Skip to main content

Evaluation Criteria

Zild Coach evaluates conversations using configurable criteria. Criteria can be structured into categories aligned with operational goals.

Core Evaluation Categories

1. Accuracy

  • Was the response factually correct?
  • Did the agent use correct knowledge?
  • Was structured data captured correctly?

2. Instruction Adherence

  • Did the agent follow defined policies?
  • Were prohibited actions avoided?
  • Was escalation triggered appropriately?

3. Resolution Quality

  • Was the issue fully resolved?
  • Was next step clearly defined?
  • Was user intent satisfied?

4. Communication Quality

  • Tone consistency
  • Clarity of explanation
  • Conciseness
  • Professional language

5. Escalation Handling

  • Was escalation necessary?
  • Was escalation triggered at the right time?
  • Was sufficient context passed to human?

Scoring Model

Each category may include:
  • Weighted score
  • Pass / fail criteria
  • Threshold triggers
  • Compliance flags
Example scoring structure:
  • Accuracy: 30%
  • Resolution: 30%
  • Communication: 20%
  • Policy adherence: 20%
Weights can be customized per Tenant.

Policy Violation Detection

Zild Coach can flag:
  • Unauthorized data disclosure
  • Policy non-compliance
  • Improper escalation
  • Incomplete information capture
Violations can trigger alerts or workflow actions.

Custom Criteria

Enterprises can define:
  • Industry-specific compliance checks
  • Sales qualification standards
  • Internal SLA benchmarks
  • Brand tone enforcement
Evaluation criteria should align with business objectives.

Aggregated Performance Analysis

Scores can be aggregated by:
  • Agent
  • Group
  • Channel
  • Time period
This enables:
  • Trend analysis
  • Team comparison
  • Performance benchmarking
  • Continuous improvement tracking

Summary

Zild Coach provides a structured evaluation framework that ensures:
  • AI quality control
  • Human performance monitoring
  • Compliance enforcement
  • Continuous operational improvement
Evaluation criteria convert conversations into measurable performance standards.