Genesys Cloud - Main

Good afternoon, I am attempting to drum up support for an idea: Agent Empathy currently populates on 25% | Genesys Cloud Ideas Portal

My supporting information on the why and potentials for how...

Issue Summary: Empathy scoring currently populates on only 23–30% of interactions due to no associated phrasing, resulting in a division by zero. These interactions are excluded from scoring, leading to an inflated empathy score for agents who have minimal but positive interactions.

While I do not have the answer, I would like to suggest review of the "request" below as a potential solution to the skewed aggregate or overall score at agent level.

Request:

Review and modify the empathy scoring logic to:

1. Include all interactions in the scoring denominator, even those with 0 empathetic and 0 unhelpful phrases.

2. Prevent division by zero errors by implementing a fallback or default logic (e.g., score of 0 when both counts are zero).

Below is my original text in full…I wanted to draw out a few points a get the conversation started around how this can be solutioned.

Care Ticket Request: Empathy Scoring Logic and Reporting Inconsistencies

Issue Summary: Empathy scoring currently populates on only 23–30% of interactions due to no associated phrasing, resulting in a division by zero. These interactions are excluded from scoring, leading to an inflated empathy score for agents who have minimal but positive interactions.

Observed Behavior:

1. Most of our calls have no score.

2. Interactions with true zero values (0 empathetic, 0 unhelpful) are excluded from scoring.

3. This exclusion skews the average empathy greatly.

4. Agents with very few positively scored interactions can appear to have a perfect empathy score of 100.

5. Agents with very few negatively scored interaction can appear to have an overall of (-100).

6. When null (unscored) interactions are treated as zero, the average empathy score aligns more closely with sentiment scores, suggesting a more accurate reflection of agent behavior.

Impact:

1. The current scoring model does not provide a reliable indicator of agent behavior.

2. It creates a misleading representation of empathy performance.

3. It limits the ability to use empathy scores for coaching, quality assurance, or performance evaluation.

Request:

Review and modify the empathy scoring logic to:

1. Include all interactions in the scoring denominator, even those with 0 empathetic and 0 unhelpful phrases.

2. Prevent division by zero errors by implementing a fallback or default logic (e.g., score of 0 when both counts are zero).

Supporting Data:

1. Average empathy score (current logic): mid-60s

2. Average sentiment/sentiment trend score: 26

3. Average empathy score (if nulls = 0): aligns with sentiment/sentiment trend

Thank you, GC Reporting and Analytics, for your input. 

#Calculations
#Metrics
#PerformanceExports

------------------------------
Glenda Kingston
------------------------------

Update: An additional idea has been submitted around this post: https://genesyscloud.ideas.aha.io/ideas/DARSTA-I-380

------------------------------
Glenda Kingston
------------------------------

Original Message:
Sent: 06-17-2025 17:08
From: Glenda Kingston
Subject: Agent Empathy Scores

Good afternoon, I am attempting to drum up support for an idea: Agent Empathy currently populates on 25% | Genesys Cloud Ideas Portal

My supporting information on the why and potentials for how...

Issue Summary: Empathy scoring currently populates on only 23–30% of interactions due to no associated phrasing, resulting in a division by zero. These interactions are excluded from scoring, leading to an inflated empathy score for agents who have minimal but positive interactions.

While I do not have the answer, I would like to suggest review of the "request" below as a potential solution to the skewed aggregate or overall score at agent level.

Request:

Review and modify the empathy scoring logic to:

1. Include all interactions in the scoring denominator, even those with 0 empathetic and 0 unhelpful phrases.

2. Prevent division by zero errors by implementing a fallback or default logic (e.g., score of 0 when both counts are zero).

Below is my original text in full…I wanted to draw out a few points a get the conversation started around how this can be solutioned.

Care Ticket Request: Empathy Scoring Logic and Reporting Inconsistencies

Issue Summary: Empathy scoring currently populates on only 23–30% of interactions due to no associated phrasing, resulting in a division by zero. These interactions are excluded from scoring, leading to an inflated empathy score for agents who have minimal but positive interactions.

Observed Behavior:

1. Most of our calls have no score.

2. Interactions with true zero values (0 empathetic, 0 unhelpful) are excluded from scoring.

3. This exclusion skews the average empathy greatly.

4. Agents with very few positively scored interactions can appear to have a perfect empathy score of 100.

5. Agents with very few negatively scored interaction can appear to have an overall of (-100).

6. When null (unscored) interactions are treated as zero, the average empathy score aligns more closely with sentiment scores, suggesting a more accurate reflection of agent behavior.

Impact:

1. The current scoring model does not provide a reliable indicator of agent behavior.

2. It creates a misleading representation of empathy performance.

3. It limits the ability to use empathy scores for coaching, quality assurance, or performance evaluation.

Request:

Review and modify the empathy scoring logic to:

1. Include all interactions in the scoring denominator, even those with 0 empathetic and 0 unhelpful phrases.

2. Prevent division by zero errors by implementing a fallback or default logic (e.g., score of 0 when both counts are zero).

Supporting Data:

1. Average empathy score (current logic): mid-60s

2. Average sentiment/sentiment trend score: 26

3. Average empathy score (if nulls = 0): aligns with sentiment/sentiment trend

Thank you, GC Reporting and Analytics, for your input. 

#Calculations
#Metrics
#PerformanceExports

------------------------------
Glenda Kingston
------------------------------