This helps most during the onboarding phase.
The set of “auto-categories” is great for helping a new user to start getting their arms wrapped around the volume of data. However:
  1. Transactions are not always categorized properly, sometimes entirely.
  2. In addition it appears you can only “re-run categorization”
    once
    to get a modal showing
    just
    the transactions for a newly created category; if you run it a second time you are shown
    all
    transactions left to categorize.
  3. When the modal pops up it does not contain information about what the category is
    meant
    for, i.e. the description that helps a user statelessly determine what is meant for this category (“Does this insurance category include pet insurance? What did the description say again?”).
For me this results in this process:
  1. Open three separate tabs: Categories, Auto-Categorization Settings, a second Auto-Categorization Settings.
  2. In one of the Auto-Categorization Settings tabs, select a single “missing” category to create.
  3. Click “re-run categorization”.
  4. In the
    second
    Auto-Categorization Settings consult the description to make accept/reject decisions in the
    first
    Auto-Categorization Settings tab.
  5. If none of the transactions match: go to the Categories tab and manually delete the newly created category, as it’s not applicable to my set of transactions.
I find myself wincing a bit every time an auto-category turns out to be inapplicable to my scenario as I’ve got to go make a context switch, find the category in the Categories tab, and click extra buttons to clean it up.
Being able to preview what LM thinks belongs in the category before creating it would help a
lot
.
For extra points, being able to tell LM “none of those transactions fit” and clear it from “missing” categories while deferring the context switch for manually categorizing would be great.