In a multi-page statement import, page boundaries produce three kinds of stray rows: repeated column headers, running-balance carry-forward lines, and narrations that wrap onto the next page and arrive as a separate row. All three can be spotted in the review step and removed or merged before you save.
Why do page breaks cause trouble?
A statement is laid out for paper. Each page repeats the bank's heading and column titles, and many show the balance carried forward from the previous page. When a PDF is read back into rows, these printed furniture lines can look like transactions.
The three culprits and how they look
| Stray row | What it looks like in review | What to do |
|---|---|---|
| Repeated header | Words such as Date, Narration and Withdrawal in the description, with no amount or date | Remove it |
| Carry-forward line | 'Brought forward' or 'Carried forward' with a balance figure but no real movement | Remove it |
| Wrapped narration | A row with no date and no amount, only the tail end of a long description | Merge the text into the row above, then remove it |
The wrapped narration is the sneakiest. A UPI payment with a long reference may run across two lines, and if the break falls at the page end, half of it lands on the next page.
A 12-page statement, step by step
Say a year-long statement runs to 12 pages with about 40 rows a page, roughly 480 rows. Reviewing that is easier in stages.
- Look at the first row and last row first. They should be the first and last transactions of the period.
- Scroll to where each page would end, roughly every 40 rows, and look at the rows on both sides.
- Remove rows that are only headings or carry-forward lines.
- Merge wrapped narration rows into the transaction above where the review step lets you edit; otherwise remove the fragment and fix the narration after saving.
- Count the rows and compare with the statement before you confirm.
What if I find them after saving?
Stray rows that were saved will inflate totals or add junk to your categories. Find them by looking for a zero or blank amount, a very odd description, or a date that repeats. Delete them, and remember deleted transactions stay in the recycle bin for 30 days in SpendixAI.
Can I avoid the problem altogether?
Yes, partly. A CSV or Excel statement is one continuous table with no pages, so there are no page breaks to trip over. If your bank offers it, use that. If only PDF is available, split a long statement into shorter periods, such as a quarter, so there are fewer boundaries to inspect each time.
A quick test for a clean import
- The first and last dates match the statement.
- The row count matches, once you leave out headers and carry-forwards.
- Opening balance plus credits minus debits equals closing balance.
- No row is missing a date or amount.
A page-by-page rhythm
On a long statement, work in a rhythm rather than all at once. Check page one's rows, then the join with page two, then page two's rows, and so on. Because each page break is where trouble lives, you only have eleven joins to look at on a twelve-page file, not four hundred and eighty rows. At each join, ask two things: does the narration at the bottom end properly, and does the next row begin with a date and an amount? If both answers are yes, move on. This keeps a tiresome task manageable and makes it far less likely that you stop paying attention halfway.
None of this promises that every layout will parse perfectly, which is why SpendixAI shows the rows for review before saving. A few minutes spent on page boundaries now saves hunting for a mismatch across a whole year later.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

