Difference between revisions of "UseCases/Current/GenesysEngage-onpremises/BO06"

From Genesys Documentation
Jump to: navigation, search
(UCC: Updated benefit name)
Line 1: Line 1:
 
{{SMART UseCase
 
{{SMART UseCase
 +
|UCIntro=This use-case is in "Sales-Ready" status, meaning that all capabilities described in the use-case are supported by the Genesys products, but the associated services require a custom quotation. The Predictive Routing product is in Conditional Commercialization status, meaning that some capabilities are still restricted or may require PM approval.
 +
|UCOverview=Companies want to improve their business Key Performance Indicators (KPI), capitalize on innovation in Artificial Intelligence and drive business decisions with the abundance of data and context available. Predictive Routing uses machine learning to support optimization of Customer Service KPIs.
 +
 +
Customer Service KPIs are metrics measuring a customer experience or efficiency outcome of an interaction, as opposed to Sales KPIs that measure the sales outcome of an interaction. Service KPIs can be of two types:
 +
 +
*Customer experience outcome such as Net Promotor Score (NPS), Customer Satisfaction (CSAT), First Contact Resolution (FCR), and Customer Effort Score (CES)
 +
*Operational efficiency metrics such as Handle Time (AHT), transfers, hold count, hold time, cases open/closed, and back-office tasks opened
 +
 +
This use case illustrates an improvement in First Contact Resolution (FCR), captured from Genesys Info Mart or from 3rd-party surveys (for inbound voice interactions, for example).  The use case also illustrates service related KPIs, where the data for the KPIs is available in Info Mart, or another available data source.
 +
 +
Predictive Routing also applies to Sales & Marketing KPIs.  See {{Link-AnywhereElse|product=UseCases|version=Current|manual=GenesysEngage-onpremises|topic=SL06|display text=Genesys Predictive Routing for Sales (SL06) for Genesys Engage on-premises}}
 +
 +
The design of traditional routing is  to match customers to agents through skills-based or group-based logic.  The goal is to maintain a service level, rather than improving a KPI. Predictive Routing differs from traditional routing in that it uses machine learning to find signals in historical data from Genesys Info Mart and other third-party data sources.  The predictive algorithm then uses these sources to build a model that predicts the business outcome of a customer's interaction when handled by an employee.
 +
 +
The way the predictive model works to improve KPIs is by ranking agents according to their predicted impact on the business outcome. It then assigns the interaction to the highest ranked available agent.  A/B testing is the method that measures the actual impact of Predictive Routing on the target KPI, which compares the performance of Predictive Routing against the existing routing strategy.
 +
 +
 +
 +
<br />
 
|SMART_Benefits={{SMART Benefits
 
|SMART_Benefits={{SMART Benefits
 
|UCBenefitID=Improved Net Promoter Score
 
|UCBenefitID=Improved Net Promoter Score
−
|UCBenefit=Target the best agents to resolve the customer call type by reducing misroutes or repeated transactions to improve customer satisfaction measures (such as Net Promoter Score and Customer Effort Score
+
|UCBenefit=Reduce misroutes or repeated transactions to improve customer satisfaction by targeting the best agents to resolve different types of calls.
 
}}{{SMART Benefits
 
}}{{SMART Benefits
−
|UCBenefitID=Improved Employee Attrition Rate
+
|UCBenefitID=Improved Employee Attrition
 
|UCBenefit=Improve Employee Experience by routing to the Agent the work that they are more efficient or “good” at more often.
 
|UCBenefit=Improve Employee Experience by routing to the Agent the work that they are more efficient or “good” at more often.
 
}}{{SMART Benefits
 
}}{{SMART Benefits
Line 13: Line 32:
 
|UCBenefit=Reduce transfers and conferences by routing interactions to the best qualified agent and reduce handle time as more efficient workers often have a lower Average Handle Time
 
|UCBenefit=Reduce transfers and conferences by routing interactions to the best qualified agent and reduce handle time as more efficient workers often have a lower Average Handle Time
 
}}
 
}}
−
|UCIntro=This use-case is in "Sales-Ready" status, meaning that all capabilities described in the use-case are supported by the Genesys products, but the associated services require a custom quotation. The Predictive Routing product is in Conditional Commercialization status, meaning that some capabilities are still restricted or may require PM approval.
+
|UCSummary=Organizations seeking to improve the level of customer service offered to their customers realize significant benefit from Predictive Routing.  Machine learning models configured to optimize First Call Resolution are at the core of the solution.
−
|UCOverview=Business leaders want to improve their business Key Performance Indicators (KPI), leverage the innovation in Artificial Intelligence and drive business decisions with the abundance of data and context available in their business. Predictive Routing uses machine learning to support optimization of Service KPIs.
 
  
−
Service KPIs gather any types of metrics measuring a Customer Experience or Efficiency outcome of an interaction, as opposed to Sales KPIs that measure the Sales Outcome of an interaction. Service KPIs can be of two types:
+
A customer calls the contact center, and Predictive Routing uses the data captured about the customer, their journey, and the current interaction to rank all available agents according to their predicted probability of resolving the call. Configuration options manage and balance the Service Level (speed to answer) with connecting to the most suitable agent.  The result is a reduction in repeat contacts and improved FCR.
  
