Skip to content

Conversation Analytics

Conversation Analytics is the MiaRec product that analyzes every conversation automatically. This page defines each analytics concept and tells you where its results appear. For hands-on usage, see the User Guide; for configuration, see the Administration Guide.

Transcript

The transcript is the text of a conversation, produced automatically by speech-to-text. Transcripts are speaker-separated — you can see who said what — and each phrase carries a timestamp, so clicking text jumps the audio player to that moment. Conversations can be transcribed in many languages.

The transcript is the foundation for all other analytics: summaries, sentiment, topics, insights, and Auto QA are all derived from it.

Where it appears: the Transcript tab of the conversation detail page. Transcript text is also searchable through advanced search, and can be downloaded. See Reading transcripts.

Summary

The summary is a short AI-written digest of the conversation — typically the reason for the call, what was discussed, actions taken, the outcome, and next steps. Authorized users can edit a summary if the AI missed something.

Where it appears: at the top of the conversation detail page; summaries are also searchable. See Conversation summaries.

Sentiment score

The sentiment score expresses the emotional tone of a conversation as a number from -100 (most negative) to +100 (most positive). Each conversation receives three scores — overall, agent, and customer — so you can distinguish an upset customer from an unprofessional agent. Scores map to five labels, from Very Negative to Very Positive.

Every score comes with a written explanation of why it was assigned, and phrases that contributed to a negative score are highlighted in the transcript as evidence.

Where it appears: on the conversation detail page, on the Sentiment dashboard tabs (including a heatmap for spotting negative-conversation patterns over time), in search filters, and as report columns. See Sentiment.

Topics

A topic is a named subject your organization wants to detect in conversations — for example "Service cancellation", "Billing question", or "Competitor mentioned". Your organization defines its own topic list; the AI reads each transcript and identifies which topics were discussed, based on the topic's description. One conversation can match multiple topics.

Topics turn thousands of individual conversations into trends: which subjects drive call volume, which are growing, and which correlate with negative sentiment.

Where it appears: the Topics dashboard tab shows topic distribution and trends; topics are also available as search filters. See Topics in the User Guide, and Configuring topics in the Administration Guide.

Insights and AI tasks

An insight is any specific piece of information the AI extracts from a conversation — the reason for the call, the product discussed, a satisfaction score, a missed sales opportunity. Insights are produced by AI tasks: configured instructions (prompts) that tell the AI what to look for and where to store the result. Summaries, sentiment scores, topics, and Auto QA results are all produced by AI tasks; your organization can also define fully custom insights.

Insight values can include the AI's explanation of how it reached the answer, often with supporting quotes from the transcript, so reviewers can trust — and verify — the result.

Where it appears: the Analytics tab of the conversation detail page groups the insight values for that conversation. See AI insights in the User Guide and Creating custom insights in the Administration Guide.

Custom fields

A custom field is a data field your organization adds to conversations — for example "Call Reason", "Product", or "CSAT". Custom fields can be filled in manually by users, populated automatically by your telephony integration, or written by AI tasks. They are searchable, can appear as columns in conversation lists and reports, and numeric fields can be shown as dashboard metrics.

Where it appears: on the conversation detail page and as search, report, and dashboard dimensions. See Custom fields in the Administration Guide.

CX metrics (CSAT, NPS, NES)

CX metrics are customer-experience scores — such as customer satisfaction (CSAT), Net Promoter Score (NPS), or NES — estimated by AI for each conversation. In MiaRec, a CX metric is a configured insight: an AI task scores the conversation and writes the value into a custom field, together with the AI's reasoning. Your organization chooses which metrics to measure and how they are defined.

Where it appears: on the conversation's Analytics tab and as dedicated dashboard tabs showing the metric across your organization. See CX metric dashboards.

Entity recognition and redaction

Entity recognition detects specific kinds of information inside transcripts — names, organizations, dates, amounts, card numbers, and similar entities.

Redaction removes sensitive information from conversations: matched content is masked in the transcript and replaced with silence in the audio. Organizations use redaction to keep payment card data, personal identifiers, and other sensitive details out of stored conversations. Redaction rules are defined by your administrator and can use recognized entities as detection patterns.

Where it appears: redacted words show as masked text in transcripts, with the matching audio silenced. See Data redaction in the Administration Guide.