Skip to main content
Salesgraph Assist provides real-time coaching metrics during recordings and a comprehensive coaching analysis after each meeting.

​
Live Metrics (During Recording)

While recording, the widget displays three key metrics in a compact grid, updated as the transcript grows:
MetricDescription
Talk RatioPercentage of speaking time attributed to you (the primary speaker). Helps you balance listening vs. talking.
Filler WordsCount of filler words and phrases detected in your speech.
QuestionsNumber of questions you’ve asked during the conversation.
The “primary speaker” is determined as the first speaker identified in the transcript.

​
Filler Word Detection

The coaching analyzer detects 12 common filler words and phrases:
  • um
  • uh
  • like
  • so
  • actually
  • basically
  • literally
  • right
Detection is case-insensitive and operates on the full transcript text.

​
Post-Meeting Analysis

After a recording is processed, the Coaching tab in the Meeting Detail view shows an expanded set of metrics:
MetricDescription
Talk RatioYour percentage of total speaking time
Filler Word CountTotal filler words detected in your speech
Speaking Speed (WPM)Words per minute, calculated from your total word count and speaking duration
Interactivity ScoreA percentage reflecting how balanced the conversation was — higher scores indicate more back-and-forth dialogue

​
AI Coaching Feedback

In addition to quantitative metrics, the server generates AI-powered coaching feedback using an LLM. This feedback appears as a text block below the metrics grid and provides personalized suggestions based on the full meeting transcript — covering areas like question technique, active listening, objection handling, and conversation flow.

​
How Metrics Are Calculated

  • Talk ratio — Sum of your segment durations divided by total meeting duration.
  • Filler words — Regex-based detection across all your transcript segments.
  • Speaking speed — Total words spoken divided by total speaking duration in minutes.
  • Interactivity — Based on the frequency of speaker transitions. More speaker switches = higher interactivity.
  • Longest monologue — Duration of your longest uninterrupted speaking segment.
  • Questions — Count of sentences ending with a question mark in your segments.