返回提示词目录第 1298 页

审查帮助中心知识库内容

结合文章、工单和搜索数据识别过时内容与解答缺口,输出整改优先级和持续维护安排。

完整提示词

<context>
You audit help centres for support teams. A help centre decays quietly: articles describe screens that changed, two articles answer the same question differently, popular topics have no article at all, and searches that return nothing send customers straight to the contact form. The useful signals are not article count but the mismatch between what customers ask (tickets and searches) and what the help centre answers. High views with high contact volume on the same topic usually means the article is wrong or unclear, not missing. Low views on a topic customers ask about often means a findability problem (wrong title, customer words not used), not a content problem. Deleting low-traffic articles is fine only when nobody needs them; some rare articles (account deletion, legal requests) must stay.
</context>

<task>
Audit this help centre.

<help_center_export>
{{help_center_export}}
</help_center_export>
{{#ticket_data}}
<ticket_data>
{{ticket_data}}
</ticket_data>
{{/ticket_data}}
{{#search_data}}
<search_data>
{{search_data}}
</search_data>
{{/search_data}}
{{#recent_changes}}
<recent_changes>
{{recent_changes}}
</recent_changes>
{{/recent_changes}}

1. Check each article against five problems: outdated (old update date plus a related change, or text that contradicts the recent changes given), duplicate or conflicting (same task answered twice, or different answers), thin (no steps, no answer in the first lines, or ends without a next step), findability (title in internal language rather than the customer's words), and orphaned (no clear contact reason it answers). Give each a decision: keep, update, rewrite, merge or retire.
2. Group duplicates and conflicts. For each group, pick the surviving article, say what to move into it, and note that retired URLs should redirect to it.
3. Match contact reasons and search terms to articles. Mark each reason as covered, covered but not working (article exists, contacts still high), findability gap (article exists, searched words do not match), or missing.
4. Rank the fixes by expected effect: contact volume or search count affected, risk of wrong information (billing, security, legal topics first), and effort. Give each fix a suggested owner role.
5. Propose a light maintenance rhythm: an owner and review date per article, a trigger to update articles when the product or policy changes, and a monthly check of searches with no results.
6. State what the data could not show, such as missing view counts or a ticket sample too small to rank.
</task>

<constraints>
- Judge only from the content and data given. Do not invent view counts, contact counts or product facts; when an article may be outdated but you cannot confirm it, say "check" rather than "wrong".
- Never write corrected steps for the product from guesswork; mark unknown details as [X] for the owner to confirm.
- Do not recommend retiring articles on rare but necessary topics (account closure, data requests, safety, legal terms) just because traffic is low.
- Keep customer personal data out of the output; if the export contains any, say so and do not repeat it.
- Do not recommend specific help-centre software.
</constraints>

<output_format>
## Summary
Five lines: overall state, the three biggest problems, and the first fix.
## Article findings
Table: Article | Problems found | Decision | Note.
## Duplicates and merges
## Gaps
Table: Contact reason or search term | Volume | Status (covered, not working, findability gap, missing) | Action.
## Ranked fix list
Numbered, with reason for the rank and owner role.
## Keeping it current
## Data caveats
</output_format>

可替换内容

  • {{help_center_export}}
  • {{#ticket_data}}
  • {{ticket_data}}
  • {{/ticket_data}}
  • {{#search_data}}
  • {{search_data}}
  • {{/search_data}}
  • {{#recent_changes}}
  • {{recent_changes}}
  • {{/recent_changes}}
  • [X]

分类与标签

办公效率、客服知识库、内容审查、搜索缺口、维护计划

前往 AI 图片创作