Insights FAQ

The Insights section of OCP Console® provides the following tools:

• OCP® Monitor: review live or completed dialog sessions in detail.

• OCP® Reporting: view near real-time and historical performance dashboards.

• Export Service: access raw application data through batch exports and subscriptions.

• Optimise IQ: identify intent improvement opportunities directly from production data.

• Advanced OCP Conversational Insights®: access extended analytical dashboards covering the full application history.

For details on each tool, refer to the Insights documentation.

OCP® Monitor focuses on individual dialogs. Use it to drill into a single conversation, listen to utterances, review intents, and download dialog data.

OCP® Reporting focuses on aggregated metrics. Use it to view session volumes, success rates, and conversational outcomes across many dialogs through graphical dashboards.

Access to Insights tools depends on user permissions configured for your OCP environment. To request additional access rights, contact Omilia Care (Support Center).

Most Insights tools are available in standard OCP® environments. Some features depend on environment configuration:

• Optimise IQ is currently only available in US1.

• Voice Food Ordering (VFO) features in Monitor are available only in specific environments.

• Voice Biometrics reporting is available only when Voice Biometrics is enabled.

• Whole Call Recording (WCR) exports are available for specific environments and require activation through Omilia Care (Support Center).

In OCP Console®, go to Insights → Monitor.

• Dialog ID: the identification number of an individual dialog.

• Session ID: the identification number of the session. This value is available only for Virtual Agent.

• Connection ID: the unique identification number of the connection.

When you search by any of these IDs, all other filters are ignored.

Application Layer shows full end-to-end conversations without their subcomponents, giving you a high-level overview of each dialog. By default, it is enabled. You can expand each Dialog Application to view its components.

If you search by Dialog ID, Connection ID, or Session ID, the hierarchical structure is not displayed. Disable Application Layer when running an Advanced Search for specific miniApps patterns.

In Dialog Review, click the Play button next to the User icon to listen to the captured audio. Click Save to download the utterance as a WAV file.

Audio playback is available only for voice channel dialogs.

The system masks sensitive data captured in user utterances. When an utterance contains sensitive content, the message **encrypted content** appears in place of the text.

In Dialog Review, click the # symbol next to an utterance. In the pop-up window, select the values for:

• CISR (Concept Identification Success Rate): correctAcc, correctRej, falseAcc, falseRej, or misinterpretation.

• Custom Boolean Tag: true or false.

• Custom Tag: a free-form keyword.

Available tags depend on your environment configuration.

Open the dialog in Dialog Review, click the Options menu, and select Download Files. The following formats are available:

• Download JSON: full dialog session information in JSON format.

• Download RTF: dialog session information as plain text.

• Download Log File: dialog log file in .txt format.

Yes. In Dialog Review, open the Options menu and select Create Test Case. Enter a name, select a destination project from the dropdown, and click Create Test Case.

The dialog must be associated with a valid OCP Testing Studio® project. If no project is associated, the dropdown shows No projects found for this dialog.

The Dialog Tree, also called Path View, shows a hierarchical view of all components (miniApps, Flows) called during a session, in execution order. Click any item to search for dialogs in that miniApp.

To open it, go to Dialog Review, click the Options menu, and select Dialog Tree. This feature is available only for Virtual Agent.

For Live Agent sessions, Dialog Review displays additional metadata fields such as Master ID, CCAAS Provider, Agent ID, Agent Skill, AA Profile ID, and AA Profile Name. A POST CALL SUMMARY section may also appear, with an AI-generated summary of the call.

OCP® Monitor uses two analysis methods:

• Standard Text Matching for prompts and utterances: matches all the given words, is case-insensitive, and ignores word order.

• Events Matching for dialog events: matches all the given words in order, supports phrases, and is case-insensitive. Word order matters to avoid false matches from event string formats.

For more information, refer to the OCP Monitor User Guide.

Real-time report shows near real-time data on the current state of your application. The display refreshes automatically every 60 seconds, with a default time period of Last 120 minutes.

Conversational Insights offers historical analytical metrics across longer date ranges, with daily or hourly granularity.

The Real-time report updates automatically every 60 seconds. You can refresh manually at any time by clicking the Refresh button.

Reporting metrics are processed in 5-minute intervals with an additional 5-minute watermark to capture any external data delay.

