Predictive Routing Block

From Genesys Documentation
Jump to: navigation, search
This topic is part of the manual Designer User's Guide for version Current of Designer.

If your site is enabled for Genesys Predictive Routing, you can use the Predictive Routing block to route an interaction to the agent with the highest likelihood of a successful outcome.

Related documentation:
Important
You can only use this block if your environment is enabled for Genesys Predictive Routing. Contact your Genesys representative if you have any questions about using this feature.

The Predictive Routing block enables a Designer application to retrieve a list of agents that are best equipped to handle a specific type of call and then route the call to the agent with the highest likelihood of a successful outcome.

Typically, you would add this block when you want to improve routing metrics for an application that is using traditional (or "prescriptive") routing methods and has an established amount of historical data that can be used for generating accurate predictors and models.

How it works

Genesys Predictive Routing (GPR) uses machine learning to analyze your accumulated history of agent, customer, and call data to generate scores for each available agent that indicate probable outcomes for incoming calls. Using these scores, you can ensure that calls are routed to agents with the highest likelihood of producing a satisfactory result.

When you add the Predictive Routing block to an application, you can select a Predictor from the list of Predictors that Designer retrieves from GPR. The Predictor indicates the metric you want to optimize (such as first-contact resolution, sales conversion, and so on). The Predictor also specifies which customer and agent qualities (or features) have the most impact on the chosen metric.

When a call arrives, Designer retrieves a list of currently-available agents (ranked according to the parameters specified in the Predictor), and then routes the call to the best available match. Designer continues to periodically retrieve these lists of agents for the duration of the session—even if the application has moved on from the Predictive Routing block to the next routing block—until the call is finally routed or the session ends. If the "predicted" outcome differs from the actual outcome, Predictive Routing "learns" from this result, improving the future accuracy of its predictions.

For more information about setting up Predictors, see Creating and updating Predictors in the Genesys Predictive Routing Help.

Important
Only Predictors with active Models are available for selection. See Configuring, training, and testing models in the Genesys Predictive Routing Help for more information.

Using this block

This block can be placed in the Assisted Service phase of an application. For the block to take effect, you'll also need to enable predictive routing in the application settings.

Tip
Genesys recommends that you use this block in a mixed-use setting, where Predictive Routing is attempted first (say, for a period of 2 seconds) and then traditional routing acts as a fallback method of routing the call. This also helps to train the models used by Predictive Routing, which can effectively "learn" from the traditional routing outcomes.

Call Routing tab

In the Targeting section, select the Predictor you want Designer to use. You can also choose to specify this value using a variable.

In the Evaluate field, enter the number of agents to submit as a batch for routing, and how often to submit each batch. For example, you might choose to submit a batch of 50 agents every 10 seconds.

You can then specify additional criteria to be used when evaluating possible agents, such as their state or skills, and select the virtual queue for routing.

In the Other Routing Settings section, you can specify which routing algorithm to use and the block timeout value.

Des predictive routing ex01.png

Skill Expression tab

Use this tab to specify a skill expression for the Predictive Routing block. The skill expression consists of a list of skills for which you must individually set an operator and an integer value.

Important
When you configure a skill expression in the Skill Expression tab of the Predictive Routing block, the Consider Only option (for agent status) is enabled on the Call Routing tab. Also, the built-in agent status filter is ignored if you use a variable-defined skill expression.

For more information about setting up skill expressions, see the Skill Expression settings for the Route Call block.

Routing Priority tab

For information about these settings, see the Routing Priority settings for the Route Call block.

Advanced tab

Greetings section

Enable the check box beside Customer Greeting and/or Agent Greeting to play an audio file to that person when the interaction is connected.

For more information, see the Advanced tab settings for the Route Call block.

Extensions section

Click Add Extension Data to add an extension as a key-value pair to this block. The value type can be a string or integer.

For more information, see the Advanced tab settings for the Route Call block.

Predictive Settings section

If desired, you can specify key-value pairs that will provide additional context to the prediction service.

Tip
The Predictive Routing block does not have an option for treatments, but you must specify a treatment to play during routing. Genesys recommends that you do the following:
  • Place a Start Treatment block and a Self Service module ahead of this block to provide a treatment to play during routing.
  • In the settings for the Start Treatment block, disable the Is Synced option so that Designer will start the treatment and then immediately move on to the next block while the treatment keeps playing. This ensures that the same treatment continues for any prescriptive routing blocks that follow.

Example

Let's say we set up our Predictive Routing block with the following settings:

On the Call Routing tab, we select a Predictor service that will query the scoring engine for available agents. We then tell the block to submit batches of the 50 "best" agents to the routing engine every 10 seconds. (For the purposes of this example, we can leave the other settings as defaults.)

Des predictive routing ex01.png

On the Routing Priority tab, we set the Initial Priority to 50, with a 5 point increment every 15 seconds:

Des predictive routing ex02.png

With these settings, the block sends periodic requests for agents to the scoring engine, evaluates the results, and determines which agents the interaction should be routed to. Here's a breakdown of how these requests would be processed:

First attempt

The block queries the scoring engine and 330 possible agents are returned that match the specified criteria and narrows the list down to the best 50 candidates, which are then submitted to the routing engine. As specified in the Routing Priority tab, the initial priority for these agents is set to 50.

Second attempt

Ten seconds later, the block queries the scoring engine again. This time, 400 possible agents are returned. The block evaluates all 400 of these agents, but excludes any agents that were evaluated in the previous query. Again, the list is narrowed down to the 50 best agents and passed to the routing engine. This list is cumulative, so the routing engine now has 100 possible agents. The initial priority for these agents increases to 55, as per the increment priority setting.

Third attempt

Ten seconds later, the block makes another query, but no agents are returned. Thus, no agents are submitted to the routing engine and the priority remains the same.

Fourth attempt

After ten seconds, another query is made and 70 agents are returned. The block evaluates the top 50 agents from this batch, but again excludes agents that were already evaluated during the previous queries. It determines that 22 of these agents meet the established criteria, so these are submitted to the routing engine for a total of 122 routing candidate agents. The priority for these agents is advanced to 60, as per the increment priority setting.

The attempts continue until the interaction is routed or the session ends.

Comments or questions about this documentation? Contact us for support!