The prompt
Score 10 of 10, Excellent
Use it when You have a month of ticket subjects or summaries and want to show the product team which problems cause the most contacts, with counts and examples.
- {{ticket_export}}
- {{period}}
- {{known_issues}}
- {{audience}}
Full prompt and why it works
System prompt
You are a support operations analyst. You group tickets by the underlying problem, count honestly, and point to the kind of fix that would stop the contacts.
User message
<tickets> {{ticket_export}} </tickets> <context> Period covered: {{period}} Known issues and existing help articles: {{known_issues}} </context> The tickets above are one per line, each with an ID and a subject or summary. I will share the result with {{audience}}, who decide what to fix next, so it should point to fixes rather than list complaints. Produce three parts: 1. A table of the top contact reasons (at most 12), with the columns Reason, Count, Share of tickets, Example ticket IDs (up to 3), and Likely fix type. Fix type is one of: product bug, product design, help content, policy or process, billing, or unknown. Order the table by count. 2. For the top three reasons, two or three sentences each on what customers are trying to do and where they get stuck, citing ticket IDs. 3. Quick wins: reasons a help article or a saved reply could reduce, especially where a known issue or article already exists but customers still ask. Rules: - A reason must be specific enough to act on, for example "Can't find where to download invoices" rather than "Billing". - Each ticket counts once, under its main reason. Tickets with too little information go under "Unclear" and still count. - Report tallies as numbers. Do not replace them with words like "many" or "most". - Leave customer names, email addresses and other personal details out of the output. If there are too many tickets to count reliably in one pass, say so, mark the counts as approximate with "~", and suggest how to split the export.
Variables
- {{ticket_export}}
- One ticket per line with its ID and subject or summary, exported from your help desk. Remove personal details first.
- {{period}}
- The dates the export covers.
- {{known_issues}}
- Known bugs and existing help articles on these topics, or "none".
- {{audience}}
- Who reads it, for example "the product team's monthly review".
Why it works
- What good looks like. A reason must be specific enough to act on, with an example of the difference, so the report says "Can't find where to download invoices" rather than "Billing".
- Format. Each reason gets a fix type, which turns a list of complaints into a list of jobs for product, help content or billing.
- If unsure. With too many tickets to count in one pass, it marks counts as approximate instead of presenting guesses as tallies.