How do I correctly to interpret the predictive routing score when integrating it into complex Architect decision trees?
We are currently observing significant variances in the predictive routing scores assigned to agents within our primary support queue. The environment is Genesys Cloud EU1, and we are utilizing the standard predictive routing model without custom attribute weighting at this stage. The issue manifests when agents with high historical performance metrics are consistently bypassed in favor of agents with lower historical handle times but higher availability. This contradicts the expected behavior of the algorithm, which should prioritize skill match and historical success rates.
In the Performance dashboard, the ‘Agent Performance’ view shows these bypassed agents with superior quality scores and lower abandonment rates. However, the ‘Queue Activity’ view indicates that the predictive routing engine is assigning lower priority scores to them during peak hours. We have verified that the skill profiles are correctly mapped in the Architect flow, and the ‘Set Attributes’ tasks are functioning as expected. There are no error logs in the conversation detail views, suggesting the issue is not a configuration error but rather a misunderstanding of the scoring logic.
We require clarity on how the predictive routing algorithm balances historical performance against real-time availability. Specifically, does the system penalize agents who are currently in a ‘Not Ready’ state, even if their historical metrics are superior? Additionally, how does the algorithm handle agents who have recently completed long interactions? We suspect that the ‘cooldown’ period might be influencing the scores, but the documentation does not provide specific details on this behavior.
Any insights into the internal weighting mechanisms or configuration adjustments that could align the routing behavior with our performance metrics would be greatly appreciated. We aim to optimize agent utilization and ensure that the most qualified agents are handling the most complex interactions.