If you fix the same merchant by hand every month, the answer is a rule: tell your tracker once that any transaction whose text contains a certain word belongs in a certain category. Then the fix runs by itself on every future import. This walkthrough uses US dollars and covers method only, so nothing here depends on any particular bank or country's rules.
Count how much time the repeat fixing really costs
Suppose you buy coffee at the same place about 14 times a month, and your tracker files it under Shopping instead of Eating Out. If you fix each one, that is 14 taps, and next month another 14. If the average coffee is $4.75, the monthly amount sitting in the wrong bucket is 14 × $4.75 = $66.50. Over a year that is $798 reported under the wrong heading, plus a lot of annoyance.
Compare that with one rule written in two minutes. The arithmetic is not subtle.
Pick the merchants worth a rule
Not every merchant deserves one. Sort last month's transactions by merchant and look at the ones you have fixed more than once.
- Weekly or monthly payees: the grocery chain, the gym, the phone company, the streaming service.
- Payees whose text is messy: names with store numbers or city codes tacked on.
- Payees that are always the same thing, such as a pharmacy or a petrol station.
- Skip general stores that sell everything, where the category depends on what you bought.
How do I write a rule that survives messy statement text?
Your statement may show the same store as CORNER COFFEE #1042 one week and CRNR COFFEE 1042 NYC the next. A rule written on the whole string will catch only one of them. Write it on the stable core.
- Collect three or four statement lines for the merchant and underline what they share.
- If they share a word such as COFFEE, check that no other merchant would also match it, because a loose word will start grabbing the wrong things.
- Use the longest shared fragment that is still safe, for example CORNER COFFEE.
- Set the category and, if you use sub-categories, the sub-category too.
- Run the rule on last month's transactions to see what it would have caught.
| Statement text | Contains CORNER COFFEE? | Category set by rule |
|---|---|---|
| CORNER COFFEE #1042 | Yes | Eating Out |
| CORNER COFFEE 1042 NYC | Yes | Eating Out |
| CRNR COFFEE 1042 | No | Needs a second rule or an alias |
| COFFEE BEAN WHOLESALE | No | Left alone |
Where a tracker's own learning fits in
Some trackers also learn from your history, so a fix you make once can influence later guesses. That helps, but history is a guess, while a rule is an instruction. For merchants you pay regularly, write the rule and treat history as a backup. SpendixAI, for example, applies your own rules first and history after that, and it is built around India and INR; if you use any tool that lets each expense carry its own currency, the dollar amounts above would simply be stored in dollars with a base currency for totals.
Keep a ten-minute monthly routine
After importing the month, spend ten minutes on three jobs. First, look only at the uncategorised or low-confidence rows. Second, for every merchant you had to fix, decide: rule or one-off. Third, look at the total for the two or three categories that moved most and ask whether the number feels right.
Say you work through 22 rows and decide 9 deserve rules. Next month, if the same 9 merchants appear 30 times between them, you have just saved 30 manual fixes for nine minutes of setup. The saving keeps compounding for as long as those merchants stay in your life.
Review the rules twice a year
Merchants close, rename or get replaced. Once in a while open your rule list and delete the ones for places you no longer use. Tidy rules are easier to reason about, and an old rule that suddenly matches a new merchant is a quiet source of errors.
- Rename vague rules so you know what they do.
- Remove duplicates that pull the same merchant to two categories.
- Add an alias when a merchant has changed its statement name.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

