Overview
New custom QA scores are often lower than expected when first published. While some of the low scores are caused by true misses, sometimes the issue is as simple as needing to update a category to pick up on a missing phrase. In other cases, requesting a score override or score change is the appropriate solution.
Common reasons and resolutions for low QA scores:
-
- True miss: Coach agents to improve the targeted behavior.
- Category miss: Train the category to recognize the right phrase.
- Exceptions: Utilize the score override feature to adjust the score.
- Score logic issue: Change the score's structure or logic.
True miss
True misses occur when the agent did not demonstrate the targeted behavior. Here is an example of a true miss in the QA Score details:
And in the interaction transcript:
In this example, we can see in the interaction transcript that the agent did not complete the appropriate steps to verify the customer's identity.
Category miss
Category misses occur when a category should have picked up a phrase in the transcript, but did not. Here is an example of a category miss in the QA score details:
And in the interaction transcript:
In this example, the Issue Resolution category did not hit on the phrase "I feel really good about that I'm glad I called" when it should have.
In this example, here is what a relevant phrase capture would look like:
Exceptions
Exceptions occur when there is a scenario in which the QA process is not applicable, but the situation is too rare to define and add to the score logic exclusions. Here is an example of an exception in the QA score details:
And in the interaction transcript:
In this example, the agent did not need to verify the customer's identity and contact information because the customer already provided it.
Score logic issue
In some cases, the score's logic can be changed to ensure that the QA score is only being measured on interactions that need the full QA process.
Here is an example of a QA score that failed based on the score criteria, but the outbound interaction was only 10 seconds long because the agent was not able to reach a customer:
If you are still experiencing issues with your QA score, please reach out to your CSM.