When a PDF conversion puts a long narration on two rows, the second row has text but no date and no amount. Join each such row's text to the transaction above it with a helper column, paste the result as values, then delete the leftover rows.
What the problem looks like
Bank narrations such as UPI strings, NEFT references and card merchant lines are long. The PDF wraps them onto a second printed line, and the converter turns that line into its own spreadsheet row. You end up with something like this, where the second row is only a tail:
| Date | Narration | Debit | Credit | Balance |
|---|---|---|---|---|
| 05/01/2024 | UPI-SHARMA GENERAL STORE-SHARMAGS | 640.00 | (empty) | 18,360.00 |
| (empty) | OKBANK-4012XXXX-GROCERY | (empty) | (empty) | (empty) |
| 06/01/2024 | NEFT-SALARY-ACME PVT LTD | (empty) | 52,000.00 | 70,360.00 |
Hypothetical numbers, but the pattern is the real one: a blank date, a blank amount and a bit of narration that belongs to the row above. (The word empty in the table stands for a cell with nothing in it.)
How do I tell a continuation row from a genuine row?
Do not rely on the date alone. Some statements leave the date blank for the second and third transactions of the same day. A continuation row is one where the date, debit, credit and balance are all empty and only the narration has text. Test for that, and you will not glue two real transactions together.
The helper-column formula
Assume Date is in column A, Narration in B, Debit in C, Credit in D and Balance in E, with data starting on row 2. In F2 enter the formula below and fill it down to one row past your last data row, so the final formula looks at an empty row:
- =TRIM(B2&IF(AND(LEN(TRIM(A3&C3&D3&E3))=0,LEN(TRIM(B3))>0)," "&F3,""))
Read it from the inside. A3&C3&D3&E3 joins the next row's date, debit, credit and balance into one string. If that string is empty and the next row's narration is not, the next row is a continuation, so the formula appends F3, which is that row's own full text. Because each row looks at the row below it, a narration wrapped over three lines is picked up correctly: the third line joins the second, and the second then joins the first.
Turn the result into real data
- Select column F, copy it, then Paste Special, Values, over itself. This is important: the formulas depend on neighbouring rows, so they will break as soon as you delete rows.
- Copy column F over the old Narration column B as values, or keep F as the new narration and delete B.
- Switch on filters, filter the Date column for Blanks and the Debit and Credit columns for Blanks too, and confirm every visible row is a continuation tail.
- Select the visible row numbers, right-click, Delete, then clear the filter.
If your statement leaves the date blank on legitimate same-day rows, filter on the Balance column being blank as well. Most statements print a balance against every real transaction.
Did the row count come out right?
Count before and after. Say the converted sheet has 190 rows under the heading, of which 150 are real transactions and 40 are tails. After deleting the tails you should have exactly 150. Use =COUNT($E$2:$E$500) on the Balance column, which only counts rows with a numeric balance, and compare it with the number of transactions you counted on the statement or with the difference in the running balance.
Then spot-check three or four of the longest narrations to make sure the pieces joined in the right order and there is a single space between them, not none. If two words have run together, edit those few cells by hand.
A shorter way next time
- Some converters have a setting to unwrap text; look before cleaning by hand.
- If the bank offers a CSV or Excel download, narrations usually arrive in one cell, so the whole problem disappears.
- SpendixAI reads PDF, CSV and Excel statements and lets you review rows before saving them. How well a given PDF layout reads varies, so treat the review step as part of the job.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

