How Savo Turns Conversations into Insights
From spoken words to scored dimensions — the pipeline that makes signal visible
A Signal Session produces a transcript. That transcript is the raw material — but the insights in your report are the result of a structured scoring process that happens after the conversation ends. Here's how it works.
How the Analysis Works
After a Signal Session completes, Savo automatically analyzes the transcript to surface the insights in your report.
The first step is identifying what's meaningful. Not every statement in a conversation is useful signal — Savo filters the transcript for moments that actually speak to the research topic. The moments that make the cut become Evidence Units: specific, citable excerpts from the conversation. Evidence Units are classified by quality:
- Story — a rich narrative account that provides full context
- Example — a concrete specific instance without full narrative
- Claim — a direct assertion without supporting context
Stronger evidence carries more weight in the scoring.
The second step is scoring. Each dimension in your Signal Event has a rubric that defines what strong, moderate, and weak evidence looks like. Savo evaluates the Evidence Units for each dimension and produces a score on the 1–5 scale, along with a confidence rating (High / Medium / Low) reflecting how much quality evidence was available. When evidence is insufficient to score a dimension reliably, the system returns ""No Score"" rather than a low-confidence estimate — a gap in evidence is itself meaningful data.
Where the Results Go
Scored dimensions, confidence ratings, and Evidence Units flow into your Insights report. The report organizes findings across four views: Summary (the big picture), Deep Dive (dimension-level detail), Evidence Explorer (the citable moments behind each finding), and Survey (if your event included a survey component).
Each finding in your report traces back to specific Evidence Units — and each Evidence Unit traces back to a specific moment in a specific session. Nothing in the report is opaque.
See Understanding Dimension Scores and Confidence Levels and Drilling Down into Evidence for more on reading your results.