Review Management
Use this for review analysis, reputation management, reply drafts, complaint/report drafts, product insight mining, uploaded review CSV analysis, and rating improvement strategy.
MCP workflow
- Use
app-store-connect-mcp. - Single review: read with
get_review_data. - Review list: read with
get_reviews_data. - Use
unansweredOnly,badRatingOnly, andmaxRatingto narrow review lists. - Use project context for tone, support links, known issues, and response policy.
- Analyze review health, themes, response opportunities, and complaint/report eligibility.
- Validate with
validate_ai_companion_datatargetrevieworreviewList. - If asked to apply, fill drafts with
fill_revieworfill_reviews.
Fill fields
Single review:
responseBodycomplaintReasonconcernType
Review list:
reviews: array or map byreviewId, each withresponseBody,complaintReason, and/orconcernType
CSV review analysis
Use this mode when the user uploads or references a CSV with all reviews. CSV analysis does not require MCP unless the user wants to fill replies/complaints back into ASO.dev.
Before analysis:
- Identify the CSV columns. Common columns: review ID, rating, title, body/text, locale/country, app version, date, developer response, response date, device/platform.
- Normalize dates, ratings, locales, versions, and empty values.
- Deduplicate repeated rows by review ID or exact title/body/date.
- Segment by rating, locale/country, app version, date period, answered/unanswered status, and review length.
- If the CSV is large, sample examples for quotes but compute counts/themes across the full file.
Analyze for:
- Sentiment by rating and text: positive, neutral, negative, mixed.
- Themes: bugs/crashes, UX friction, pricing/paywall, onboarding, performance, account/login, localization, support, content quality.
- Feature requests: missing capabilities, repeated "wish it had..." patterns, competitor comparisons.
- Product risk: regressions by app version/date, severe bugs, refund/cancel intent, trust/privacy concerns.
- ASO language: phrases users use to describe value, outcomes, and use cases.
- Support opportunities: unanswered low-star reviews, stale responses, high-impact replies.
CSV output should include counts and representative examples. Do not paste large raw review dumps.
Review analysis framework
Classify themes:
| Theme | What it means | Action |
|---|---|---|
| Bugs/crashes | Technical failure, broken flow, regression | Escalate fix, respond with known status only |
| Feature requests | Users ask for missing capability | Track frequency and user segment |
| UX complaints | Confusing, slow, frustrating, hard to use | Prioritize UX/product review |
| Pricing/paywall | Too expensive, unclear subscription, trial confusion | Use monetization-strategy if repeated |
| Praise | Value, delight, favorite features | Thank and learn conversion language |
| Competitor mentions | Comparisons or switching reasons | Mine positioning gaps |
Review health signals:
- Average rating and trend, if available
- Recent low-star volume
- Unanswered negative reviews
- Response rate and response freshness
- Repeated bug/UX/pricing themes
- Complaint/report candidates
Reply rules
- Be calm, specific, and brief.
- Prioritize timely replies; aim for 24-48 hours for high-impact negative reviews when operationally possible.
- Do not argue with the reviewer.
- Do not reveal internal data.
- Do not promise fixes unless they are confirmed.
- Do not ask for personal information in public.
- For angry reviews, acknowledge the issue and move to support.
- Do not ask the user to change their rating.
- Do not copy-paste identical replies across many reviews.
Response framework
Use HEAR for negative reviews:
- Hear: acknowledge the specific issue.
- Empathize: show you understand the frustration.
- Act: explain what is known, fixed, or being investigated.
- Resolve: invite direct support with the right contact path.
Templates are starting points only:
- Bug: acknowledge, mention known fixed version only if confirmed, invite update/support.
- Feature request: thank them, say it is shared with the team, avoid promising roadmap.
- Vague negative: apologize briefly, ask for details through support.
- Positive: thank them and echo the benefit they named.
Complaint/report rules
- Report only when the review appears abusive, spam, fraudulent, irrelevant, or otherwise policy-violating.
- Do not report valid negative feedback just because it is low-rated.
- Keep complaint reasons factual and short.
Rating prompt strategy
Use for advice only; this MCP skill fills review drafts, not app code.
- Trigger only after a positive value moment.
- Avoid first session, onboarding, errors, crashes, failed payment, or support frustration.
- Good moments: completed task, achieved goal, streak, saved time/money, successful restore/export/share.
- Respect platform limits and do not incentivize reviews.
- If bad reviews are caused by product issues, fix the issue before increasing prompt volume.
Product insight mining
Use reviews to extract:
- Most loved features and exact user language for metadata/screenshots.
- Top recurring bugs or UX failures.
- Monetization objections and subscription confusion.
- Feature demand frequency.
- Competitors mentioned and switching triggers.
Feature request scoring:
| Signal | Meaning |
|---|---|
| Frequency | How many reviews mention the request |
| Rating impact | Whether request appears in low-star or churn-risk reviews |
| Segment | Locale, version, user type, or country where it appears |
| Strategic fit | Whether it matches app positioning and roadmap |
| Effort/risk | Whether it needs product, backend, design, pricing, or policy work |
Output
# Review Health and Response Work
App:
Mode:
Rating/trend:
Filters:
CSV source:
## Health summary
- Total loaded reviews:
- Low-star reviews:
- Unanswered reviews:
- Main risk:
## Sentiment
| Segment | Positive | Neutral | Negative | Mixed | Notes |
|---|---:|---:|---:|---:|---|
## Themes
| Theme | Count | Rating impact | Evidence | Action |
|---|---:|---|---|---|
## Missing features / requests
| Request | Count | User segment | Evidence | Product recommendation |
|---|---:|---|---|---|
## Drafts
| Review ID | Rating | Theme | Action | Draft |
|---|---:|---|---|---|
## Product insights
| Insight | Evidence | Recommended owner/action |
|---|---|---|
## Rating improvement plan
1.
2.
3.
## Validation
- CSV columns mapped:
- Counts based on full data:
- Tone safe:
- No private data:
- Complaint justified:
- MCP validation:
- Ready for fill: