How AI evaluations work, scoring criteria, behaviour tracking, and interpreting your results
After every practice session, Oliver AI generates a detailed evaluation of your performance. Evaluations measure how well you demonstrated key sales behaviours during the conversation.
Oliver AI uses two complementary evaluation systems. Understanding both helps you get the most out of your performance data.
Behaviour rubrics are detailed, criteria-based evaluations tied to specific sales skills. The platform includes eight built-in behaviours:
Each behaviour is scored on a 1--5 rubric scale (Novice to Expert) with weighted evaluation criteria. For example, Objection Handling is evaluated on acknowledgment, clarification, value response, confirmation, and next steps -- each carrying a specific weight.
For full details on each behaviour's rubric and criteria, see Sales Behaviours.
Call evaluations provide a broader, overall assessment of your performance across the entire call. These scores are stored as part of the call evaluation record and include:
| Metric | Description |
|---|---|
| Overall Score | Your aggregate performance rating for the call |
| Communication Clarity | How clearly and effectively you communicated your message |
| Product Knowledge | How well you demonstrated understanding of the product |
| Objection Handling | How effectively you addressed concerns raised during the call |
| Engagement | How well you kept the customer engaged throughout the conversation |
| Closing Effectiveness | How successfully you guided the conversation toward a commitment |
| Customer Satisfaction | The predicted satisfaction level of the customer based on the interaction |
These two systems work together: behaviour rubrics give you detailed, skill-by-skill coaching, while call evaluations give you a high-level snapshot of your overall performance on each call.
In addition to the two evaluation systems above, every practice session is automatically scored against the SPICED discovery framework. SPICED evaluates five dimensions of your discovery conversation: Situation, Pain, Impact, Critical Event, and Decision.
Unlike the 1--5 rubric scale, SPICED uses binary pass/fail questions (11 total across the five dimensions) and produces a 0--100 score per dimension and overall. This gives you a framework-specific view of how thoroughly you explored the prospect's situation during the call.
SPICED results appear in a dedicated SPICED tab on the call results page and as a card on your practice history detail pages. Your dashboard also shows SPICED radar charts and trend lines once you have enough session data.
For full details, see SPICED Discovery Scorecard.
Each behaviour is scored on a 1-5 scale:
After each practice session, you are taken directly to the Call Results page. This is a two-panel layout that shows your transcript alongside detailed analysis.
The left side of the screen displays the full conversation transcript in a chat-bubble style:
Scroll through the transcript to revisit specific moments in the conversation.
The right side contains four tabs: Feedback, SPICED, Analytics, and Objections.
The Feedback tab is your primary performance view.
Score Ring -- A large animated ring displays your overall score out of 100:
Score Breakdown -- A scorecard lists each evaluation criterion with a progress bar and score. Click any criterion to expand it and see:
Coaching Cards -- Three colour-coded cards appear below the scorecard:
The Analytics tab shows conversation metrics that help you understand your communication patterns:
| Metric | Description |
|---|---|
| Talk / Listen Ratio | A visual bar showing what percentage of the conversation you spoke vs. the customer. If you talked more than 70%, you will see a warning. |
| Filler Words | Count of filler words used (um, uh, like, you know) |
| Talk Speed | Your average words per minute |
| Longest Monologue | The longest uninterrupted stretch of speaking (in seconds) |
| Questions Asked | How many questions you asked during the call |
The Objections tab lists every customer objection detected during the call:
If no objections were detected, the tab shows an empty state.
The Call Results page includes action buttons that vary by role:
For employees:
For managers:
For owners and admins reviewing a rep or location result, Chat with Coach opens the role-aware Coach panel with language focused on team patterns, location signals, setup gaps, or organization priorities rather than treating the call as your own.
Note: While the evaluation is being generated, a processing banner appears at the top of the results page. The analysis typically completes within a few seconds.
Managers can access evaluations for all team members:
Note: Evaluation scores are generated by AI and should be used as coaching guides, not absolute performance measures. The value comes from identifying patterns and tracking improvement over time.