Custom questions

Custom Questions turn your own quality or compliance rules into structured results on every new call. Each enabled question is evaluated against the transcript and can return:

• a Pass or Fail verdict;
• supporting transcript evidence;
• a detected name when relevant;
• optional structured details specific to the question.

The feature is enabled for your account by Speechlytics, and questions are configured per account. Sub-accounts receive their parent's results.

Write an effective question

A question is a scoring instruction, not just a short label. State the behavior, who must perform it, and the exact pass and fail conditions.

Salesperson Name in Introduction
- Goal: Did the salesperson state their name in the opening greeting?
- Initiator: Salesperson only.
- Pass: Name stated during the opening greeting.
- Fail: Name appears only later in the call or during sign-off.

For consistent results:

• check one behavior per question;
• identify whether the agent, customer, or either speaker must perform it;
• describe edge cases explicitly;
• ask only about information that can be heard in the call;
• avoid rules that depend on unseen CRM or ticket data.

Disabling a question keeps its history but stops it from being applied to future calls. Editing a question does not re-grade calls that were already processed.

Import and export

Questions can be exchanged as a semicolon-separated CSV without a header:

Question text;true;
Another question;false;

The final value is the enabled flag. Accepted values include true/false, yes/no, y/n, and 1/0. Import matches existing questions by normalized text and updates their enabled state; changed wording creates a new question.

Read call results

Results are returned in customQuestions, with the overall pass rate and each question's verdict, evidence, detected name, extra details and scoring rule, from:

• GET /api/v1/calls/{id}/status;
• GET /api/v1/calls.

A missing customQuestions value means the call was not evaluated. It is intentionally different from an evaluated call where every question failed.

Aggregating results

When you aggregate pass and fail rates across calls, count only calls that contain Custom Question results; calls that were not evaluated are not failures.

New or edited questions affect future processing. Reprocess historical calls if they must be evaluated against a changed rule set.

Troubleshooting

Symptom	Likely explanation
Custom Questions is not visible	The capability is not available for the account.
A call has no results	No enabled questions existed when it was processed, or grading did not complete.
A new question has no history	Questions do not retroactively grade older calls.
Import created a second question	Its wording no longer matched the existing normalized text.
A verdict looks wrong	Review the evidence and transcription, then make the rule more explicit.