Testing and quality
Call analysis
Know the outcome without replaying the call. Get a summary, quality score and the details you asked for.
#What every call gets
| Field | What it is |
|---|---|
| Summary | A few sentences on what the caller wanted and what happened. |
| Outcome | The one-word result. See Outcomes and end reasons. |
| Intent | Why they called, in a few words. |
| Caller name | Picked up from the conversation and saved to the contact. |
| Follow-ups | Tasks someone should do, added to your follow-up list. |
| Quality score | 0 to 100 for how well the assistant handled it, with any issues noted. |
#Customise it
On the assistant's Analysis tab:
summary_prompttext- How to write the summary, such as "Start with the caller's name and whether they are a new patient."
success_rubrictext- How to judge whether the call went well.
success_scalescore_10 | pass_fail | descriptivedefault score_10- A score from 1 to 10, pass or fail, or a word from excellent to poor.
structuredup to 40 fields- Fields to extract from every call, each with a name, a type (
string,number,boolean,enumordate), a description and, for enums, the options.
Structured fields
[
{ "name": "budget", "type": "number", "description": "The buyer's maximum budget in dollars" },
{ "name": "timeline", "type": "enum", "options": ["now", "3 months", "6 months", "later"],
"description": "When they want to move" },
{ "name": "pre_approved", "type": "boolean", "description": "Has a mortgage pre-approval" }
]#Where results appear
On the call page's Analysis tab, in GET /calls/{id} and in the call.completed webhook under analysis: success and success_reason from the rubric, scale, and your fields under data.