Skip to content

NPS (Net Promoter Score)

NPS is traditionally based on a survey question (“How likely are you to recommend us to a friend or colleague?” on a 0–10 scale). In Conversation Analytics, NPS is often implemented as an AI-estimated proxy based on the conversation content.

Recommendation: be explicit with stakeholders that this is AI-estimated NPS, unless you are extracting an explicitly stated rating from the conversation.

Common NPS implementation options

Choose one approach and document it internally to avoid confusion:

  1. Extract explicit NPS rating
    Use when agents ask the NPS question and customers answer with a number.
  2. Infer NPS from sentiment + outcome (proxy)
    Use when no explicit question is asked; AI estimates likely rating.
  3. Hybrid
    Extract explicit rating when present; otherwise return “Unknown”.

Recommendation: Use Hybrid approach—extract explicit ratings when customers state them, otherwise return "Unknown". This avoids inventing numbers when NPS is not discussed in the conversation.

Configuration overview (tenant view)

  • Create/verify a numeric Custom Field (0–10) or dropdown (Promoter/Passive/Detractor)
  • Enable the prebuilt NPS AI Task (if available) or create a tenant task
  • Configure thresholds/buckets:
  • 0–6 = Detractor
  • 7–8 = Passive
  • 9–10 = Promoter
  • Test and validate on representative conversations

Where NPS appears

  • Conversation Details → Analytics
  • Dashboards (distribution by bucket)
  • Search (e.g., NPS <= 6)

Prompt recommendations

  • Prefer JSON output
  • Include “Unknown” when evidence is insufficient
  • Provide a short explanation