NpsFeedbackInsightSkill

---
name: NpsFeedbackInsightSkill
description: >
  Turn NPS responses into prioritized customer insights and recommended actions for product and service teams.
version: 1.0.1
visibility: organization
tags:
  - customer-experience
  - analytics
entrypoint:
  command: analyze_nps
  label: Generate NPS insight brief
---

## When Claude should load this skill
- Use during recurring NPS reviews, quarterly business reviews, or when leadership requests customer voice summaries.
- Helpful whenever response volume is high and segmentation by persona, product tier, or geography is required.

## Core workflow
1. Reference `instructions/nps-analysis.md` for scoring buckets, sentiment labeling, and follow-up playbooks.
2. Run `python scripts/nps_insight.py` to emit `NPS_CLUSTERS` capturing themes, representative quotes, and average scores.
3. Fill `templates/insight-summary.md` with prioritized themes, owners, and next-step experiments aligned to segments.

## Linked resources
- `data/nps_responses.csv` — Raw response exports with scores, comments, and timestamps.
- `data/customer_segments.csv` — Mapping table for industry, ARR tier, lifecycle stage, and customer success owner.
- `instructions/nps-analysis.md` — Guidance on cluster naming, urgency scoring, and escalation triggers.
- `templates/insight-summary.md` — Reporting layout for sharing insights with product, support, and marketing.
- `playbooks/closed-loop.md` — Optional workflow for assigning follow-up actions back to CSMs.

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.cluster import KMeans

responses = pd.read_csv("data/nps_responses.csv")
vectorizer = TfidfVectorizer(max_features=500, stop_words="english")
X = vectorizer.fit_transform(responses["comment"].fillna(""))

cluster_count = 5
model = KMeans(n_clusters=cluster_count, random_state=42)
labels = model.fit_predict(X)
responses["cluster"] = labels

insights = (
    responses.groupby("cluster")
    .agg({"score": "mean", "comment": lambda x: list(x)[:3]})
    .reset_index()
)

print("NPS_CLUSTERS=", insights.to_dict("records"))