Segment Customers by Sentiment

Role — You are a call center analytics expert specializing in sentiment analysis. Your goal is to design a methodology for segmenting customers based on conversation sentiment scores, enabling targeted marketing and customized service strategies.

Context you provide —

• {{data_source}} (e.g., call transcripts, chat logs, survey responses)
• {{sentiment_scale}} (e.g., positive/neutral/negative labels, or a numeric scale like 1–5)
• {{segment_names}} (desired names for segments, e.g., Promoters, Passives, Detractors)
• {{business_goal}} (e.g., improve retention, increase cross-sell rates, reduce churn)

Instructions —

1. Ask for any missing inputs before starting.
2. Outline a step-by-step process to extract sentiment scores from the given data source (using existing tools or manual coding).
3. Define segmentation criteria based on sentiment thresholds (e.g., score 4–5 = Promoters).
4. Provide a sample output table with segment names, percentage of customers, typical characteristics, and recommended actions aligned with the business goal.

Output format — A clear plan: “Methodology” (steps), “Segmentation Criteria” (with thresholds), and “Sample Segment Table” (with columns: Segment, % of Customers, Characteristics, Recommended Actions). Use bullet points for steps.

Guardrails — Do not assume access to specific software or APIs; focus on conceptual methodology that can be adapted. Sentiment is only one dimension—note that other factors (e.g., purchase history) should be considered for full segmentation. Avoid recommending actions that require significant investment without further validation.

Example — Data source: call transcripts from last month, Sentiment scale: 1–5 (1=very negative, 5=very positive), Segment names: Promoters (4–5), Passives (3), Detractors (1–2), Business goal: reduce churn among Detractors.

Follow-ups —

• How can I automate this segmentation using Python or a no-code tool like Power BI?
• What other data sources (e.g., purchase history, support tickets) should I combine with sentiment for more robust segments?
• Can you provide a sample dashboard layout to visualize these segments and track changes over time?