AI Text Analysis for Surveys, Reviews & Feedback
Text Response Hub
undefined Customer Experience Analysis 2025
AI-based synthesis of verified Trustpilot reviews (122 responses)
Uncovering the sources of customer happiness and dissatisfaction
Executive Summary & Next Steps
Customer Sentiment Performance:
Beats 99.3% of benchmark datasets for positivity
What delights
- Positive impact on sleep
- Comfortable and soft feel
- Effective in blocking light
What to tune
- Mixed feedback on weight
- Issues with fit or slipping off
Suggested Next steps
- Add clear fit/adjustment guidance on PDP + packaging; align claims with real use.
- Prototype an adjustability tweak/accessory; monitor ‘slipping’ mentions post-release.
- Gauge demand for weight/design variants via a quick survey or waitlist poll.
Snapshot: Sep 30, 2025
See Data Capsule. for scope & source.
Sentiment Breakdown
What this shows: Positivity of review text
Average sentiment score (0-10) across all reviews, showing percentile rank.
This is a preview. The full data and complete insights are available in the full report.
This is a preview. The full data and complete insights are available in the full report.
Global Themes & Frequencies
What this shows: Share of reviews that mention each theme (n = 122).
Ranked themes (all mentions):
Theme frequency across reviews (multi-label; totals can exceed 100%; n = 122). See Data Capsule
Voice of the Customer
Full reviews are not available in this preview version.
However, click any active Show details button above to see a subset of review data.
Data Capsule
Source: Trustpilot ()•Coverage: 122 reviews•Snapshot: Sep 30, 2025•Review dates: May 24, 2023 → Sep 24, 2025•Metrics: sentiment (0–10), theme frequencies•Details in Methodology
Methodology
- Preprocessing: De-duplication (exact/near duplicate detection), language filtering (English), PII redaction, light normalization (lowercasing, punctuation).
- Sentiment: A model assigns polarity per review; we map this to 0–10 (higher = more positive) and compute the mean across reviews. Outliers are lightly clipped; no length weighting (long reviews don’t count more than short ones).
- Themes: We embed review text, cluster semantically similar snippets, and label clusters with an LLM. Reviews can belong to multiple themes (multi-label). Frequencies are the share of reviews mentioning each theme.
- Benchmarking: “Top 0.7%” is a percentile rank vs our internal Trustpilot corpus of brand–review datasets from a comparable period; percentiles refresh on a set cadence.
- Reliability: We report n (sample size), show distributions where helpful, and note biases (public-review skew, theme overlap). Confidence intervals for the mean sentiment are available upon request.
- Limitations: AI models can miss nuance; sarcasm/idioms may be imperfectly scored; clustering can drift as new data arrives; this is a snapshot at the stated date.
Prepared by Text Response Hub
About
Text Response Hub turns free-text feedback (surveys, reviews, support logs) into clear, actionable insights.
Metadata
Version: Text Response Hub v.1.02 • Snapshot date: Sep 30, 2025 • Corpus: Trustpilot ()
Disclaimer
This analysis is intended for internal decision support. It summarizes patterns in review text using AI methods and may miss nuance; interpretations are subjective. Results are provided “as is,” without warranties, and should be evaluated alongside other evidence. See Data Capsule and Methodology for scope, assumptions, and limitations.
Contact
joel@textresponsehub.com • textresponsehub.com
© 2026 Text Response Hub. All trademarks belong to their respective owners.