−
*Customer Experience or Service outcome such as Net Promotor Score (NPS), Customer Satisfaction (CSAT), First Contact Resolution (FCR), and Customer Effort Score (CES)
+
The outcome data feeds back into the machine learning model to inform future predictions. Impacts on KPIs and the performance of the machine learning models are available via real-time reports.
−
*Operational efficiency metrics such as Handle Time (AHT), transfers, hold count, hold time, cases open/closed, and back office tasks opened
 
  
−
This use case focuses on improving First Contact Resolution (FCR), captured from Genesys Info Mart or from 3rd-party surveys (for inbound voice interactions, for example) but may also be extended to other service-related KPIs, and other channel types with custom PS effort. Wherever applicable, this use case documents the impacts of choosing another channel type or another KPI, captured from Genesys data or from 3rd-party sources.<br />
+
<br />
−
Predictive Routing also applies to Sales & Marketing KPIs. See [https://repository.docs.genesys.com/Draft:UseCases/Current/Genesys Engage/SL06 Genesys Predictive Routing for Sales (SL06) for Genesys Engage]
 
−
 
 
−
Traditional routing is designed to match customers to agents through skills-based or group-based logic to maintain a service level rather than improving KPIs. Unlike traditional routing, Predictive Routing uses machine learning to find signals in historical data to build a model that predicts the business outcome of a customer's interaction when handled by an employee. This model improves KPIs by ranking agents before making the match with customers. This model also addresses the operational challenges that occur in understaffing and overstaffing scenarios while balancing the service level with improving a KPI.
 
−
 
 
−
 
 
−
Predictive Routing has built-in A/B Testing to demonstrate the uplift of the KPI provided through use of machine learning. Predictive Routing leverages a variety of Genesys or third party data sources in order to build high quality predictors. In particular, Predictive Routing supports a native integration with Performance DNA in order to obtain a granular representation of the agent's profile.
 
−
|UCSummary=The contact center manager wants to increase the percentage of the interactions achieving First Contact Resolution (FCR). The Predictive Routing solution can assist in achieving this objective. FCR can be obtained either from:
 
−
 
 
−
*Genesys Historical Reporting data, enabling you to define FCR. For example, FCR 7 is defined as the number of customers who didn't make a subsequent contact for the same service type in 7 days.
 
−
*A 3rd-party outcome data source (such as a survey).
 
−
 
 
−
Predictive Routing:
 
−
 
 
−
*Uses machine learning, a subset of Artificial Intelligence, to compare feedback of the actual outcome with the predicted outcome, helping to improve future agent-to-customer matches.
 
−
*Ranks agents predicted to resolve the inquiry on first contact.
 
−
*Provides the finest-grain match of customer contact with agent to improve FCR.
 
−
*Provides an FCR uplift using continuous learning to rank the expected FCR for agents servicing customers.
 
−
*In many customer environments, anchoring the optimization on a customer experience metric like FCR, CSAT, or NPS also drives benefits in adjacent KPIs such as a reduction in handle time, transfers, or multi-agent events (MAE) with manageable impact to service levels.
 
−
 
 
−
The result is that repeat contacts are reduced and FCR improves.
 
 
|PainPoints=* Frustrated customers or repeat interactions – increased number of transfers to identify the right person to resolve the issue​
 
|PainPoints=* Frustrated customers or repeat interactions – increased number of transfers to identify the right person to resolve the issue​
 
* Resource inefficiency and high costs – unable to identify the right agent to quickly solve the issue​
 
* Resource inefficiency and high costs – unable to identify the right agent to quickly solve the issue​
Line 61: Line 58:
 
* At least one Inbound or Digital SMART USE Case (see use-case inter-dependancies section for details)
 
* At least one Inbound or Digital SMART USE Case (see use-case inter-dependancies section for details)
 
|BusinessImageFlow={{SMART BusinessImageFlow
 
|BusinessImageFlow={{SMART BusinessImageFlow
−
|BusinessFlow='''Model Creation'''
+
|BusinessFlow=This business flow shows the use case from the perspective of the customer and agent.
−
 
 
−
The following flow shows how a model is created. The main actor of this flow is typically a Business Analyst / Data Analyst in charge of the model creation. The Analyst is a trained professional from Genesys, a partner, or a customer organization.
 
−
|BusinessImage=https://www.lucidchart.com/documents/edit/825b125b-ee08-4705-8e5b-249e7cee457b/0
 
−
|BusinessFlowDescription=#The team Lead / Supervisor and the Analyst agree on the outcome metric to be used. This use case uses FCR as the reference metric.
 
−
#The Analyst gathers customer profile, interaction profile, agent profile, and FCR data from Info Mart, and optionally from 3rd-party data sources.
 
−
#The Analyst analyzes the data to determine correlating factors/predictors and verify if the data is suitable for a predictive model.
 
−
#The Analyst creates a predictive model based on the available Info Mart data set.
 
−
#The Analyst reviews the quality of the predictive model and potential for uplift. If the quality is satisfactory, the model can be provisioned.
 
−
 
 
−
The modeling process described above may be extended to incorporate the following changes:
 
−
 
 
−
*Integration of additional 3rd-party data sources for customer profile (such as CRM), agent profile (such as WFO), content analysis data (speech or text), or outcome data (such as NPS surveys or CRM or case management)
 
−
*Selection of other KPI(s) to optimize based on Info Mart data (such as AHT) or 3rd-party data (such as NPS)
 
−
 
 
−
The selection, analysis, and integration of this data into the predictive model requires a project-based implementation that is supported by the Predictive Routing product, but not described in this use case. Contact Genesys Professional Services for details.
 
−
}}{{SMART BusinessImageFlow
 
−
|BusinessFlow='''Predictive Routing First Contact Resolution'''
 
−
 
 
−
This business flow shows the use case from the perspective of the customer and agent.
 
 
|BusinessImage=https://www.lucidchart.com/documents/edit/7deb8693-2294-4b64-8b50-9b3e115b9a89/0
 
|BusinessImage=https://www.lucidchart.com/documents/edit/7deb8693-2294-4b64-8b50-9b3e115b9a89/0
−
|BusinessFlowDescription=#The customer contacts the company using one of the available inbound channels (such as voice, e-mail, chat, mobile, work item, or Apple Business Chat).* This inbound interaction may be the result of a proactive rule on a web or mobile application.
+
|BusinessFlowDescription=#The customer contacts the company using the inbound voice channel. This inbound interaction may be the result of a proactive rule on a web or mobile application.
 
#One of the Inbound use cases for the corresponding media type handles the interaction. The context data is captured depending on the interaction and engagement type.
 