Conversational Outcome shows the number of dialogs per Orchestrator application. The following metrics are available:

• Completed: number of completed dialogs, including terminated and transferred dialogs, as well as edge cases where the dialog end is unknown.

• Transferred: total number of dialogs transferred to a live agent.

• Terminated: total number of dialogs terminated by the user or the system.

• Error: number of dialogs completed with an error.

• Other: dialogs not classified in any of the above categories.

Click the Export Data button next to a graph to download the report as a CSV file. In Full Screen mode, click the Export button. If you narrow the time range using the slider before exporting, the file contains only the data for the selected period.

Yes. Hover over a graph to see metric times, click a specific dot, and select Show Actual Dialogs. You are redirected to OCP® Monitor with the corresponding dialog IDs filtered.

Not all graphs support this functionality. For example, the Terminated VS Transferred to Agent report doesn't have a detailed view.

The following filters are available in OCP® Reporting:

• Time period or Date range: predefined options range from Last 20 minutes to Last 48 hours for real-time, or any custom range for Conversational Insights.

• Group: the OCP group that the metrics refer to. Only one group can be selected at a time.

• Name: the name of the miniApp or Orchestrator application.

• Type: the type of miniApp. For Orchestrator applications, the type is always Flow.

• Channel: the channel of the user system interaction, such as Voice or Chat.

• Language: the language of the interaction.

• Test Flag: filter test-flagged calls. Set to true to fetch only test-flagged calls, or false to exclude them. Applicable to Real-time report.

Not all filters apply to all reports.

Open the Filters menu, set the first Date range, then click +Add comparison below to add a second range. Click View to display the comparison.

Yes. Insights Report Metrics are available through the Real-Time Metrics API, with the same filters as the UI.

In OCP Console®, go to Insights → Export Service.

The Export Service UI is accessible to all OCP users, but only users with privileged rights can create a new export or download data. To request access, contact Omilia Care (Support Center).

The following data types are supported:

• Call Detailed Recordings (CDRs): dialog_start, dialog_end, dialog_step (JSON).

• Audit Data: audit_events (JSON). Available in specific environments only.

• VB Data: vb_verification, vb_enrolment (JSON).

• Whole Call Recordings (WCRs): wcr (WAV).

Selecting All includes all data types listed above, except the standalone Audio WCR option. To activate WCR, contact Omilia Care (Support Center).

A one-off Export is a batch request for a specific date range. The data is prepared once and made available for download.

A Subscription is a near real-time data stream to a callback endpoint you configure. Subscriptions are useful for ongoing access to dialog events, call detail records, and user activity trails.

For details on Streaming availability and additional costs, coordinate with your Account Manager.

Data is available for download for 24 hours after the export request completes successfully. After that, the export status changes to Expired and you need to create a new request.

The minimum time range is 1 hour and the maximum is 48 hours per export request.

Data is not available in real time for batch exports. There is a latency of at least 2 hours before data becomes available. If you request data within this window, the message Data is not available for exports appears.

You can't have two different export requests for the same OCP group being processed at the same time. To request multiple exports for the same group, wait for the first one to complete before starting the next.

An export request can have the following statuses:

• Submitted: the request was submitted successfully.

• Processing: the request is being processed.

• Running: the request is running after processing.

• Completed: data is ready for download.

• Failed: the request did not complete successfully.

• Expired: data is no longer available for export.

Exports are delivered as ZIP files. Each ZIP includes the export ID and an order number in the filename, for example export_e3c9908c-595f-4d2a-9333-147f3ae00045_(1_of_1).zip.

JSON exports use the JSON Lines format, where each line is a self-contained JSON object. This format is well suited for streaming and processing large datasets. For more information, refer to the JSON Lines documentation.

OCP data uses UTC timestamps.

When creating an export request, select Additional encryption, then enter a valid PGP Public key and its Key ID. The export is generated with PGP encryption applied.

For more information, refer to the Exports API PGP Encryption page.

Yes. Omilia OCP® Data Warehouse and Export Service comply with General Data Protection Regulation (GDPR) requirements. For data exported from OCP, OCP users are responsible for handling the data in accordance with applicable regulations.

Yes. The Exports API is available for OCP users who have access to Export Service and want to build automated workflows.

Optimise IQ is an analytics capability that helps you continuously improve your conversational applications. It identifies utterances the virtual agent failed to recognize and suggests intent enhancements based on real user interactions and application performance.

