---
name: customer-review-theme-analysis
description: Clusters customer reviews into evidenced themes, separates praise from complaints, quantifies sentiment and product areas, and surfaces representative quotes with source identifiers. Use for review mining, voice-of-customer reporting, product feedback synthesis, or reputation monitoring.
license: Apache-2.0
metadata:
  adlass.categories: "customer-support/voice-of-customer"
  adlass.industries: ""
  adlass.tags: "customer-reviews,voice-of-customer,themes,sentiment,product-feedback"
  adlass.adaptation: "mapping"
  adlass.source: "original"
  adlass.version: "1"
---

# Customer review theme analysis

## Purpose

Turn free-text customer reviews into a stable set of themes that product, support, and marketing can act on. The analysis preserves the original wording and shows how often each theme occurs.

## Scope

Analyze the supplied review corpus for the selected period, channels, products, and languages. Code topic, sentiment, rating context, recurrence, and representative evidence while retaining review-level traceability.

**Excluded:** identifying reviewers, responding to reviews, proving causation, weighting an unprovided sample, or treating one anecdote as a population estimate.

## Data basis

- Review table columns `review_id`, `date`, `rating`, `channel`, `product`, `text`, `language`, and `verified`.
- Product taxonomy and sentiment or theme coding guide.
- Prior theme register with `theme_id`, definition, owner, and trend status.
- Customer segment and response-policy references.

## Result

A theme analysis report, coded review sheet, and prioritized issue register.

## Quality criteria

- Every coded review retains `review_id`, date, rating, channel, and source text citation.
- Theme definitions are mutually distinguishable and include inclusion and exclusion examples.
- Counts reconcile to the review population after language, duplicate, and spam dispositions.
- Sentiment is reported as coded polarity, not as a psychological claim.
- Priority combines stated business criteria with volume and severity evidence.

## Instructions

Deduplicate only when review text, author marker, and timestamp support the match; keep uncertain duplicates separate. Translate only when a translation source is supplied and preserve the original language. A review can receive multiple themes, but the report must state whether counts are review-level or mention-level. Quote only the minimum representative text and mask personal data. Do not rank a theme without naming its denominator and priority rule.

## Adapt before use

- Map review fields and product taxonomy to the source table.
- Set language handling and sentiment labels.
- Define severity, volume, recency, and ownership rules.
- Add prior theme register and privacy requirements.