#One of the Inbound use cases for the corresponding media type handles the interaction. The context data is captured depending on the interaction and engagement type.
 
#Genesys queues the interaction until at least one agent with the required skill(s) is available.
 
#Genesys queues the interaction until at least one agent with the required skill(s) is available.
−
#Predictive Routing ranks the agents against the FCR probability for that specific interaction and returns a ranked list with values.
+
#Predictive Routing ranks the agents according to the probability that they will resolve that specific interaction first time, and returns a ranked list with values.
−
#Genesys checks if the rank for at least one agent is above the threshold.
+
#Genesys checks if the score for at least one agent is above the configured score threshold.
−
#If no agents are available within the configured timeout, the routing expands the potential target pool of agents, such as by reducing the required skill level.
+
#If no agents are available within the configured timeout, the routing expands the potential target pool of agents, by reducing the score threshold.
−
#If yes, Genesys distributes to the best available agent based on the predictive model (the agent with the highest rank) and the routing rules.
+
#If agents over the score threshold are found, Genesys distributes to the best available agent based on the predictive model (the agent with the highest rank) and the routing rules.
−
#The agent disconnects the interaction.
+
#After dealing with the customer call, the agent disconnects the interaction.
 
#The outcome is mapped to an InfoMart attribute (for example, a disposition code or custom key value pair).
 
#The outcome is mapped to an InfoMart attribute (for example, a disposition code or custom key value pair).
 
#Optional: The customer is offered a survey. The answer to the survey is stored in a third-party system.
 
#Optional: The customer is offered a survey. The answer to the survey is stored in a third-party system.
 
#Optional: Outcome data, such as case management closure, is produced and stored by a third-party application.
 
#Optional: Outcome data, such as case management closure, is produced and stored by a third-party application.
 
}}
 
}}
−
|BusinessLogic=Parameters and Business Rules - Create Predictive Model
+
|BusinessLogic='''Routing Step 1'''
  
−
'''Model Step 1'''
+
* Using a supported channel and media type, the customer creates an interaction that captures the defined metric outcome.
−
 
+
*Prerequisite: This use case requires inbound call routing.
−
*The business representative of the company decides on the KPI to be improved by Predictive Routing. This use case selects First Contact Resolution as determined by repeat callers for the Service Type.
 
−
*The company decides which metrics to use, in this case the values for First Contact Resolution X (such as 7 days) and First Contact Resolution Y (such as 35 days), to determine the periods during which the customer has contacted the company more than once.
 
−
*At least once a day the FCR metrics are evaluated against the customers that have contacted the company over the last X and Y days for each agent and service type.
 
−
*Other KPI types can be considered through a project-based implementation.
 
−
 
 
−
'''Model Step 2'''
 
−
 
 
−
*The user gathers the FCR, Agent Profile, Interaction profile, and Customer profile data from Genesys Info Mart.
 
−
*Predictive Routing then ingests the data into the system as the basis for the predictive model.
 
−
*Other Genesys or 3rd-party data sources can be considered through a project-based implementation.
 
−
 
 
−
'''Model Step 3'''
 
−
 
 
−
*The user assesses the data for predictability of FCR:
 
−
**The data is analyzed for variance in First Contact Resolution by intention and agent. The results indicate whether a predictive model can generate significant improvement from the data.
 
−
**Each data point is analyzed for its correlating factor to FCR.
 
−
*The user prepares to create the model.
 
−
 
 
−
'''Model Step 4'''
 
−
 
 
−
*The system creates the predictive model to rank the FCR for customer-to-agent matching.
 
−
*The model can be made available for further analysis before provisioning into production.
 
−
 
 
−
'''Model Step 5'''
 
−
 
 
−
*The potential uplift of FCR using the predictive model is analyzed for effectiveness and the predictive model is then provisioned for use within customer-to-agent matching.
 
−
**The analyst evaluates the model to determine the potential uplift in FCR for this predictive model.
 
−
**The system shows the expected mathematical accuracy of the model through taking a percentage (such as 80%) of the data to train the model and then using the remaining percentage (such as 20%) of the data to predict the outcome with the model. Then the system compares the predicted results with the actual results and provides model accuracy information.
 
−
 
 
−
====Parameters and Business Rules – Predictive Routing FCR====
 
−
'''Routing Step 1'''
 
−
 
 
−
*Through Inbound, Digital, Workload Management, and other use cases, the customer creates an interaction on any media type that is stored within Info Mart for the defined FCR metric. This use case supports all inbound channels involving Genesys routing. See Use Case Interdependencies for details.
 
−
*Prerequisite: This use case requires one or more use cases handling inbound interactions.
 
  
 
'''Routing Step 2'''
 
'''Routing Step 2'''
  
−
*The interaction use case identifies the primary intention of the customer (Service Type) and the initial target skill expression is set. The necessary predictors identified during model creation are available to the interaction and can be passed to Predictive Routing.
+
*The interaction use case identifies the primary intention of the customer (Service Type) and sets the initial target skill expression.
−
*The system integrates 3rd-party data sources for customer profiles during the interactions qualification process. Any required additional customer or agent profile data available to the interaction in run time can be integrated through a project-based implementation.
 
  
 
'''Routing Step 3'''
 
'''Routing Step 3'''
  
−
*This step queues the interaction and is designed to cover both agent surplus and customer surplus scenarios. When either one or multiple agents are available (agent surplus scenario), the flow immediately proceeds. Otherwise, Genesys queues the interaction until an agent is available (customer surplus scenario). The system starts to balance the service level with the business KPI through maintaining priority.
+
*This step queues the interaction and is designed to cover both agent surplus and customer surplus scenarios. When either one or multiple agents are available (agent surplus scenario), the flow immediately proceeds. Otherwise, Genesys queues the interaction until an agent is available (customer surplus scenario).
  
