You keep recategorising the same merchant because each import starts from a guess, and your correction does not tell the tracker what that merchant is. Write a rule once and the guess stops. Below is a checklist, with figures in Canadian dollars, for deciding which merchants to fix, how to write the rule and how to check it worked. It describes method only, with no banking or tax specifics.
Before and after, in dollars
Say a hardware store sells you both cleaning supplies and garden tools, and your tracker alternates between Household and Garden. In four months you spend C$62.40, C$118.00, C$45.75 and C$89.90. The total is C$62.40 + C$118.00 + C$45.75 + C$89.90 = C$316.05. Because the labels wobbled, Household shows C$108.15 and Garden shows C$207.90. A single decision (everything from this store is Home and garden) puts the full C$316.05 on one line.
| View | Household | Garden | Home and garden |
|---|---|---|---|
| Before the rule | C$108.15 | C$207.90 | none |
| After the rule | none | none | C$316.05 |
The checklist
- Export or open last month's list and sort by merchant.
- Mark every merchant where you changed the category more than once in three months.
- For each, write down the text that stays constant across lines.
- Decide whether the merchant is always one category, or mixed.
- For always-one merchants, write a rule. For mixed ones, set a default or leave them to manual review.
- Re-run the rules on the past three months.
- Look at the category totals and check each against your memory of the month.
How do I know if a merchant is mixed?
Look at the amounts. A pharmacy chain that sells both prescriptions and cosmetics will show small and large amounts with no pattern. A single-purpose payee like a gym or a phone provider has the same amount each month. If amounts are the same, a rule is nearly always safe. If amounts vary wildly, be careful.
- Same amount each month: rule.
- Varying amount, one purpose: rule.
- Varying amount, varying purpose: default plus review.
What if the statement text keeps changing?
Processors sometimes rename a payee, or a business is bought by another and its descriptor changes. When that happens, an old rule stops matching and the repeat fixes return. Keep an eye out for a merchant that suddenly shows up uncategorised, then add the new spelling to the rule or create an alias.
Where tools differ
Some apps learn from every correction, others only follow explicit rules, and some combine the two. SpendixAI applies your own rules first and learns from your history after that, with an AI step only for what is left and only if you switch AI on. It is built for India and INR, so this article is not a claim that it is sold in Canada. The method above works wherever your tool supports rules.
A worked check of the time saved
Say you re-label 12 rows a month and each takes about 20 seconds. That is 240 seconds, or four minutes, every month, which is 48 minutes over a year. Writing six rules at three minutes each costs 18 minutes once. After that the same four minutes simply stop appearing, and the saving continues for as long as the merchants stay in your routine.
The time is the smaller gain. The larger one is trust: when you stop correcting by hand, your category totals stop depending on whether you remembered to correct them.
Closing check
- Do your three biggest categories match your sense of the month?
- Is there any merchant still uncategorised at month end?
- Do you know where your rules live, so you can edit them?
Tick all three and the repeat work is largely gone.
Useful ideas start with a clearer picture of your own spending.
Explore your reports

