Getting started

Does correcting a category once fix future imports?

Does correcting a category once fix future imports? Corrections become history that helps match similar merchants, but a rule is the firmest fix.

Editorial illustration for Does correcting a category once fix future imports?

Correcting a category once usually helps future imports, because your correction becomes part of your history, which is used to categorise similar merchants later. It will not catch every row, though. If you want a merchant to land in the right category every time, a rule is the firmest fix, since rules are applied before history and before any AI.

What happens when I correct a category?

You change a transaction from, say, Shopping to Groceries. Two things follow. That row now carries the category you chose. And the correction is remembered as history for that merchant. When the next statement arrives with the same or a similar merchant, the tracker has your earlier decision to draw on.

In SpendixAI the order of categorising is: your own rules, then your history and similar merchants, then an AI model for what is left if AI is on. Corrections feed the second step.

Why is a rule still firmer than a correction?

History is a strong hint, and a hint can be outweighed. A merchant you categorised two different ways in the past, or a similar-looking name that was categorised differently, can pull the answer the wrong way. A rule is an instruction: when this text appears, use this category. Because rules run first, they win over history. If a merchant should always be in one category, such as your monthly broadband or the milk delivery, write the rule.

MethodHow firmBest for
Correct a category by handBecomes history, a good hintOccasional merchants
Add a merchant aliasJoins spellings into one nameMerchants printed several ways
Write a ruleApplied first, firmestRecurring merchants that always belong in one place

A before and after: month one versus month two

Say your first import in October has 150 rows and the tracker places 90 of them from the default categories and your rules, leaving 60 for you. You correct 40 of them in one sitting and write 6 rules for the merchants that recur, such as your grocer, the milk supplier and fuel. Those 40 corrections are now history.

In the November import of 155 rows, the 6 rules place the repeat merchants outright. History places many of the others, because they are similar to something you fixed. Instead of 60 rows to deal with, maybe 25 are left, and most of those are shops you have not visited before. The numbers are invented, but the shape is typical: each month the queue shrinks, then levels off at the genuinely new.

What will a correction not fix?

  • Rows that were already imported: an old transaction keeps its category until you change it.
  • A merchant with several spellings that the tracker does not know are one shop. Add an alias.
  • A merchant that really belongs in different categories on different days.
  • Anything held for review because AI is off and no rule or history applies yet.

How do I make my corrections count?

  1. Correct the repeat merchants first, since they pay off most.
  2. Be consistent: pick one category per merchant and stick with it.
  3. Write a rule for any merchant you correct a third time.
  4. Merge look-alike spellings with an alias.
  5. After the next import, check which rows the tracker got right without your help.

A few weeks of consistent corrections build up a history that does more than any settings screen.

SPENDIXAI JOURNAL

Useful ideas start with a clearer picture of your own spending.

Explore your reports