 
'''Routing Step 4'''
 
'''Routing Step 4'''
  
−
*Once one or more agents are available, the necessary Customer Profile, Interaction Profile, Agent Profile, and predictor information is passed to Predictive Routing to match the customer interactions against the agent for optimal outcome. This process caters to both customer surplus and agent surplus scenarios.
+
*The predictor information along with necessary call information (Customer Profile, Interaction, Agent Profile) passes to Predictive Routing as a scoring request once one or more agents are available.. The relevant machine learning model processes the request resulting in a score for each available agent for that interaction. This process caters to both customer surplus and agent surplus scenarios.
  
 
'''Routing Step 5'''
 
'''Routing Step 5'''
  
−
*The score for each of the interactions against each of the agents is returned to routing to bias the customer-to-agent matching towards the agent(s) that can deliver the highest first contact resolution.
+
*The customer-to-agent matching provides the interaction score for each of the agents to the routing engine to promote the agents that can deliver the highest benefit to the target KPI. The model compares the score of the highest ranked agent to the configured minimum score threshold, in an agent surplus scenario. If not, then the interaction is held, pending either a higher ranked agent becoming available, or the threshold reducing.
−
*The system works in both customer and agent surplus modes. In a customer surplus scenario, the system delivers the highest priority interaction(s) and when multiple interactions are targeting the available agent, the highest priority and then the highest FCR score applies.
+
*In a customer surplus scenario, where multiple interactions are waiting when an agent becomes available, the agent’s scores for each waiting interaction are compared with the minimum score threshold.  If the agent score exceeds the threshold for at least one interaction, the system proceeds to routing the highest scoring interaction.  If not, then the agent remains unassigned, pending either a lower scored interaction becoming available, or the threshold reducing.
  
 
'''Routing Step 6'''
 
'''Routing Step 6'''
  
−
*If at least one of the FCR values is above the threshold, the interaction is routed to the agent with the highest FCR value.
+
*The minimum score threshold reduced over time according to the preconfigured fallback strategy.
−
*The system delivers the interaction normally, handling any ring on no answer and exception situations (applicable to voice, chat or email) as defined in the underlying use case.
+
*The checks in Routing Step 5  repeat regularly until the identification of an agent or interaction occurs.
−
*The customer and the agent are connected.
+
**Normal target expansion, such as relaxing skill level as configured within the underlying distribution strategy, occurs.
 +
**The continual reprioritization of the interaction also occurs as do any treatments and the queued customer experience.
  
 
'''Routing Step 7'''
 
'''Routing Step 7'''
  
−
*Optionally: If no appropriately scored agent become available or the available agent does not have a sufficiently high FCR value, the value threshold is reduced after the configured amount of time.
+
*The system delivers the interaction usually, handling any ring on no answer and exception situations as defined in the underlying use case.
−
*Normal target expansion, such as relaxing skill level as configured within the underlying distribution strategy, occurs.
+
*The customer and the agent connect.
−
*The continual re-prioritization of the interaction also occurs as do any treatments and the queued customer experience.
+
 
 +
'''Routing Step 8'''
 +
 
 +
*The interaction ends when the call disconnects the customer or agent.
  
 
'''Routing Step 9'''
 
'''Routing Step 9'''
  
−
*The outcome of the interaction is captured through the agent desktop or a server-side process. Genesys APIs are invoked automatically or after an agent action to map the outcome to a Genesys interaction attribute: custom attached data or disposition code.
+
*The agent desktop or a server-side process captures the outcome of the interaction. Genesys APIs invoke automatically or after an agent action to map the outcome to a Genesys interaction attribute: custom attached data or disposition code.
 
*Info Mart captures this attribute with the Info Mart interaction record.
 
*Info Mart captures this attribute with the Info Mart interaction record.
  
 
'''Routing Step 10'''
 
'''Routing Step 10'''
  
−
*Optionally, the customer receives a survey (the survey results are not connected with Genesys and are intended to evolve with the survey use cases)
+
*Optionally, the customer receives a survey.
−
*The survey is completed (optionally) and the outcome is collected and stored by a 3rd-party application.
+
*If the customer completes (optionally) the survey, it collects and stores the outcome through a 3rd-party application.
  
 
'''Routing Step 11'''
 
'''Routing Step 11'''
  
−
*Optionally, the outcome data is produced and stored by a third-party application.
+
*Optionally, a third-party application produces and stores the outcome data.
−
 
+
|DistributionLogic=The inbound use case provides details of the distribution of an interaction to an agent. Refer to the flow above to understand how Predictive Routing influences the distribution logic.
−
====Scenarios with Performance DNA Integration (Optional)====
+
|CustomerInterfaceRequirements=There are no specific customer interface requirements.
−
=====<u>Scenario 1: Genesys Predictive Routing and Agent profile enrichment with PDNA data</u>=====
+
|AgentDeskRequirements=Target agents can review Attached Data/Case Data when an interaction routes to their Agent Desktop.
−
 
+
|RealTimeReporting=Predictive Routing real-time reports include:
−
{{{!}} class="wrapped confluenceTable tablesorter tablesorter-default stickyTableHeaders wikitable" role="grid"
 
−
{{!}}- class="tablesorter-headerRow" role="row"
 
−
! class="confluenceTh tablesorter-header sortableHeader tablesorter-headerUnSorted" style="width: 81.8125px;" role="columnheader" scope="col" data-column="0" aria-label="Section Name: No sort applied, activate to apply an ascending sort" data-mce-style="width: 81.8125px;" tabindex="0" aria-disabled="false" aria-sort="none"{{!}}Section Name
 
−
! class="confluenceTh tablesorter-header sortableHeader tablesorter-headerUnSorted" style="width: 900.812px;" role="columnheader" scope="col" data-column="1" aria-label="Description: No sort applied, activate to apply an ascending sort" data-mce-style="width: 900.812px;" tabindex="0" aria-disabled="false" aria-sort="none"{{!}}Description
 
−
{{!}}- role="row"
 
−
{{!}} class="confluenceTd" style="width: 81.8125px;" data-mce-style="width: 81.8125px;"{{!}}Business Flow
 
−
{{!}} class="confluenceTd" style="width: 900.812px;" data-mce-style="width: 900.812px;"{{!}}
 
−
#The Team Lead / Supervisor and the Analyst agree on the outcome metric to be used. This use case uses FCR as the reference metric.
 
