Extraction criteria
The structured data fields Violet fills in from each call transcript - one per screening question.
After every call, Violet reads back through the transcript and extracts structured answers into your extraction criteria. Think of them as columns in a spreadsheet - each criterion is a field that gets filled in for every contact who spoke with Violet.
How criteria relate to questions
There's usually one extraction criterion per in-call question. If you ask "Are you currently looking for new nursing roles?", you'd create a boolean criterion called something like Actively looking. Violet will mark it Yes or No based on what the candidate said.
Not sure what criteria to create? Use the Generate with AI button in the campaign editor - it reads your in-call questions and drafts a matching set of criteria for you to review and adjust.
See In-call questions for guidance on writing questions that extract cleanly.
Criterion types
| Type | Use it when you want... | Example |
|---|---|---|
| Boolean | A yes/no answer | Holds AHPRA registration |
| Number | A numeric value | Years of experience |
| Selector | One of a fixed set of options | Shift preference: Days / Nights / Rotating |
| Text | A short free-text answer | Preferred work location |
Tips for writing good criteria
- Match the question closely. If your question is "What's your earliest available start date?", name the criterion Earliest start date and set the type to Text (or Date if that's available for your plan).
- Keep names short. The criterion name appears as a column header in your results table, so Available July is better than Is the candidate available to start in July.
- Stick to what you asked. Don't add criteria for information Violet never asked about - the extraction will be unreliable.
Reviewing extracted data
Once calls start coming in, you can see each contact's extracted values in the campaign results view. You can filter, sort, and export from there to push shortlisted candidates into your next step.