Feature
Coach from the transcript, not memory.
After a call is transcribed, Fresno scores how much the rep talked versus the prospect and pulls out objections with a category and a confidence score. Low-confidence results stay out of the way.
What actually gets measured
Coaching runs on the call transcript. It does not invent a score from silence or a manager's notes.
Rep vs prospect ratio
Each analysis returns the share of talk time for the rep and for the prospect, plus a confidence score. You can see whether a call was a monologue or a conversation.
Detected, categorized, scored
Objections come back as the quote, a category, and a confidence. Filter to high-confidence hits (0.8+) so 1:1s start from things the model is actually sure about.
Typed decode or it fails
LLM output is decoded through a typed schema. A bad response is an error — it does not get written as coaching data.
Signals across calls
Call records already roll up per rep (volume and deal-score). Talk-time and objection totals can be merged in once per-call analysis is stored — they are not invented from call count.
Same call, more than coaching
Calls already record and transcribe in Fresno. Coaching is another read of that transcript — next to synced calendar and email activity, not a separate product.