Learn why you should simulate results before applying a scorecard for AutoQA evaluation.
This article supports Super User and Supervisor roles only.
As PeopleQX leverages AI to evaluate conversations, your scorecard needs to be crafted in certain ways to drive the accuracy of AI scoring. Testing is a way to build confidence in writing effective questions.
Here’s how to run a test:
After adding a question, the test button will be activated. Test one question at a time, or send the entire scorecard for testing.
Samples are required for testing. Conversations from the assigned sub-unit will be pre-selected as samples. However, samples can be managed in a way that better suits your needs.
Test results will display as ‘Y’ or ‘N’, or ‘N/A’ and you can click on the bottom arrow icon to display the scoring indicators.
There are two scenarios why a result is marked ‘N/A’:
- question was not selected for AutoQA
- in a question logic setup, the evaluation was ended or the follow-up question was skipped.
You can use the information on the right to check for the possible reason.
Apply for evaluation
If you’re satisfied with the test results, you can proceed to have AutoQA evaluate conversations and score agents. Select the action from the upper right corner of the screen.
You will be prompted for a confirmation and upon confirming, this scorecard will be queued for AutoQA evaluation to start.
Upon application, you’ll be redirected to the list of scorecards with a notification confirming the scorecard is en route for AutoQA evaluation.
⭐Although testing is not mandatory, you’re highly encouraged to try it out to ensure the scorecard’s quality is as good as it can get.