Core workflows

Categorization

How a category is chosen, what similarity means, and when learning happens.

Decision order

  1. 1

    Rules

    Exact / fuzzy pattern

  2. 2

    History

    Past patterns

  3. 3

    Other

    Needs review

TrackIt does not call an LLM per CSV row. Categories come from deterministic mappings and similarity against your history. Full labels: Categories.

In the UI

  • Category chip opens a dropdown with search, up to three suggestions (similarity / confidence / learned count), and the full list.
  • Needs review can Apply top suggestions (up to 25 rows, similarity floor about 0.55).
  • Import confirm can create or strengthen a merchant mapping when you override.
Suggestion rows
Dropdown-option layout: category dot, name, and similarity meta.

Learning side-effect

Bulk review and Rules Apply to Other are different tools for the same goal: fewer Other rows. See Needs review.