(SEARCH("ATM", B2)), "ATM", IF(ISNUMBER(SEARCH("POS ", B2)), "Card", "Check")))
SEARCH finds text anywhere in the cell and ignores capitals. It returns a number when it finds the text, and ISNUMBER turns that into TRUE. Note the space after POS in the formula. Without it, the word DEPOSIT would be caught, because it contains the letters POS. Filter on Check afterwards and tag those few rows by hand.
Why does the tag matter?
- UPI and card rows are real spending, so they belong in your category totals.
- ATM rows are cash movement; counting them and later counting the cash spent double counts the same money.
- NEFT and IMPS rows may be rent, EMI, or a transfer to your own account, and each needs a decision.
- Fee rows are small but regular, and add up over a year.
A worked example
Say a statement for the month has 90 debit rows. The formula marks 58 as UPI, 17 as Card, 4 as ATM and 11 as Check. Check: 58 + 17 + 4 + 11 = 90. You then open only the 11 Check rows, which is a few minutes, and find six rent and EMI transfers, three bank fees and two cheque payments. Instead of opening 90 rows, you have opened 11.
What happens when you import into SpendixAI?
Choose the account before importing, then review the rows before saving. Categorisation runs in stages: your own rules first, then your history and similar merchants, and an AI model only for what is left and only if you have AI on. A correction you make becomes part of your history, so you do not need to repeat it next month. Layouts vary, so review the preview and do not assume every row was read correctly.
Quick check before you move on
- Write down how your bank spells each type.
- Run the formula and filter on the Check rows.
- Mark ATM and own-account transfers so they stay out of spending.
- Save the clue list so your formula stays the same next month.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

