Scoring and the debrief

Two things make Revyn's scores worth trusting: they are decided per turn while you are still on the call, and the part that decides is deterministic. A language model proposes evidence; a rules layer decides what it is worth. Charm does not move the number.

#Turn scores

Every turn you speak is graded on three dimensions, and the running totals are shown on the call screen:

• Discovery. Did you learn something you did not know? Open questions, follow-ups on what the buyer said, and confirming what you heard all count. Monologues do not.
• Objection handling. When the buyer pushed back, did you acknowledge, isolate and answer, or did you talk past it?
• Momentum. Is the call moving towards the terminal? Proposing a next step, getting a small yes, and reading the buyer's mood correctly all move it forward. Repeating yourself moves it back.

Scores are cumulative across the call, so an early stumble is recoverable and a strong close is rewarded. The dimension scores also feed your skill-line progress in the practice programme.

#Coaching cues

Cues fire mid-call when the scenario expected something you did not do: a discovery question skipped before a price was quoted, a competitor mentioned and not addressed, a next step never proposed. They are short, private, and never shared with a manager. If a cue keeps appearing, it is telling you something.

#Framework scoring: MEDDPICC

Every call is also scored against MEDDPICC: Metrics, Economic buyer, Decision criteria, Decision process, Paper process, Identify pain, Champion, Competition. For each field the engine records whether it was uncovered, partially uncovered or missed, with the transcript line as evidence.

Three rules keep this honest:

1. Evidence is quoted, never inferred. A field is only marked uncovered if a specific line in the transcript establishes it.
2. Stage sets the expectation. A cold open is not expected to uncover the paper process. The engine grades against what a call at that stage should reach, so a great discovery call is not penalised for not negotiating.
3. Hard ceilings. Some misses cap the overall score regardless of everything else: no next step, no identified pain, or a price quoted before any qualification.

The same engine scores imported real calls, so practice and reality are graded on one scale.

#How commitment is decided

The buyer committing (booking the meeting, agreeing to the pilot) is not up to the language model. A server-side progression engine tracks the state of the conversation and only allows a commit when the scenario's conditions are met: enough pain surfaced, the key objection handled, a concrete next step proposed. If those conditions are not met, the buyer will not say yes however nicely you ask. Every call ends at one of five graded terminals or a turn cap.

#Reading the debrief

The debrief has six parts, in the order you should read them:

1. The turn. The single exchange that decided the call: your line, what it did to the buyer, and the line that would have landed better.
2. Outcome and terminal. Which ending the call reached and how it was graded.
3. Dimension scores with the turns that moved each one.
4. MEDDPICC with per-field evidence.
5. The transcript, annotated turn by turn, with mood changes marked.
6. The recording.

Ask Revyn can read every debrief you own. "Why did discovery score low on my last call?" is a fair question to ask it.

#Manager view

In an org, managers see aggregate scores, trends, and the framework fields a team keeps missing. They do not see coaching cues. Individual debriefs are visible to managers by default in an org; see Security and privacy.