−
#The Analyst gathers customer profile, interaction profile, agent profile, and FCR data from Info Mart, and optionally from 3rd-party data sources.
 
−
#The Analyst evaluates PDNA data and<span class="inline-comment-marker valid" data-ref="fd39fe19-e136-4745-8d99-ee4099a05423">attaches it to Agent Profile.</span>Analyst is able to flexibly choose and decide which data he wants to pull from PDNA. PDNA could contain multiple agent performance metrics. Agent skill proficiency through Assessments and Learning Items is core meter to PDNA Performance evaluations.
 
−
#PDNA strands may also include data from QM systems grading agents on their performance in past interactions.
 
−
#The Analyst analyzes the data to determine correlating factors/predictors and verify if the data is suitable for a predictive model. PDNA data is key data source impacting FCR prediction
 
−
#The Analyst creates a predictive model based on the available Info Mart data set, along with PDNA enriched Agent profile data
 
−
##The Analyst reviews the quality of the predictive model and potential for uplift. If the quality is satisfactory, the model is provisioned.
 
−
{{!}}- role="row"
 
−
{{!}} colspan="1" class="confluenceTd" style="width: 81.8125px;" data-mce-style="width: 81.8125px;"{{!}}Operational Flow
 
−
{{!}} colspan="1" class="confluenceTd" style="width: 900.812px;" data-mce-style="width: 900.812px;"{{!}}
 
−
#First time model training<span class="inline-comment-marker valid" data-ref="eebbe85c-66ee-4530-97d7-dc0013231dba">assumes that PDNA data is available for historical interactions</span><br />
 
−
#Agents will have the PDNA data updated on a nightly basis.
 
−
#PDNA scores are recalculated based on new trainings that an agent undergoes
 
−
#For any agent going through new training their disjoint (local) model is de-activated and they get scored using a global model<span class="inline-comment-marker valid" data-ref="50d072a1-9fcc-42f3-a7ac-d1b770172212">until they build critical mass</span>of new interactions to retrain a local model
 
−
#Nightly PDNA data updates and model re-trainings happen automatically to keep Predictive Routing lift and accuracy up.
 
−
{{!}}- role="row"
 
−
{{!}} colspan="1" class="confluenceTd" style="width: 81.8125px;" data-mce-style="width: 81.8125px;"{{!}}Reporting
 
−
{{!}} colspan="1" class="confluenceTd" style="width: 900.812px;" data-mce-style="width: 900.812px;"{{!}}New reports need to be defined to measure PDNA impact towards Predictive model performance and Lift
 
−
{{!}}}
 
−
 
 
−
 
 
−
=====<u>Scenario 2:Daily Agent Survey using PDNA for improved scoring</u>=====
 
−
 
 
−
{{{!}} class="relative-table wrapped confluenceTable tablesorter tablesorter-default stickyTableHeaders wikitable" role="grid"
 
−
{{!}}- class="tablesorter-headerRow" role="row"
 
−
! class="confluenceTh tablesorter-header sortableHeader tablesorter-headerUnSorted" role="columnheader" scope="col" data-column="0" aria-label="Section Name: No sort applied, activate to apply an ascending sort" tabindex="0" aria-disabled="false" aria-sort="none"{{!}}Section Name
 
−
! class="confluenceTh tablesorter-header sortableHeader tablesorter-headerUnSorted" role="columnheader" scope="col" data-column="1" aria-label="Description: No sort applied, activate to apply an ascending sort" tabindex="0" aria-disabled="false" aria-sort="none"{{!}}Description
 
−
{{!}}- role="row"
 
−
{{!}} colspan="1" class="confluenceTd"{{!}}Business Flow
 
−
{{!}} colspan="1" class="confluenceTd"{{!}}
 
−
#A customer data analyst has deployed an FCR optimization model which uses PDNA data as input.
 
−
#PDNA application is configured by IT to survey Agents in the beginning of their shift
 
−
#PDNA calculates scores based on survey response of agents
 
−
#PDNA pushes latest strand scores to Predictive Routing agent profiles to be used for scoring
 
−
#Through out the day PDNA score are used as input to Predictive routing models for FCR optimization
 
−
{{!}}- role="row"
 
−
{{!}} colspan="1" class="confluenceTd"{{!}}Operational Flow
 
−
{{!}} colspan="1" class="confluenceTd"{{!}}
 
−
#PDNA is configured to push data to Predictive Routing
 
−
#Daily sync of PDNA strands occurs right after the beginning of the agent shift
 
−
#*''Note that we may need enhancement to push on new data as opposed to by specific time''
 
−
#Model re-training is NOT required
 
−
#Model refresh/re-training cadence continues based on the business criteria of KPI lifecycle, Agent Churn and Customer Churn
 
−
{{!}}- role="row"
 
−
{{!}} colspan="1" class="confluenceTd"{{!}}Reporting
 
−
{{!}} colspan="1" class="confluenceTd"{{!}}New reports needs to be defined to view the PDNA sync information shared with GPR
 
−
{{!}}}
 
−
|DistributionLogic=The details of the distribution of an interaction to an agent are defined in the underlying inbound use cases. Refer to the flow above to understand how Predictive Routing influences the distribution logic.
 
−
 
 
−
Predictive Routing provides a routing lever that can be used to control how customer-to-agent matching behaves in customer surplus mode to distribute the interactions based on agent occupancy.
 
−
|CustomerInterfaceRequirements=N/A
 
−
|AgentDeskRequirements=Target agents can review Attached Data/Case Data when an interaction is routed to their Agent Desktop.
 
−
|RealTimeReporting=Predictive Routing realtime reports include:
 
−
 
 
−
'''Queue KPIs By Predictive Model'''
 
−
 
 
−
Monitor your Queue performance with the prediction scores provided by Genesys Predictive Routing to optimize your business KPI.
 
