Predictive Model for Contact Center Agent Performance Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In contact centers, existing performance measurement metrics often fail to accurately assess agent effectiveness, particularly in scenarios where transactions or purchases are not made, leading to potential customer dissatisfaction and unfair comparisons across different campaigns or demographics.
Innovation Solution
A predictive model is employed to determine agent performance by calculating the cumulative expected likelihood of success for each call, allowing for comparison with actual outcomes to identify agents performing above or below expectations, independent of campaign type and customer demographics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If call duration is used as the performance metric, then it is easy to measure and process, but it does not accurately measure agent effectiveness in resolving customer issues
Solution Approach 1:
The patent introduces a predictive model as an intermediary that processes multiple call parameters (duration, outcome, customer demographics, campaign type) to generate a normalized performance score. This mediator transforms easily measurable raw data into an accurate effectiveness metric that accounts for external factors beyond agent control.
Solution Approach 2:
The patent changes the performance measurement parameter from simple call duration to a composite effectiveness score derived from multiple parameters including call outcome, duration, customer demographics, and campaign type. This parameter transformation enables accurate measurement of agent effectiveness while maintaining ease of data collection.
2Productivity
If agents are evaluated based on absolute performance metrics without normalization, then comparisons are simple, but agents working on different campaigns or with different customer demographics are unfairly compared
Solution Approach 1:
The patent applies equipotentiality by normalizing agent performance scores to account for external factors such as campaign type and customer demographics. This creates equal measurement conditions for all agents, ensuring that differences in performance are due to agent skill rather than unequal starting conditions or external variables.
3Productivity
If agents focus on maximizing measured performance parameters like call volume, then productivity increases, but customer satisfaction may decrease due to rushed or insufficient call resolutions
Solution Approach 1:
The patent implements feedback by using call outcomes and customer satisfaction data as inputs to the predictive model. This feedback loop ensures that performance measurement reflects both productivity and quality, preventing agents from optimizing solely for call volume at the expense of customer satisfaction.
Data Source
AI summary
An agent's performance is measured using a predictive model calculating an expected probability of success for each outbound communication, such as a call, handled by the agent that has reached the desired (“right”) party. An agent's cumulative actual performance value is maintained based on each “successful” contact, as indicated by a disposition code provided by the agent. A cumulative expected probability value of a “successful” communication is also maintained based on each call that reaches the right party where the expected probability of a “successful” call is determined by the predictive model. The agent's performance value can be determined by comparing the cumulative actual performance value with the cumulative expected probability of success value. The agent's performance value can then be compared to the performance value of other agents to identify agents performing above-expectations or below-expectations.