Historical call selection (includes near real-time call selection)

Historical call selection refers to the selection of recorded interactions after the interaction has concluded. Any interaction captured by the recorder is eligible. An interaction can be stored on the Recorder or on Archive media.

Historical call selection is campaign-based. When performing call selection based on campaigns, ensure that your campaign has been configured before attempting to configure call selection in the Recorder Analytics Rule.

Any interaction included in one or more campaigns is evaluated against the Recorder Analytics Rule conditions. If an interaction is not included in the campaign, the Recorder Analytics Rule ignores the interaction.

Near real-time call selection

Historical call selection also optionally includes near real-time call selection. Near real-time requires a campaign, but refers to historical call selection which occurs a short interval after the interaction concludes.

Use of campaigns to select interactions

Use campaigns for:

Automatic enrollment: create a campaign that is configured to locate interactions that are ideal candidates for enrollment.

Do not use real time interactions to enroll users in voice biometrics. Enrollment requires audio of a reasonable duration by a speaker of known identity and with good audio quality. Making a determination of duration, speaker identity, and audio quality recorded is difficult while the interaction is in progress.
Watch list detection: create a campaign that selects interactions for detection on completed interactions. For example, you obtain a new voiceprint model for a target and create a campaign to process historical interactions to see if that person contacted you.
Metadata Detection - Create a campaign that is configured to locate interactions on which you want to perform Metadata Detection.

When to use historical call selection

Use historical call selection to analyze interactions anytime after the interaction concludes.

For example:

The risk management team creates a voiceprint model for a new target. You create a campaign to help you identify interactions with a high probability of a match to the target for the previous six months of historical activity.
You have a large watch list of targets and want to check the previous 12 months of interactions to check for possible target detections for analysis by the risk management team.

When to use near real-time call selection

Use near real-time call selection when you need to access the interaction identified in the alert upon receipt of the notification, such as to follow procedures for a target detection. In a near real-time scenario, an alert recipient can use Risk Management to find and replay the interaction immediately. Do not confuse real-time call selection (concurrent with interaction capture) with near real-time call selection (historical).

For example, you could use near real-time call selection when you have a small list of targets actively taking over accounts.

Steps for a typical implementation of near real-time call selection are:

In your deployment, you determine that interactions are accessible in Risk Management approximately five minutes after the interaction ends. Note that on-site testing is required to determine the appropriate interval for each deployment.
You create a campaign for a Recorder Analytics Rule that analyzes all captured interactions for targets in a watch list.
You configure the Campaign Configuration section of the Recorder Analytics Framework component to query for interactions a minimum of five minutes after the interaction ends.

Since the Recorder Analytics Framework is a server role, its campaign settings affect all voice biometrics campaigns, not just the campaign used for near real-time processing. As such, with this configuration option you cannot restrict campaign processing to non-peak processing times, such as at night.

If the interval between the conclusion of the interaction and its availability in Risk Management increases over the course of the day, consider adding more centralized voice biometrics servers. Adding one or more servers can enable campaigns to keep pace with processing the interaction volume.
After the campaign works through the backlog of captured interactions, the campaign checks for targets on interactions in near real time. When a detection occurs, a notification is delivered to the risk management team.
The risk management team accesses and replays the interaction in Risk Management for further action according to the policies and procedures of your organization.

Differences between campaigns used by Recorder Analytics Rules and Archive

Both Recorder Analytics Rules and Archive use campaigns. However, the two features generally do not use the same campaigns. After you have selected the interactions based on the campaign criteria, the actions performed with those interactions are fundamentally different.

For Archive, you configure the campaign to archive the interactions on media.
For the Real-Time Analytics (RTA) Framework, you configure the campaign to locate interactions for specific tasks, such as for enrollment or for watch list detection.

Interaction metadata updates

Historical processing generates an updated XML file with tagging used to update Contact Database records.

The analytics system generates a new Marking job XML file. The original XML file associated with the interaction is NOT updated.

Bookmark events are not updated

No new bookmarks events are added to interactions.

No historical analytics activity is archived

Although no data (such as tagging actions, case assignments, and bookmarks) produced by historical analytics activity is archived, historical analytics activity can trigger the archive of an interaction. The archive includes only the data created at the time the recording occurred.