−
 
 
−
[[File:PRM_Queue_Group_by_Model_View_3.png]]
 
−
 
 
−
'''Agent Group KPIs by Predictive Model'''
 
−
 
 
−
Monitor the Agent Group Performance with the volume of answered interactions by range of predicted scores provided by Genesys Predictive Matching to optimize your business KPI.
 
  
−
[[File:PRM_Agent_Group_By_Model_View_1_3.png]]
+
(Forthcoming) - Replace with KPI Outcome, Feature Coverage and Model Accuracy?
−
|HistoricalReporting=The historical reports available within the Predictive Matching Data Store include:  
+
|HistoricalReporting=The historical reports available within the <s>Predictive Matching Data</s> Store include: (ARE THESE IN GCXI?)
  
 
*Predictive Routing Operational Report tracks Predictive Routing operational statistics.
 
*Predictive Routing Operational Report tracks Predictive Routing operational statistics.
Line 261: Line 138:
  
 
<sup>*</sup><small>A/B reports can be developed from any standard or custom Info Mart data. If the outcomes data is NOT integrated with Info Mart, the creation of A/B reports must be evaluated as a separate effort.</small>
 
<sup>*</sup><small>A/B reports can be developed from any standard or custom Info Mart data. If the outcomes data is NOT integrated with Info Mart, the creation of A/B reports must be evaluated as a separate effort.</small>
−
|DocVersion=v 2.2.0
 
 
|GeneralAssumptions=The following use cases are exceptions where Predictive Routing cannot be integrated:
 
|GeneralAssumptions=The following use cases are exceptions where Predictive Routing cannot be integrated:
  
Line 267: Line 143:
 
*{{#mintydocs_link:topic=EE04}}:  Enable Schedule-based routing.
 
*{{#mintydocs_link:topic=EE04}}:  Enable Schedule-based routing.
 
*Self-Service Use Cases
 
*Self-Service Use Cases
−
*Outbound preview + agent reservation used for Predictive and progressive outbound
+
*Outbound preview and agent reservation used for Predictive and Progressive outbound
 
*Callback, web monitor, proactive assist, co-browse, and Knowledge Management.
 
*Callback, web monitor, proactive assist, co-browse, and Knowledge Management.
 
|CustomerAssumptions=*Customer has all compatible versions of URS, IRD, Composer, GIM, GCXI, GII and Pulse or upgrades have been scoped in to the project plan.
 
|CustomerAssumptions=*Customer has all compatible versions of URS, IRD, Composer, GIM, GCXI, GII and Pulse or upgrades have been scoped in to the project plan.
 
*Customer has necessary systems and processes in place to track results and measure impact over the life of the model
 
*Customer has necessary systems and processes in place to track results and measure impact over the life of the model
 
*Customer identification must be available and stored in Infomart.
 
*Customer identification must be available and stored in Infomart.
−
*Significant percentage of interactions require to have a customer id linked. If less than 10 percent interactions have a customer id then building models may not be feasible
+
*<s>Significant percentage of interactions require to have a customer id linked. If less than 10 percent interactions have a customer id then building models may not be feasible</s>
 
|RequiresOr=CE01, CE02
 
|RequiresOr=CE01, CE02
−
|Optional=CE16, CE18, CE19, EE14
+
|Optional=EE14
−
|Exceptions=CE12, EE04
+
|Exceptions=CE12, CE16, CE18, CE19, EE04
−
|PremiseAssumptionsAdditional=*'''<span class="s1">Predictive Routing solution is offered to On-Premise customers from the Genesys Cloud in a Cloud-only supporting Hybrid Architecture.</span>'''
+
|PremiseAssumptionsAdditional=*Predictive Routing solution is offered to On-Premise customers from the Genesys Cloud in a Cloud-only supporting Hybrid Architecture.
−
*'''<span class="s1">Requires PM approval.</span>'''
+
*Requires PM approval.
−
*<span class="s1"></span>
+
*<span class="s1"></span>Predictive Routing is offered as a managed service by Genesys Professional Services, who deals with all aspects of machine learning model creation and maintenance.  A Professional Services package is mandatory for implementation and support of Predictive Routing.
−
*FCR 7 is defined as the number of customers who didn't make a subsequent contact for the same service type within 7 days.
+
*The Standard deployment materials address Inbound voice interactions only based on Info Mart data only.
−
*The Standard deployment materials address Inbound voice interactions only, Informart data only, including FCR 7.
 
 
*Omnichannel and outbound integrations require Genesys Product Management approval and Genesys Professional Services support.
 
*Omnichannel and outbound integrations require Genesys Product Management approval and Genesys Professional Services support.
 
*Integration of additional data sources, Genesys or 3rd-party, requires a dedicated assessment and implementation by Genesys Professional Services.
 
*Integration of additional data sources, Genesys or 3rd-party, requires a dedicated assessment and implementation by Genesys Professional Services.
 
*Customer should have one or more Genesys channels.
 
*Customer should have one or more Genesys channels.
−
*Prerequisites: An implemented use case for one or more channels and Info Mart reporting. These use cases populate the predictors in the routing and the data necessary to build the models. This solution cannot use data that is not present.
+
*Prerequisites: An implemented use case for one or more channels and Info Mart reporting. These use cases populate the predictors in the routing and the data necessary to build the models.
 
*The capture and analysis of FCR KPIs is not part of Info Mart out-of-box statistics and is developed during model creation.
 
*The capture and analysis of FCR KPIs is not part of Info Mart out-of-box statistics and is developed during model creation.
 
*Routing to agents is based on skill expressions or group-based routing.
 
*Routing to agents is based on skill expressions or group-based routing.
−
*There is no industry standard definition for FCR. The definition chosen in this use case is intended to be a guideline metric that can be repeated across our customers who have not defined this metric.
+
|DocVersion=v 2.2.0
−
*PDNA integration must be evaluated and scoped by PS as part of the implementation effort as every customer may have different types of PDNA data available.
 
