Detect early churn signals from usage + behavior data

Review this customer usage data and identify accounts at elevated churn risk.

Inputs:
- Usage data over last 90 days: {{paste — daily/weekly active users, key feature adoption, login frequency}}
- Support ticket activity: {{paste — volume, sentiment, severity}}
- Engagement signals: {{paste — exec attendance at QBRs, email opens, training attendance}}
- Contract data: {{paste — renewal date, term, ASP}}

Output:

## At-risk customers (ranked)
| Customer | Risk level | Primary signal | Secondary signal | Days to renewal |
|---|---|---|---|---|

Risk level: 🔴 Likely churn / 🟡 Watch closely / 🟢 OK but trending wrong direction

## Per top-3 risk: the diagnostic question
What the CSM needs to ask in their next conversation to confirm or rule out the risk.

## Patterns across the at-risk list
Are at-risk customers concentrated in one segment, one CSM, one product area? Don't miss the structural signal.

## Customers that look OK but I'd watch
Healthy NPS, healthy usage, but one weak signal (champion left, ownership change, etc.). The not-obvious cases.

## What I'd ignore
Common false positives (e.g. summer dip in seasonal SaaS, end-of-quarter usage spikes). Don't trigger interventions on these.

Hard rule: if "renewal in <60 days" appears alongside any 🔴 or 🟡, surface it as a separate "act this week" list at the top.