I run marketing for a B2B SaaS and the scariest number on my dashboard is churn. Support tickets are full of early warning signs, but nobody reads them at scale. This pack turns ticket text into a scoring model you can actually act on.
[Claude]
You are a churn analyst. Given a CSV of 200 support tickets with columns ticket_id, customer_name, plan, days_since_signup, and ticket_text, build a churn-risk scoring model. Define 5 risk signals (e.g. downgrade talk, competitor mentions, usage frustration, billing complaints, silence). Score each ticket 0-2 per signal, sum to a 0-10 risk score, and output a table sorted by score with a one-line reason per customer. Add a 3-bullet playbook for the top 5 at-risk accounts.
[ChatGPT]
Act as a retention marketer. I will paste a support ticket from a customer who has been quiet for 30 days. Identify 3 specific churn signals in the text, rate overall churn risk as low/medium/high, and write a 4-sentence win-back email that references their exact issue. Output as: signals, risk rating, email draft, and a follow-up cadence (day 3, day 7, day 14). Keep the email under 120 words and in a warm, non-salesy tone.
[Cursor]
In this repo, create a Python script churn_scorer.py that reads support_tickets.csv and outputs churn_risk.csv. Implement: normalize ticket_text (lowercase, strip stopwords), match against a keyword dictionary of 8 churn signals with weights, compute a weighted risk score per customer, and flag customers above a threshold. Include a –dry-run flag that prints the top 10 at-risk customers to stdout without writing the file. Add a short README with usage examples.
More packs in the archive at aitoolalliance.com/tag/prompt-pack.