Narrations arrive cut off or garbled because the statement itself had them that way: some banks print a fixed number of characters per line, PDFs can wrap one description over several lines, and odd characters come from how the file was encoded. You can usually still use the rows. Add a merchant alias or a rule on the part of the text that is stable, and if the mess is serious, download the same statement in another format and import that.
What causes truncated or strange text?
- Fixed-width fields: some statements allow only a set number of characters for the description, so a long UPI narration is chopped mid-word.
- Wrapped lines in PDFs: a long description is printed across two or three lines, and a reader can pick up only the first line, or glue lines together in the wrong order.
- Encoding: a CSV saved in one character set and opened in another turns symbols such as the rupee sign into rubbish characters.
- Special characters: ampersands, slashes and accents inside a merchant name can look different once converted.
- Page breaks: a transaction that straddles two pages of a PDF can lose its tail.
None of these means the bank got the amount wrong. The amount and date columns are usually fine even when the text next to them is ugly.
Can I still categorise a row with cut-off text?
Yes, as long as the part that identifies the merchant survived. Say the full narration was a long UPI string ending in a food delivery name, but your statement shows only the first 30 characters and the name is gone. Then the row carries no merchant clue and you categorise it by hand once. If the name is there but misspelt, for example the last letter missing, that is a job for a merchant alias, which maps the odd spelling to a clean name.
| What you see | What it tells you | Sensible fix |
|---|---|---|
| Merchant name present but clipped at the end | Fixed-width field | Alias or rule on the visible part |
| Name missing, only a reference number | Text cut before the name | Categorise by hand, use the amount and date as clues |
| Rupee sign shown as strange symbols | Encoding mismatch | Re-download or re-save the file as plain CSV |
| Two halves of one narration on separate rows | Wrapped PDF line | Try the CSV or Excel version of the statement |
Should I try another file format?
If the bank lets you choose, try that. A CSV or Excel download keeps each field in its own cell, so the narration is not subject to PDF line wrapping. A PDF is built for printing, and reading text out of it depends on how it was laid out. It is worth five minutes to download the same month as a CSV and compare the narration for one or two awkward rows. Importing a statement is something you can review before saving, so you can look at the rows in the preview and decide before anything is kept. No promise can be made that every layout will read perfectly, and that is why the review step exists.
What do I do with rows that are already imported?
First decide whether it is worth fixing at all. If the amount, date and account are right and you can recognise the row, a clean category is enough. For repeat merchants, the better move is to teach the tracker once.
- Open the row and read the narration in full.
- Find the shortest stable piece of text that appears every time, usually the merchant name.
- Add an alias so that the clipped spelling becomes one clean merchant name.
- Create a rule for that name so the next import categorises it by itself.
- Re-check the next statement to see if the same row now lands correctly.
Aliases are useful here because they sit in front of everything else. Once a messy string is mapped to a clean name, your rules and your history can match it. SpendixAI categorises in stages, starting with your own rules, then your history, and only if AI is switched on, an AI model for what is left, so a clean name helps at every stage.
When is it worth asking the bank?
If the same fields are unreadable month after month, check whether your bank offers a fuller statement format, or whether the app and the web statement differ. Describe the problem in your own words, for example that long descriptions are cut after a fixed length. Never send anyone your login details for this; a statement file you downloaded yourself is all that is needed.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

