Greetings

The “Greetings” tracker detects phrases that indicate the opening of a conversation with politeness and acknowledgment—expressions like “hello,” “hi there,” “good morning,” or “how are you doing today?” By capturing these moments, teams gain insights into whether interactions start off on a friendly, warm note, enabling analyses of first-impression strategies and ensuring standards are met for a positive initial experience.

Who Benefits:

• Project Managers & Team Leads: Understand how consistently representatives start calls with friendly greetings, guiding training and scripts for welcoming interactions.
• Sales & CX Teams: Recognize if a positive initial tone correlates with improved outcomes, conversions, or customer satisfaction.
• Business Managers & Strategists: Spot patterns in greeting usage to refine interaction standards, cultural sensitivities, or brand voice approaches.
• Compliance Officers: Ensure that required greetings or disclaimers at the start of calls are properly delivered, maintaining compliance and trust.

Value Proposition:

• Positive First Impressions: Highlighting greeting moments ensures calls begin courteously, potentially improving customer mood and willingness to engage.
• Operational Insights: Structured, machine-readable output integrates with dashboards, supporting analysis of greeting consistency and impact.
• Continuous Improvement: Understanding greeting patterns leads to refinements in scripts, training materials, and representative confidence.

Data Dictionary

Objective:

Detect phrases in the conversation that indicate a greeting or welcome at the start of the call. The tracker returns occurrences of matched phrases, along with timestamps and confidence. This data helps teams evaluate how effectively representatives set a positive tone from the outset.

Proposed JSON Schema:

{
  "trackers": [
    {
      "name": "Greetings",
      "matches": [
        {
          "text": "<string>",
          "timestamp": "<ISO 8601 or relative time>",
          "confidence": <number> // 0.0 to 1.0
        }
      ]
    }
  ]
}

Data Dictionary:

If No Matches Found:
If no greeting is detected, return an empty matches array for "Greetings" or omit the tracker. If no trackers match at all, return:

{
  "trackers": []
}

Determining Confidence & Thresholds

How the Greetings Tracker Works:

• Seed the tracker with phrases commonly used at the start of a call: "hello," "hi there," "good morning," "good afternoon," "how are you doing today?"
• The system generalizes these seed phrases to identify semantically related greetings.
• Confidence indicates how strongly the detected phrase represents a greeting intent.

Confidence & Calibration:

• Initial Testing: Validate with known calls to ensure detected phrases accurately represent genuine greetings rather than unrelated positive statements.
• Refine Phrases: Add or remove seed phrases to capture cultural nuances or company-specific greetings.
• Thresholding: Downstream processes can filter low-confidence matches if necessary.
• Stakeholder Feedback: Involve CX leads or compliance officers to refine which greetings matter (e.g., required salutations for brand consistency).

Prompt Construction & Instructions

Role Specification & Reiteration:

• Present the system as a “highly experienced assistant” identifying greeting moments.
• Reiterate instructions: return only JSON, no extra commentary, follow schema.

No Hallucination:

• Only detect greetings if the transcript supports it.
• If uncertain, assign lower confidence or omit matches.

Strict Formatting:

• Return only JSON with trackers.
• If none found, return empty trackers.

Example Prompt for Implementation

System Message (Role: System):
"You are a highly experienced assistant that identifies 'Greetings' mentions in a conversation. You have:

• A transcript
• A 'Greetings' tracker defined by seed phrases indicating a polite start.

Your task:

1. Detect phrases suggesting a hello or welcome.
2. Return text, timestamp, and confidence.
3. Return only JSON as defined, no extra commentary.

Instructions & Variables (example):

{
  "tracker": {
    "name": "Greetings",
    "phrases": [
      "hello",
      "hi there",
      "good morning",
      "good afternoon",
      "how are you today"
    ]
  }
}

User Message (Role: User):
"Analyze the following transcript and detect 'Greetings' mentions:

[TRANSCRIPT_JSON][TRANSCRIPT_JSON]"

Expected Output Format:

{
  "trackers": [
    {
      "name": "Greetings",
      "matches": [
        {
          "text": "Hello, thank you for calling!",
          "timestamp": "2024-07-10T09:00:00Z",
          "confidence": 0.9
        }
      ]
    }
  ]
}