Predictive routing reference

This article collects detailed information on how predictive routing works in specific scenarios. 

Understanding the basis for agent rankings

Predictive routing scores the agents who might handle an interaction using a machine learning model. Machine learning is effective at identifying patterns. In this case, patterns identify the agents who deal most effectively with certain types of interactions. 

Note: Predictive routing currently supports voice interactions only.

When predictive routing is activated on a queue, it does the following:

  1. Creates a model using various data sources, including agent profile data, aggregated customer data (such as whether they are a repeat caller), and historic interaction data.
  2. When an interaction arrives in that queue, predictive routing retrieves data about the customer and each available agent. It uses the model to process agent and customer data in real time and returns a ranking for each available agent.
  3. This ranking represents the agents predictive routing expects to have the most positive impact on the target KPI when handling that specific interaction.

How Genesys Cloud assigns agents using predictive routing

When a voice interaction is offered to the queue, predictive routing does the following:

  1. Creates a list of all agents assigned to the queue. 
  2. Filters the list to include only agents actually working (On Queue) and who handle voice interactions. Predictive routing supports voice interactions only.
  3. Filters the list for required agent language skills. Predictive routing does not filter the list of agents for any required skills other than Language. Queues that must enforce skill requirements might not be suitable for predictive routing.
  4. Scores all agents in the list and ranks them in order. The higher the score, the higher the predicted impact on the target KPI when handling the current interaction. The highest scored agent ranks first.
  5. During the predictive routing timeout period (configured in the Queue details window), predictive routing tries to assign the interaction to the highest ranked agent available.
  6. If Genesys Cloud finds no qualifying agent during the timeout period, it routes the interaction using standard routing, which is the fallback routing method.

Predictive routing and data retention

For predictive routing to be at its most effective, you must retain at least 180 days of data. If you retain fewer days of data, you can still use and benefit from predictive routing. However, the quality and effectiveness of your AI models and predictions is likely to suffer and the resulting benefit will be reduced compared to standard routing.

The KPI processing phase 

The KPI you choose to optimize is one of the most significant parameters used to configure predictive routing. When you do any of the following actions, Genesys Cloud enters a processing phase, during which it analyzes your data and constructs the model used to optimize the KPI:

  • Start a comparison test
  • Activate predictive routing for all interactions on a queue
  • Change your KPI after you activate predictive routing

KPI processing can take several hours.

  • If you change your KPI after processing starts, Genesys Cloud deletes your model and starts over, even if you change your KPI to one that you had been using in the past.
  • If you run a comparison test and then activate predictive routing for all interactions on the queue without changing the KPI, no processing phase happens. The change is immediate.
  • If you hover your cursor over the Predictive [processing…] label, a pop-up shows the routing method used during the processing period.

While the KPI processing happens, Genesys Cloud routes interactions on the affected queue in the ways described below:

  • For benefit assessment – Benefit assessment has no effect on routing, so Genesys Cloud routes interactions using your current routing method.
  • When activating a predictive routing comparison test – Genesys Cloud routes all interactions using the routing method you configured for comparison with predictive routing. During this phase, the Routing column on the Queue list window shows Predictive [processing…].
  • When activating predictive routing – Genesys Cloud routes all interactions using standard routing. During this phase, the Routing column on the Queues list window shows Predictive [processing…].
  • When you change the predictive routing KPI – If you had set up a comparison test, Genesys Cloud routes all interactions using the routing method you configured for comparison with predictive routing. If predictive routing was routing all interactions, Genesys Cloud routes them using standard routing. During this phase, the Routing column on the Queues list window shows Predictive [processing…].
    • Note: Changing the KPI is resource-intensive and can have significant consequences for your routing. For example, improvements using predictive routing for one KPI on a queue do not indicate that another KPI will also improve. Genesys strongly recommends that you always return to standard or bullseye routing and then run a new benefit assessment evaluation on the queue before changing your KPI configuration.
  • When predictive routing fails to activate correctly – If you are trying to start a comparison test, Genesys Cloud routes all interactions using the comparison method you configured. If you are trying to activate predictive routing for all interactions, Genesys Cloud routes all interactions using standard routing. When an error occurs during activation, the Routing column on the Queues list window shows the currently used routing method (standard or bullseye) with a red exclamation mark icon next to it.

Supported KPIs

The following table lists the KPIs currently available for use in predictive routing. If a KPI you want to use is listed as Beta, contact your Genesys representative for assistance.

  • To submit a request for a new KPI, add it in the Genesys Cloud Product Ideas Lab, available to logged-in users from the Genesys Knowledge Network.
Generally Available Beta
Average Handle Time (AHT) Transfer %

Routing fallback scenarios

Predictive routing mode Scenario Routing method used
Predictive routing activated, standard routing as fallback success predictive routing
  fallback – timeout standard
  failure – error standard
Comparison test mode, predictive routing vs standard routing predictive routing on, success predictive routing
  predictive routing on, fallback – timeout standard
  predictive routing on, failure – error standard
  predictive routing off standard
Comparison test mode, predictive routing vs bullseye routing predictive routing on, success predictive routing
  predictive routing on, fallback – timeout standard
  predictive routing on, failure – error bullseye
  predictive routing off bullseye