Reviewing calls
Calls you can access in a call analysis group appear in one sortable table: native outbound Call Center calls, audio uploads, Genesys imports, and added transcripts. Navigate to Call analysis → Calls from the main menu and pick a group from the Contact center group selector at the top left. Outbound Call Center calls appear here only after you enroll them through a Human outbound call Flow.
If your organization runs Call analysis in Groups, the selector lists only the Call Analysis groups you can access. As a team lead you see only calls with at least one of the agents you coach on them; calls of agents without a team lead, and calls with no agent, are visible to admins only. With the Agent role you see only calls you handled yourself.
Calls list
The table shows the most recently added calls first. Click the arrow next to Call date to order by when the calls took place instead, which differs when you have imported historical recordings. Click any row to open the call detail, with its transcript, summary, and evaluation results.
Use the Columns button to choose which columns are visible. Your choice is remembered per browser, so the view you set up stays that way next time you open the page.
Every data column supports inline filtering: look for the filter icon in the column header. A dot on the icon means the filter is active. Filters combine as you would expect: within a single column the values are alternatives, so picking Failed and Pending returns calls in either state, while filters in different columns all have to match. Applying a filter returns you to the first page of results.
Clear all filters appears next to the Columns button whenever a filter is active. Use it when a filter sits on a column you have since hidden, since its filter icon is hidden along with the column. Switching to another group also clears the filters, because agents, topics, and evaluation rules are configured per group.
The table displays these columns:
| Column | Description |
|---|---|
| Audio file | File name, import/upload audio state, or source recording availability |
| Source | Call Center outbound, Import, or Upload |
| Call date | When the call took place, sortable |
| Duration | Call length, sortable |
| Agents | Contact center agents on the call |
| Topics | Detected topics shown as chips |
| Transcription | Processing status of the transcript |
| Enhancement | Processing status of transcript enhancement |
| Summarization | Processing status of the summary |
| Evaluation | Processing status of the evaluation run |
| Taxonomy fit | Whether the call fits your configured topic taxonomy |
| Last updated | When the record last changed, sortable |
Topics is hidden by default, as are the evaluation columns for rules not marked primary. Use the Columns button to bring them in.
Filtering by source
Use the Source column filter to select Call Center outbound, Import, or Upload. Close the filter popover to apply your selection. Selecting several sources includes calls from any selected source; other column filters still have to match.
Filtering by audio state
The Audio file column shows a file name once the audio is stored, and a status chip when it isn't: Queued for download, Downloading..., Download failed, or Transcript only for calls that were added as a transcript with no audio at all.
Its filter covers both. Either type part of a file name to narrow by name, or tick any of the four import/upload audio states, then click Apply. Ticking several states widens the search, so Download failed plus Transcript only returns both kinds. These audio-state filters exclude native recordings; use Source to isolate native outbound calls.
A call is in one of these states only while it has no file name, so searching a name and ticking a state at the same time returns nothing. Use one or the other.
Filtering on Download failed is the quickest way to find calls whose audio never arrived. To retry them, delete the failed calls and run the import again: re-importing on its own skips them, because the original records still count as imported.
Filtering by duration
The Duration filter takes a minimum and a maximum in seconds, and either bound can be left empty. Setting only a minimum of 600 finds every call over ten minutes; setting only a maximum of 30 surfaces the very short calls that are often misdials or abandons. Calls without a known duration are not returned by a duration filter.
Filtering by processing status
Transcription, Enhancement, Summarization, and Evaluation each filter on the same set of states: Pending, Processing, Completed, Failed, and Skipped. Filtering one of them on Failed is the fastest way to find calls that need reprocessing. A skipped step renders as an empty cell. Filtering on Skipped matches only steps explicitly marked skipped, not steps that have not started yet, which also render blank.
A step stays in Processing for as long as it's still being retried; there's no separate "timed out" state. Once retries are exhausted, the step shows as Failed.
Call Center outbound calls
Native call details show View source call, a Flow version link to the version associated with the call, and the Flow step that enrolled the recording. For published or pinned flows, the link opens the locked version used for that call. A draft version may show later saved edits; those edits do not change the call's saved execution definition. Employee names follow the linked Stellar user rather than an ingestion-time copy.
Viewing native analysis requires access to both the destination group and the source Call Center call, including its Groups restrictions. That same source boundary applies to playback, transcripts, exports, reports, and insights.
Native audio stays with the source call. In the Audio file column, Call Center recording means the audio is available; Source audio unavailable means it can no longer be played or re-transcribed. You can still rerun summaries and evaluations from the existing transcript. Audio retention is independent of derived analysis and report retention.
Studio has no action for deleting a source call for privacy reasons. On request, an administrator (through the API) or Stellar support performs privacy deletion of the source call. It removes the source call's recording, phone numbers, and transcript, and also purges the linked private analysis and any affected report or insight content.
Call Analysis uses the recording's participant annotations to combine the employee and external party into mono audio for playback and transcription. System-audio channels are excluded; the original source recording stays unchanged.
In the enhanced native outbound transcript, the employee and Called party labels follow speech activity in the original participant channels, not the transcription provider's speaker IDs. Employee turns show the contact center agent's name when available. If PII redaction is enabled or the name is unavailable, they show Human Agent instead. Turns without clear channel evidence are labeled Unknown speaker. The raw view retains the provider's original labels.
Native outbound transcript enhancement does not assume a greeting, introduction, stated purpose, verification, or fixed conversation sequence. It preserves the observed content rather than filling in an expected call script.
Evaluation columns
Every evaluation rule configured for the group gets its own column, named after the rule. Rules marked primary are visible by default and the rest are hidden, so use the Columns button to add the ones you want. Adding a rule in the group's settings makes a new column available here without any further setup.
Each column shows that rule's score for the call as a chip, and a dash when the call has no score for it. The column's filter lists exactly the score values the rule is configured with, so you can pick, say, Failed on a compliance rule to review only the calls that fell short. Selecting values on two different rules requires both to match.
Evaluation filters only match calls that have been evaluated, and calls that were never evaluated cannot be isolated from these columns today. Filtering the Evaluation column narrows to the calls whose evaluation ran and reached a given state, which is not the same set.
Next steps
- Enroll outbound Call Center calls through a Human outbound call Flow
- Import call recordings from Genesys to populate the calls list
- Configure evaluation rules to add scoring columns
- Review conversations handled by your voice agents