Classify survey comments for churn risk with Claude Haiku, Supabase and n8n and get a weekly theme digest
A step-by-step build that tags each open-text survey response with themes, sentiment and a churn signal using a cheap Claude model, alerts on risky ones, and emails a weekly summary of themes.
Evidence: The source shows its work: steps, screenshots, code or data you can inspect.
The business problem
A SaaS founder collects survey comments but reads only a fraction of them, so churn warnings in the text go unnoticed.
What was tried
A survey with a 0 to 10 score and one comment field sends responses by webhook with metadata added from the product database. Two Supabase tables hold responses and weekly snapshots. One workflow calls Claude Haiku per response to assign one to three themes from nine, set sentiment, a churn signal and a summary as JSON, and emails an alert if the signal is raised. A Monday workflow aggregates the week, asks Claude for a short narrative and emails a digest with themes, quotes, churn patterns and one action. After about 50 responses the founder reviews ten against the tags and adjusts the definitions.
What was reported (not reported)
The author says the build takes about two hours and about $5 a month in API calls, and claims about 15 minutes of automation instead of 8 hours of manual review per cycle. No accuracy or case-study results are provided.
Limitations
The churn rules are heuristics with no evidence that they predict churn, and no classification accuracy is measured beyond a manual check of ten responses. Retry handling, the exact model version and personal-data handling for stored emails and comments are not covered, and the $5 estimate excludes n8n, Supabase, Resend and survey tool costs. The article links to the author's paid build service.
What you need
A webhook-capable survey tool, access to user metadata, and accounts for Supabase, n8n, the Anthropic API and Resend, plus basic SQL and n8n familiarity.
Sources
- Ekofi Pulse (Substack) ↗ Tutorial, published October 2, 2026
Source published: October 2, 2026. Last reviewed here: October 11, 2026. Spot a mistake? Tell us.
Tools in this workflow
- Claude: Classifies responses and writes the digestTry Claude
- n8n: Runs both workflowsTry n8n
- Supabase: Stores responses and weekly snapshotsTry Supabase
- Resend: Sends alerts and the digestTry Resend
Related workflows
Send a Monday business metrics email from Stripe, Notion and Sheets with Claude writing the narrative in n8n
A shared n8n template pulls last week's Stripe revenue, Notion deal pipeline and a Sheets row of support numbers, has Claude write a three-paragraph summary, and emails and posts it to Slack.
Draft help-centre-cited support replies and propose refunds that a person must approve
An open-source support workbench classifies tickets into three tiers, drafts replies that cite help-centre articles, reads billing data in read-only test mode, and only proposes refunds that need a named approver.
Install a free pack of six Claude skills for positioning, customer research, pricing and go-to-market
An open-source set of Claude skills that builds a shared context file, then produces positioning, customer research, competitive briefs, pricing and packaging and launch plans that build on each other.
Turn a freelance service into priced packages and a productized offer with six AI prompts
A newsletter post gives a nine-step sequence and six prompts to find the outcome clients buy, standardise repeatable parts, design three package tiers, price by value and build a productized service.