Anomaly Detection Dashboard

AI-generated summary

Genesys Cloud's Anomaly Detection Dashboard is a forthcoming feature designed to identify unusual patterns in contact center performance metrics by comparing detected values against expected values and generating anomaly scores. The system monitors 14 supported queue performance metrics across inbound and outbound voice, message, and email interactions, including Interaction Offered, Interaction Transferred, Interaction Handled, Average Handle Time, Longest Handle Time, Average Hold Time, Average ACW Time, Average Speed of Answer (ASA), Max Wait Time, Service Level, Abandon Rate, Answered Percentage, and Average Talk Time. Analysis occurs at 15-minute intervals while model training and result generation happen nightly. Access requires granular permissions across multiple signal categories, including Queue Anomaly , Queue Anomaly Query, Anomaly Detection Settings, and Monitored Queues views, with additional permissions needed for editing settings, exporting events, and viewing queue configurations. Users can access the dashboard via Menu > Analytics > Anomaly Detection and export current data through the Export function. The dashboard interface includes an anomaly trend chart displaying anomaly distribution over selected time periods with adjustable intervals, where hovering reveals specific dates, times, and anomaly counts. An anomaly table provides detailed information about each detected anomaly, including affected metrics, queues, media types, detected versus expected values, anomaly scores, and consensus analysis showing how many detection models identified a data point as anomalous. The dashboard automatically adjusts display times according to the selected time zone and indicates partial data availability through checkered bars. Filter functionality allows users to narrow displayed anomalies by queue, metric, media type, division, time window, and anomaly score range. The Show filter option enables users to display all results, detected anomalies only, or false positives only. Users can mark anomalies as false positives when detected anomalies represent expected behavior or lack relevance to analysis, which improves model accuracy over time. False positives are excluded from the standard Anomalies view but remain accessible through filtering options, enabling users to distinguish between anomalies requiring investigation and those that do not

Feature coming soon

PrerequisitesClick to expand

The following permissions:

To access the Anomaly Detection dashboard:

Signals > Queue Anomaly Summary > View
Signals > Queue Anomaly Query > View
Signals > Anomaly Detection Settings > View
Signals > Anomaly Detection Monitored Queues > View

To edit anomaly detection settings:

Signals > Anomaly Detection Settings > Edit

To view/search anomaly events:

Signals > Queue Anomaly Query > View
Signals > Queue Anomaly > View
Signals > Queue Anomaly Summary > View

To export anomaly events:

Signals > Queue Anomaly Export Job > Add
Signals > Queue Anomaly Export Job > View

To edit anomaly:

Signals > Queue Anomaly > Edit

To view analysis:

Signals > Queue Analysis > View

To view the configuration of available queues:

Signals > Anomaly Detection Available Queues > View

To view the configuration of monitored queues:

Signals > Anomaly Detection Monitored Queues > View

To edit the configuration of monitored and available queues:

Signals > Anomaly Detection Monitored Queues > Edit

The Anomaly Detection dashboard helps you identify unusual patterns in contact center performance metrics. The dashboard compares detected metric values with expected values and provides an anomaly score to help you identify and investigate unexpected changes in performance.

Anomaly detection supports inbound and outbound voice, message, and email interactions for monitored queues. Genesys Cloud analyzes signals for monitored queues at 15-minute intervals. Model training occurs nightly, and the analysis results used for anomaly detection, including expected values, bounds, and deviation, are generated nightly. Predictions for monitored signals are then generated every 15 minutes. For more information on anomaly detection, see Overview of anomaly detection for queues.

Use the anomaly trend chart to view the distribution of detected anomalies over selected time period. To view the date, time interval, and number of anomalies detected during that interval, hover over a bar.

The anomaly table provides details about each detected anomaly, including the affected metric, queue, media type, detected and expected values, anomaly score, analysis time window, and consensus analysis.

To access your Virtual Agent performance dashboard, click Menu > Analytics > Anomaly Detection.

To export the current data, click Export.

Anomaly trend chartClick to expand

The anomaly trend chart displays the distribution of detected anomalies over the selected time period. The chart automatically adjusts its time intervals based on the time range selected in the dashboard filter.

Each bar represents the anomalies detected during a specific time interval. Hover over a bar to view:

Date — The date and time interval of the anomalies.
# of anomalies — The number of anomalies detected during that interval.

The dashboard displays times according to the selected time zone.

Note: Checkered bars indicate that only partial data is available for that time period.

Anomaly Detection settingsClick to expand

Filter for dashboardClick to expand

Use the Filter option to narrow the anomalies displayed on the dashboard to only the data that’s relevant to you.

