What classification does
Classification determines what each uploaded file represents and which candidate it belongs to. For comparison workflows, multiple files can form one candidate submission. For validation, a file is mapped to an expected document type.
Review the proposed structure
- Open the Classification or Candidates step.
- Start with items marked uncertain, unassigned, or conflicting.
- Open the document preview and confirm its identity from the content, not only the file name.
- Move the file to the correct candidate or expected type.
- Create, rename, merge, or split candidate groups only when the source documents support that boundary.
- Confirm the structure before starting analysis.
Common examples
- A cover letter and proposal PDF from the same supplier belong to one candidate.
- Two resumes with similar names remain separate candidates unless they are clearly versions of the same submission.
- A single combined PDF may contain multiple logical document types; review detected boundaries before accepting the classification.
- An unexpected supporting file can remain attached to the candidate even when it does not map to a required type.
Why this matters
Incorrect grouping can make evidence appear missing, assign it to the wrong candidate, or distort consistency checks. Classification review is the last inexpensive place to correct the evidence scope before AI evaluation uses credits.