 
}}
 
}}

Revision as of 14:59, December 9, 2020

This topic is part of the manual Genesys Engage On-Premises Use Cases for version Current of Genesys Use Cases.
Important
This use-case is in "Sales-Ready" status, meaning that all capabilities described in the use-case are supported by the Genesys products, but the associated services require a custom quotation. The Predictive Routing product is in Conditional Commercialization status, meaning that some capabilities are still restricted or may require PM approval.
Place CX and agent efficiency at the center of your routing decisions using AI to match each customer interaction with the best agent

What's the challenge?

Today’s contact centers generate large volumes of data and have outgrown legacy skill and queue-based routing for matching customers and agents. It is almost impossible to optimize for metrics such as First Call Resolution (FCR) or Average Handling Time (AHT) because thousands of if-then rules have to be built and managed.

What's the solution?

Genesys Predictive Routing works in real-time, using AI to analyze 100s of data points to discover patterns to match customers to the best agents. With Genesys Predictive Routing, contact centers can improve customer experiences, grow revenue, improve efficiency, and optimize for important KPIs.

Use Case Overview

Story and Business Context

Companies want to improve their business Key Performance Indicators (KPI), capitalize on innovation in Artificial Intelligence and drive business decisions with the abundance of data and context available. Predictive Routing uses machine learning to support optimization of Customer Service KPIs.

Customer Service KPIs are metrics measuring a customer experience or efficiency outcome of an interaction, as opposed to Sales KPIs that measure the sales outcome of an interaction. Service KPIs can be of two types:

  • Customer experience outcome such as Net Promotor Score (NPS), Customer Satisfaction (CSAT), First Contact Resolution (FCR), and Customer Effort Score (CES)
  • Operational efficiency metrics such as Handle Time (AHT), transfers, hold count, hold time, cases open/closed, and back-office tasks opened

This use case illustrates an improvement in First Contact Resolution (FCR), captured from Genesys Info Mart or from 3rd-party surveys (for inbound voice interactions, for example). The use case also illustrates service-related KPIs, where the data for the KPIs is available in Info Mart or another available data source.

Predictive Routing also applies to Sales & Marketing KPIs. See Genesys Predictive Routing for Sales (SL06) for Genesys Engage on-premises

Traditional routing matches customers to agents through skills-based or queue-based logic. The goal is to maintain a service level, rather than to improve a KPI. Predictive Routing differs from traditional routing in that it uses machine learning to detect patterns in historical data from Genesys Info Mart and other third-party data sources. The predictive algorithm then uses these sources to build a model that predicts the business outcome of a customer's interaction when handled by an employee.

The predictive model works to improve KPIs by ranking agents according to their predicted impact on the business outcome. It then assigns the interaction to the highest ranked available agent. A/B testing measures the real-world impact of Predictive Routing on the target KPI by comparing Predictive Routing performance against the existing routing strategy.



Use Case Benefits*

The following benefits are based on benchmark information captured from Genesys customers and may vary based on industry, lines of business or Genesys product line:

Use Case Benefits Explanation
Improved Customer Experience Reduce misroutes or repeated transactions to improve customer satisfaction by targeting the best agents to resolve different types of calls.
Improved First Contact Resolution Improve First Contact Resolution by routing the interaction to the most proficient Agent available to handle it.
Reduced Employee Attrition Improve Employee Experience by routing work to the Agent that they are more efficient or “good” at more often.
Reduced Handle Time Reduce transfers and conferences by routing interactions to the best qualified agent and reduce average handle time by having more efficient workers take each interaction.
*You can sort all use cases according to their stated benefits here: Sort by benefits

Summary

Organizations seeking to improve the level of customer service offered to their customers realize significant benefits from Predictive Routing. Machine learning models configured to optimize metrics such as First Call Resolution are at the core of the solution.

A customer calls the contact center, and Predictive Routing uses the data captured about the customer, their journey, and the current interaction to rank all available agents according to their predicted probability of resolving the call. Configuration options manage and balance the Service Level (speed to answer) with connecting to the most suitable agent. The result is a reduction in repeat contacts and improved FCR.

The outcome data feeds back into the machine learning model to inform future predictions. Impacts on KPIs and the performance of the machine learning models are available via real-time reports.



Use Case Definition

Business Flow

This business flow shows the use case from the perspective of the customer and agent.

Business Flow Description

  1. The customer contacts the company using the inbound voice channel. This inbound interaction can be the result of a proactive rule on a web or mobile application.
  2. One of the Inbound use cases for the corresponding media type handles the interaction and captures interaction context data. The exact data captured depends on the interaction and engagement type.
  3. Based on the interaction context, Genesys selects an initial group of agents with the required skill(s) as possible routing targets to handle the interaction.
  4. Predictive Routing calculates the scores of the agents in the target group using a machine learning model that takes into account the agents' historic performance on similar interactions.
  5. When there are multiple agents available, Genesys attempts to route the interaction to the available agent with a highest score.
  6. If there is an interaction surplus and an agent becomes ready, Genesys selects an interaction from the queue taking into account the priority of each waiting interaction, the score the agent has for each interaction, and the time the interactions were queued.
  7. If no agents are available within the configured timeout, the routing strategy expands the potential target pool of agents by reducing the skill requirements and then repeats the target agent selection using Predictive Routing.
  8. After dealing with the customer call, the agent disconnects the interaction.
  9. The outcome is mapped to Genesys Info Mart attribute (for example, a disposition code or custom key-value pair).
  10. Optional: The customer is offered a survey. The answer to the survey is stored in a third-party system.
  11. Optional: Outcome data, such as case management closure, is produced and stored by a third-party application.

Business and Distribution Logic

Business Logic

Routing Step 1

  • Using a supported channel and media type, the customer creates an interaction. As the interaction is handled and traced through your environment, data is captured that enables you to determine the outcome for the metric you want to optimize.
  • Prerequisite: This use case requires inbound call routing.

