GitHub only
Analyze a churn spike
Devin queries a SQLite warehouse to explain a churn spike, tests competing explanations, and delivers an interactive HTML report with the SQL behind every number.
Steps
- Enter your GitHub user or org. The links and the prompt on this page use it.
- Create your own copy of the starter repo. Open GitHub's new-repository page for devin-demo-churn-analysis, choose your account as the owner, keep the name
devin-demo-churn-analysis, and click Create repository. - Make sure Devin can access the new repo under Settings → Connections → GitHub.
- Start a session at app.devin.ai and paste the prompt below.
Prompt
Our CFO noticed churn roughly tripled in September 2026 and needs to know why before the board meeting. The data is in https://github.com/<you>/devin-demo-churn-analysis (SQLite warehouse snapshot, see README).
Investigate like a senior analyst: quantify the spike, segment it (region, plan, channel, payment method, cancel reason), test competing hypotheses against the release notes and support tickets, and separate correlation from causation. Deliver an interactive HTML report (charts, filterable tables, the SQL behind every number) with an executive summary up top, a root-cause conclusion with confidence level, the size of the revenue impact, and 3 concrete recommendations. Commit the report and the queries to the repo in a PR.What to watch
- It segments the spike several ways before drawing a conclusion.
- It separates correlation from causation and states its confidence.
You're done when
- There is an interactive HTML report with an executive summary, charts, filterable tables and SQL.
- It recommends three concrete actions, all committed in a PR.