Click Filter to open the filter panel, and then set one or more of the following criteria:

Filter	Description
Queue	Select one or more queues for which you want to view detected anomalies.
Metric	Select the metrics for which you want to view anomaly data.
Media	Select the media types for which you want to view anomaly data.
Division	Select a division to limit the results to anomalies associated with that division.
Time window start	Select the start date and time of the period for which you want to view anomaly data.
Time window end	Select the end date and time of the period for which you want to view anomaly data.
Minimum score	Enter the minimum anomaly score to include in the results.
Maximum score	Enter the maximum anomaly score to include in the results.
Show	Select any of the following required types of results to display: All — Displays both detected anomalies and results identified as false positives. Anomalies — Displays detected anomalies. False positives — Displays results identified as false positives.

Click Apply to apply the filter criteria.

Queue anomaly metricsClick to expand

When you enable anomaly detection monitoring for a queue, Genesys Cloud monitors supported queue performance metrics for unusual behavior.

The following metrics are available for queue anomaly detection:

Metrics	Description
Interaction Offered	The number of interactions offered to a queue.
Interaction Transferred	The number of interactions transferred to a queue.
Interaction Handled	The number of interactions handled in a queue.
Average Handle Time	The average amount of time spent handling interactions. Calculated by: `tHandle_sum / tHandle_count`
Longest Handle Time	The longest amount of time spent handling an interaction.
Average Hold Time	The average amount of time interactions were placed on hold. Calculated by: `tHeldComplete_sum / tHeldComplete_count`
Average ACW Time	The average amount of time spent on after-call work. Calculated by: `tAcw_sum / tAcw_count`
Average Speed of Answer (ASA)	The average amount of time before an interaction is answered. Calculated by: `tAnswered_sum / tAnswered_count`
Max Wait Time	The maximum amount of time before an interaction is answered.
Service Level	The service-level performance for the queue.
Abandon Rate	The proportion of offered interactions that were abandoned. Calculated by: `tAbandon_count / nOffered_sum`
Answered Percentage	The percentage of offered interactions that were answered. Calculated by: `(Answered / Offered) × 100`
Average Talk Time	The average amount of time spent talking during interactions. Calculated by: `tTalk_sum / tTalk_count`

Anomaly detection supports these metrics for supported inbound and outbound voice, message, and email interactions.

Dashboard metricsClick to expand

The dashboard also displays the total number of anomalies detected during the selected time period above the table.

The dashboard provides the following information for each detected anomaly:

Metrics	Description
Detected Time	The date and time when Genesys Cloud detected the anomaly.
Metric	The performance metric for which the anomaly was detected. For the complete list of metrics available for anomaly detection, see Queue anomaly metrics section in this article.
Queue	The queue associated with the detected anomaly.
Media	The media type associated with the anomaly, including voice, email, or message.
Detected Value	The actual value of the metric that Genesys Cloud identified as anomalous.
Expected Value	The value that Genesys Cloud expected for the metric based on the anomaly detection analysis.
Anomaly score	Displays the confidence or strength of the detected anomaly. A higher percentage indicates that the detected value differs from the expected behavior.
Time window start	The beginning of the time interval that Genesys Cloud analyzed for the anomaly.
Time window end	The end of the time interval that Genesys Cloud analyzed for the anomaly.
Consensus analysis	Displays how many anomaly detection models identified the data point as anomalous. For example, 3 of 3 models indicate that all three models identified the anomaly.

Anomaly details chart

Select a metric to open its details panel. The metric details panel provides additional information about the selected anomaly, including a line chart that shows the metric values and anomaly detection bounds for the selected period:

Gray bounds — Represent the anomaly detection bounds calculated for the metric.
Blue line — Represents the actual metric values from Analytics.

Note: The blue line represents actual analytics values, not the expected values generated by anomaly detection.

Mark an anomaly as false positive

If a detected anomaly represents expected behavior or is not relevant to your analysis, you can mark it as a false positive. Marking an anomaly as a false positive helps distinguish it from anomalies that require further investigation, improving the accuracy of the anomaly detection models over time.

To mark an anomaly as a false positive:

In the anomaly table, locate the anomaly that you want to mark as a false positive.
Click More at the end of the anomaly row.
Click Mark as false positive.

Alternatively, select a metric to open its details panel, and then click Mark as false positive.

Anomalies marked as false positives are excluded from the Anomalies view. After an anomaly is marked as a false positive, you can use Filter > Show > False positives to view anomalies identified as false positives. Use Show > All to view both anomalies and false positives.

[NEXT] Was this article helpful?

Get user feedback about articles.