Routing Step 2

  • The interaction use case identifies the customer's primary intention (Service Type) and sets the initial target skill expression.

Routing Step 3

  • This step queues the interaction and is designed to cover both agent surplus and customer surplus scenarios. When either one or multiple agents are available (agent surplus scenario), the flow immediately proceeds. Otherwise, Genesys queues the interaction until an agent is available (customer surplus scenario).

Routing Step 4

  • The predictor information along with necessary call information (Customer Profile, Interaction, Agent Profile) passes to Predictive Routing as a scoring request once one or more agents are available. The relevant machine learning model processes the request, resulting in a score for each available agent for that interaction. This process handles both customer surplus and agent surplus scenarios.

Routing Step 5

  • The customer-to-agent matching provides the interaction score for each of the agents to the routing engine to identify the agents that can deliver the highest benefit to the target KPI. In an agent surplus scenario, the model compares the score of the highest ranked agent to the configured minimum score threshold. If the score is below the threshold, then the interaction is held until a higher-ranked agent becomes available or the threshold is reduced.
  • In a customer surplus scenario, where multiple interactions are waiting when an agent becomes available, the agent’s scores for each waiting interaction are compared with the minimum score threshold. If the agent score exceeds the threshold for at least one interaction, the system proceeds to route the highest scoring interaction. If not, then the agent remains unassigned until a lower-scored interaction becomes available or the threshold is reduced.

Routing Step 6

  • The minimum score threshold is reduced over time according to the preconfigured fallback strategy.
  • The checks in Routing Step 5 repeat regularly until an agent-interaction matchup meets the threshold requirements.
    • Normal target expansion, such as relaxing skill level as configured within the underlying distribution strategy, occurs.
    • The continual reprioritization of the interaction also occurs, as do any treatments and the standard queued customer experience.

Routing Step 7

  • The system delivers the interaction, handling any ring on no answer and exception situations as defined in the underlying use case.
  • The customer and the agent connect.

Routing Step 8

  • The interaction ends when the customer or agent disconnects the call.

Routing Step 9

  • The agent workspace or a server-side process captures the interaction outcome. Genesys APIs invoke, either automatically or after an agent action, to map the outcome to a Genesys interaction attribute, which can be custom attached data or a disposition code.
  • Genesys Info Mart captures this attribute in the Info Mart interaction record.

Routing Step 10

  • Optionally, the customer receives a survey.
  • If the customer completes the survey, the system collects and stores the outcome through a 3rd-party application.

Routing Step 11

  • Optionally, a third-party application produces and stores the outcome data.

Distribution Logic

The inbound use case provides details of the distribution of an interaction to an agent. Refer to the flow above to understand how Predictive Routing influences the distribution logic.

User Interface & Reporting


Agent UI

Target agents can review Attached Data/Case Data when an interaction routes to their Agent Workspace.

Reporting

Real-time Reporting

Predictive Routing does not include real-time reports. Operational reports are available in the Predictive Routing UI.

Operational reports include:

  • KPI Outcome
  • Feature Coverage
  • Model Accuracy

Historical Reporting

The historical reports available through GCXI are the following:

  • Predictive Routing Operational Report - tracks Predictive Routing operational statistics.
  • Predictive Routing A/B Testing Report - tracks A/B testing results for Predictive Routing models and predictors.
  • Predictive Routing Agent Occupancy Report - tracks Agent Occupancy while Predictive Routing is being used to optimize routing.
  • Predictive Routing Daily Queue Statistics Report - tracks KPIs for each Queue while Predictive Routing is being used to optimize routing.
  • Predictive Routing Detail Report - interaction-level detail data about Predictive Routing use and its impact on KPIs.

*A/B reports can be developed from any standard or custom Info Mart data. If the outcomes data is NOT integrated with Info Mart, the creation of A/B reports must be evaluated as a separate effort.

Customer-facing Considerations

Interdependencies

All required, alternate, and optional use cases are listed here, as well as any exceptions.

All of the following required: At least one of the following required: Optional Exceptions
None

Inbound

Digital

Workforce Engagement

Digital

Inbound

Outbound

Self-Service and Automation


General Assumptions

  • Requires Product Management approval.
  • Predictive Routing solution is offered to on-premises customers in a hybrid architecture that incorporates core functionality served from the components deployed in your own environment.
  • Predictive Routing is offered as a managed service by Genesys Professional Services, who deal with all aspects of machine-learning model creation and maintenance. A Professional Services package is mandatory for implementation and support of Predictive Routing.
  • The standard deployment materials address Inbound voice interactions based on Genesys Info Mart data only.
  • Integration of additional data sources, whether Genesys or 3rd-party, requires a dedicated assessment and implementation by Genesys Professional Services.
  • Customer must have implemented a use case for one or more channels and have deployed Genesys Info Mart reporting. These use cases populate the data used to build predictors and models, which direct how interactions are routed. Note that the capture and analysis of FCR KPIs is not part of Genesys Info Mart out-of-box statistics and is developed during model creation.

Note the exceptions where Predictive Routing cannot be integrated listed in the interdependencies section:

  • Self-Service use cases
  • Outbound preview and agent reservation used for Predictive and Progressive outbound

Customer Responsibilities

  • Customer has already optimized traditional routing strategies and processes and wants to achieve further improvements.
  • Customer has all compatible versions of URS, IRD, Genesys Info Mart, GCXI, and Pulse; or upgrades have been scoped in to the project plan.
  • Customer has the necessary systems and processes in place to track results and measure impact over the life of the model.
  • Customer identification is available and stored in Genesys Info Mart.



Document Version

  • Version v 2.2.1 last updated December 9, 2020

Retrieved from "https://all.docs.genesys.com/UseCases/Current/GenesysEngage-onpremises/BO06 (2025-06-18 21:19:43)"
Comments or questions about this documentation? Contact us for support!