Find mismatches between ratings and text

Spot deceptive 4 and 5-star reviews that contain serious product bugs, billing complaints, or hidden cancellation risks.

You getDiscrepancy Summary, high-Priority Mismatches and action Items

Prompt

Prompt
You are a quality assurance and user research analyst. Examine the paired review ratings and customer comments provided below to find severe sentiment discrepancies.

- Flag false positives: reviews with 4 or 5 stars that report critical bugs, poor support, or billing issues.
- Flag false negatives: low-star reviews (1-2 stars) where the text indicates satisfaction or user error rather than product failure.
- Categorize each flagged comment by underlying issue (for example: Billing, Performance, Missing Feature, Support).
- State what immediate risk each mismatch represents for product retention.

Output format:
- Discrepancy Summary (counts of positive-score complaints vs negative-score compliments)
- High-Priority Mismatches (table with columns: Star Rating, Quoted Text, Actual Sentiment, Root Issue)
- Action Items (3 immediate steps to address flagged accounts)

Here are the reviews:

Example output

Example for: 15 app store and in-app feedback submissions

Discrepancy Summary

  • High ratings with critical complaints: 3 instances
  • Low ratings with positive sentiment: 1 instance
  • Total reviews requiring manual review: 4 out of 15

High-Priority Mismatches

Star RatingQuoted TextActual SentimentRoot Issue
5 Stars"Love the tool daily but the latest export bug corrupted our client invoice files."Highly FrustratedCritical Bug / Data Loss
4 Stars"Great UI, but support ignored my billing double-charge ticket for two weeks."Disappointed / Churn RiskBilling Support
5 Stars"Solid software, though we will have to cancel next month if SSO is not added."Conditional / Flight RiskMissing Enterprise Feature
1 Star"Awesome team and great app, just accidentally tapped one star on my phone!"Highly PositiveRating Error

Action Items

  • Contact the customer experiencing corrupted exports immediately to offer manual recovery and alert engineering.
  • Route the billing issue to finance leadership to issue a credit before renewal.
  • Reach out to the user who misclicked a 1-star review to ask if they would update their rating.
Written by AI from this prompt. Your results will vary.

Paste paired survey responses (numerical rating alongside written feedback) into Gemini, Claude, or ChatGPT. Ask the model afterward to draft a polite follow-up email template for the most critical customer.

Make it repeatable

Keep using this prompt without pasting it in every time.

Keep exploring

More in Data Analysis