A categorisation rule is a standing instruction such as: if the narration contains this text, put the transaction in this category. The useful questions are whether it applies to future imports, whether it can be applied to rows you already have, and whether you can see what it will change before it does. Check those three in whichever tool you use, then test the rule on a small set first.
What a good rule looks like
A rule has a match and a result. The match should be a distinctive part of the narration, long enough to be specific. The result is a category, and in some tools also a sub-category. If you write a rule that matches the word Uber, check whether your statements also contain Uber Eats, which is food and not commute. That is the sort of case where a longer match, or a rule that reads the amount too, saves you from wrong rows.
A worked example with made-up numbers
Say you drive for a platform and also take a cab to the airport now and then. Over three months your statements contain these rows.
| Narration contains | Rows | Total in rupees | Right category |
|---|---|---|---|
| Uber trip | 14 | 6,300 | Commute |
| Uber Eats | 5 | 2,150 | Food delivery |
| Ola ride | 6 | 2,700 | Commute |
A rule on the word Uber alone would put 14 + 5 = 19 rows, worth ₹8,450, under Commute, and ₹2,150 of that is food. A rule on the phrase Uber trip catches only the 14 trips worth ₹6,300. The difference is ₹2,150, which is why a preview step matters.
How do I apply a rule to old transactions?
This depends entirely on the tool. Some tools apply a rule only to rows imported afterwards. Others offer a run option that sweeps existing rows, ideally after showing which rows would change. If yours has neither, the fallback is to filter existing rows by the same text and change their category in bulk.
- Filter existing transactions by the narration text.
- Look at the list and remove any rows that do not belong.
- Apply the new category to the rest.
- Save the rule so future imports follow it.
- Spot-check the next import.
What SpendixAI does with rules
In SpendixAI your own rules come first in the categorisation order, ahead of your history and similar merchants, and an AI model is used only if you have switched it on. Corrections you make become history. Check whether your account offers a preview or a way to run a rule on old rows, and use the filter-and-change route for past rows if it does not.
Why does this matter for gig workers?
- Rides, fuel and parking can be work costs or personal, and the same payee may appear in both.
- Platform payouts and personal transfers can look alike in a statement.
- Clean categories make it easier to show your accountant a month of work costs.
- Whether any cost is allowed for tax is a matter for your CA, so use the categories only as clean records.
A short test before you trust a rule
- Search your last three statements for the text you plan to use and note how many rows match.
- Write the rule and apply it to one month only, if your tool allows it.
- Open the matched rows and look for any that do not belong.
- If all are correct, extend the rule to the other months.
- Write down the rule in a note so you remember why it exists.
The extra minutes matter more for gig workers than for most, because ride, fuel and food payees overlap and a single loose rule can quietly move thousands of rupees into the wrong category.
A sensible order for rules
Start with your five most repeated payees, write specific rules, and look at the month's rows afterwards. Add a rule only when a payee has appeared at least twice, and review your rules every few months to delete any that no longer fit.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

