返回提示词目录第 1402 页

应用商店评论分析与改进归纳

按版本、平台和评分归纳评论主题,区分发布故障与长期诉求,输出回复思路及团队行动项。

完整提示词

<context>
You analyse public app store reviews. They have their own quirks. The rating a store shows mixes old and new reviews, while what matters is the trend by app version. In-app rating prompts shape who reviews: a prompt after a success moment brings a wave of short five-star reviews, and removing it makes the average drop with nothing else changed. One-star spikes usually follow a release (crashes, a forced login, a removed feature, a price change) and fade if fixed, while lasting complaints (missing features, pricing, ads) persist across versions. Platforms differ: the same feature can work on iOS and break on some Android devices. Reviews in other languages hold country-specific issues (payments, localisation). And reviewers are a small, self-selected group, so reviews show what bothers people, not how common it is.
</context>

<task>
<reviews>
{{reviews}}
</reviews>

<app_context>
{{app_context}}
</app_context>

1. Rating picture: average rating and count by app version and platform, and the trend. Note any change in rating prompts that would explain shifts.
2. Themes by rating: themes in one-to-two-star, three-star and four-to-five-star reviews, with counts and short quotes.
3. Release issues versus lasting complaints: for each negative theme, whether it starts with a specific version (release issue) or appears across versions (lasting complaint).
4. Platform and country differences: themes concentrated on one platform, device mention or country.
5. Reviews worth a reply: reviews where a reply could help (bug now fixed, misunderstanding, a workaround exists, a support route is needed), with a reply approach. Avoid replies that argue or promise features.
6. Actions: for engineering (release bugs with version and platform), for product (lasting complaints), and for the rating prompt if it is badly timed.
</task>

<constraints>
- Do not treat review counts as prevalence among all users; recommend checking crash data or analytics for release issues.
- Quote only text given and keep quotes short; no reviewer names.
- Do not suggest asking only happy users to review, offering incentives for ratings or other practices that break store rules or mislead.
- If version or date is missing, say which conclusions that prevents.
{{> output/uncertainty}}
</constraints>

<output_format>
## Rating picture
Table: version | platform | reviews | average rating.

## Themes by rating
Table: rating band | theme | count | quote.

## Release issues versus lasting complaints
Table: theme | first version seen | across versions? | type.

## Platform and country differences
Bullets.

## Reviews worth a reply
Table: review summary | reason to reply | reply approach.

## Actions
Grouped by engineering, product, rating prompt.

## Questions
What to confirm.
</output_format>

可替换内容

  • {{reviews}}
  • {{app_context}}
  • {{> output/uncertainty}}

分类与标签

办公效率、应用商店评论、用户反馈分析、版本问题、产品改进

前往 AI 图片创作