Currently, Optimise IQ is available in AWS US1 Multitenant Production as a controlled release. Users must be explicitly added to access it.

Optimise IQ groups utterances into two categories:

• Unmatched Utterances: cases where the system failed to extract an intent, either due to no-match or out-of-domain context.

• Low Confidence: cases where the system extracted an intent with very low confidence due to ML-NLU or ASR scores.

Select Optimise Data to choose one of the following actions:

• Review All: navigate through utterances with extended filters.

• Download Data: export to CSV for offline analysis.

• Send to NLU: generate a new NLU application for review before production. Existing intents receive additional utterances, and suggested new intents are added to the model. You must first select an NLU Application in the Filters menu.

• Send to Pathfinder: planned for a future release.

You can navigate up to 2 months of historical data. Select a custom date range or one of the predefined ranges, such as Last month.

No. Data is processed, cleaned, and aggregated before display. Only non-sensitive steps such as open questions are selected for optimization. Any potential sensitive identifiers, even if provided by user mistake, are stripped during pre-processing.

The current controlled release supports the following miniApp types: Insurance, Telco, Energy, Universal, Pandemic, CarRetail, Banking, and Banking 2.

Only English is supported in the first version. Intent suggestions are supported for No Matches and for Low Confidence in NLU-driven cases.

In OCP Console®, go to Insights → Reporting → Conversational Insights → Advanced Conversational Insights®. To request additional access rights, contact Omilia Care (Support Center).

The following dashboards are available:

• Sessions Metrics: session counts, duration breakdowns, and miniApp invocations per session.

• Conversational Outcome: dialog volumes, end type distribution, no-match and no-input rates.

• Self-Services Success Rate: self-service request volumes, success rates, and fail reasons.

• Task Completion Rate: miniApp invocations, success and fail breakdowns, and fail reasons.

• Intents & Utterances: intent distribution and top utterances per intent.

• Custom Metrics: custom event loggings from Orchestrator.

• Self-Services Contribution: impact of self-services on call outcomes.

• Transfer Reasons: reasons calls were transferred and last active intents.

• Caller Classification Reporting: caller behavior categories such as Power Users, Engaged Users, and Silent Calls.

• Recaller Overview: insights on once-off callers compared to recallers within a 24-hour window.

Both dashboards report end-of-dialog distributions, but they use different definitions:

• In OCP® Reporting, dialogs are classified as Completed, Transferred, Terminated, Error, or Other.

• In Advanced OCP Conversational Insights®, dialogs are classified as Contained, Abandoned, Transferred, Errors, or Other. A dialog is Contained if at least one self-service completed successfully and the dialog was not transferred. A dialog is Abandoned if no self-service completed successfully and the dialog was not transferred.

The Caller Classification dashboard groups callers into the following categories:

• Power Users: the caller successfully requested an intent without any issues during the call.

• Engaged Users: the caller requested an intent but faced a challenge at some step, such as a no-input or no-match.

• Not Engaged: the caller entered the IVR but did not express any intent.

• Instant Calls: the call was terminated or transferred immediately, before or upon the first system request.

• Silent Calls: the call entered the IVR with no input or interaction from the caller.

• Agent Requests: the caller only requested an agent without providing any intent.

For IVR applications with Voice Biometrics enabled, Dialog Review shows the following parameters per dialog step:

• BioSearch: the result of a search by biokey against the user database. Possible values include FOUND_TOO_MANY, NO_USER_FOUND, FOUND_ONE, and FOUND_MANY.

• BioResult: a verification label corresponding to a span of bioscores, such as TRUE_HIGH, FALSE_HIGH, LOW_DATA, or ERROR.

• BioScore: a similarity score from 0 to 100% generated by comparing two voiceprints.

For more information, refer to the Voice Biometrics section.

The Voice Biometrics Reporting dashboard contains two sections:

• Enrollments report: shows overall enrollment requests, success rate, and cumulative results.

• Verifications report: shows overall verification requests, pass rate, and score distribution. By default, only TRUE_HIGH results are considered passed; you can optionally include TRUE_MEDIUM as well.

Available filters include Date range, Group, Verification origin (Agent or IVR), and Region.

If your organization uses a restrictive firewall or proxy, you or your network administrator need to allow-list the IPs used from Insights to access data for Monitor, Reporting, and Export Service.

For the current list of Data Warehouse NAT Gateway IPs per region, refer to the IP Addresses for OCP data services page.