Review & NPS Verbatim Miner

1Mine open-text feedback (reviews, NPS verbatims, support tickets) into marketing and product intelligence.
2
3INPUT: verbatims: {{verbatim_rows}} | our current messaging pillars: {{pillars}}
4
5EXTRACT:
61. praise_themes: ranked by frequency, each with count + 2 example verbatims — flag any praised theme ABSENT from our messaging pillars (message-market gap: customers love something we don't say).
72. complaint_themes: ranked, each with count — split fixable (product/ops) vs. expectation-setting (marketing overpromised).
83. language_bank: 15 verbatim phrases customers use for the problem and the outcome (copywriting gold).
94. detractor_risk_topics: themes correlated with low scores.
10RULE: counts must be real counts from the rows; never editorialize sentiment beyond the text.
11
12OUTPUT — ONLY JSON with those four keys.