Skip to main content

​
What is Data Collection?

Data Collection turns each call into a tidy record of the business facts you care about — a budget, a preferred appointment day, whether the caller already has a policy. You define the fields once on the script; after every call ends, the AI reads the transcript and fills in a value for each field. The values are stored with the call and can be read on the call, exported, sent on by an automation, or fetched through the API. It is configured per script, in its own section of the Post-Call Analysis tab — “Extract structured business data for reporting, exports, and automations.”: Call Scripts → (your script) → Post-Call Analysis → Data Collection Data Collection works after the call, from the transcript. It does not change what the assistant says or asks during the call. To actually ask a caller for a value and confirm it live, use a collection tool in your flow — for example Data Collection - v0.2 — and add a field here if you also want the answer in the call’s structured record.
Data Collection and Custom Analysis Fields both answer questions from the transcript. Use custom analysis fields for judgements about the call (was the caller qualified, did they mention a competitor). Use Data Collection for the business data the call was meant to capture.

​
Adding fields

Before you add anything the section reads No data collection fields yet. Add your first field below. Click Add Field to append a field. It starts as a Text field with the key field_1 (then field_2, and so on) and the label Field 1, and opens ready to edit. Click a field’s row to collapse or expand it, and use its delete icon to remove it.

​
Data types

For Single Select and Multi Select fields, the Options editor lists the allowed answers. Each option has a Value, a Label and an optional Description. Add Option appends a row. The AI reads only the field’s own Description when it extracts a value, so also list the allowed values there (and, for Number fields, any range or unit). Text, Number and Yes/No fields are stored as their own type. Other types are stored as text, so write the Description to say the exact format you want, such as “the date as YYYY-MM-DD”.

​
Example

A field that captures the day a caller would like a visit:
After a call where the caller said “Thursday works best”, the call’s record holds preferred_day with the value thursday (string).

​
Where the values appear

  • The call’s Data tab. Open a call from the Call Log and switch to its Data tab to see every field with its extracted value and a count such as 4 of 6 fields collected — see Call details.
  • The Call Report PDF. The data collection table is included when you export a call as a PDF.
  • Automations. A Call Completed trigger loads the values as call.structured_data, so a condition or field mapping can read a single field’s value — for example call.structured_data.fields.preferred_day.value (string, e.g. thursday). See Automation conditions and field mappings.
  • The API. The call result’s structured_data object holds the same values in its fields object, keyed by Field Key, each with a value — see the API quickstart.
Values appear once the call has finished and been analysed, usually within a few seconds of the call ending. Changes to the fields take effect on calls analysed after you save the script.

​
Writing fields that extract well

  • One fact per field — “budget and timeline” in one field gets one muddled answer. Split it.
  • Say what “not mentioned” looks like — tell the AI to leave the field empty when the caller never said it, rather than guessing.
  • Pin the format — for dates, phone numbers and codes, spell out the format you want in the Description.
  • Keep keys stable — automations, exports and API integrations refer to the Field Key. Rename the Display Label freely, but treat the key as permanent once something depends on it.
  • Keep option lists complete — give Single Select fields an option for every answer a caller could give, including one such as not_discussed, and repeat the list in the Description.

​
Next Steps