Too many uncategorised transactions usually means the app has not yet learned your merchants, not that it is broken. The fastest fix is to categorise the biggest repeat merchants once, turn those choices into rules, and let history handle the rest. Twenty minutes on the top ten names often clears most of the list.
Why so many rows come in uncategorised
Bank statements describe UPI payments with long strings: a handle, a reference number, a bank code. The same shop can appear in five spellings. Until those are tied to a merchant, nothing tells the tracker whether it is groceries or fuel. Cash withdrawals and transfers to individuals are also hard to classify because the narration says nothing about the purpose.
Work from the biggest repeaters down
- Sort the uncategorised list by merchant name, or by amount if the names are messy.
- Pick the merchant that appears most often and categorise one row.
- Turn that choice into a rule so future rows follow it.
- Repeat for the next ten merchants.
- Handle the one-off rows last, by hand, in a batch.
SpendixAI categorises in stages: your own rules first, then your history and similar merchants, and only then an AI model if you have it switched on. Corrections you make become history, and you can add merchant aliases so different spellings of one shop map to a single name.
What does the pile look like after a first pass?
Say you imported 240 rows and 160 are uncategorised. Your top five merchants account for 70 of those: a milk vendor, a grocery store, a fuel station, a food delivery app and an auto-rickshaw handle. Five rules clear 70 rows. Another ten rules for repeat names clear perhaps 40 more. That leaves 50 rows, a manageable hand review of ten minutes.
| Step | Rows left | Effort |
|---|---|---|
| Start | 160 | None |
| Five top merchant rules | 90 | About five minutes |
| Ten more merchant rules | 50 | About ten minutes |
| Hand-review the remainder | 0 | About ten minutes |
What if we do not want AI involved?
You can switch AI off per feature, set a monthly AI budget, or turn it off completely. Unmatched items are then held for review, which is exactly the list you work through by hand. When AI is on, it works from the normalised merchant name and an amount band, not your name, account numbers or balances.
Keeping it clean next month
- Review new imports the same week, while you remember the purchases.
- Add an alias when one merchant shows under several names.
- Keep a category for cash so ATM withdrawals do not stay in limbo.
- Give a transfer to a person a real purpose once, with a note.
- Check the uncategorised count at month end; aim for near zero before running reports.
Rules that work well for UPI-heavy spending
UPI narrations often repeat a stable part, such as the merchant handle or a name, followed by a changing reference. Build rules on the stable part. Rules on the full string fail the next month when the reference changes.
- Use the shop or handle name, not the reference number.
- Keep one rule per merchant instead of one per amount.
- For person-to-person payments, categorise by purpose, such as rent share or household help, and add a note.
- Revisit rules every few months and delete ones that no longer apply.
A tidy category list also makes reports trustworthy. Fix the pile now, and each later month is mostly confirming rows rather than creating work from scratch.
Useful ideas start with a clearer picture of your own spending.
Explore your reports


