Skip to content

Customer Feedback Theme Digest

Example prompt: "Every Monday morning, pull last week's customer feedback from our Zendesk tickets tagged 'feedback', our Typeform CSAT survey, and the NPS comments tab of our metrics sheet, group the comments into themes, and post a digest to our #product Slack channel with the headline issues and a couple of representative quotes."

The Problem

Customer feedback lands in three or four different places — support tickets tagged "feedback", post-purchase Typeform surveys, NPS comments dropped into a Google Sheet by the CS team, the odd Slack DM from a sales rep. Individually each one is interesting; collectively they would tell us what to fix next, but nobody has the time on a Monday morning to read 80 free-text fragments, mentally cluster them, and decide which theme is loudest this week. So the feedback either gets actioned one ticket at a time, which biases towards whoever shouted loudest, or it sits in a backlog that nobody opens.

How GloriaMundo Solves It

We build a workflow that runs every Monday morning. Three integration steps pull last week's feedback from each source — Zendesk tickets with the "feedback" tag, new rows in the Typeform survey, and the NPS comments tab in Google Sheets. A code step normalises them into a single list with source, submitted_at, and verbatim fields. An LLM step reads the whole list and clusters the comments into themes such as "onboarding friction", "billing confusion", or "missing integration X", with a count per theme and the two or three most representative verbatims for each. An integration step posts the digest to the #product Slack channel, and a second integration step appends the same digest as a new section in a "Customer Feedback Themes" Notion page so the history is browseable. Glass Box preview shows the clustering and the digest text before the workflow is approved.

Example Workflow Steps

  1. Trigger (scheduled): Runs every Monday at 09:00.
  2. Step 1 (integration): Read Zendesk tickets closed in the last 7 days with the "feedback" tag — subject, body, ticket URL.
  3. Step 2 (integration): Read Typeform responses to the CSAT survey from the last 7 days.
  4. Step 3 (integration): Read new rows in the "NPS comments" tab of the metrics sheet since the last run.
  5. Step 4 (code): Normalise the three sources into a single list of {source, submitted_at, verbatim, link} records and dedupe near-identical comments.
  6. Step 5 (LLM): Cluster the comments into themes; for each theme produce a name, a count, a one-sentence description, and the two or three most representative verbatims (with source links).
  7. Step 6 (integration): Post the digest as a Slack message to #product, themes ordered by count.
  8. Step 7 (integration): Append the same digest as a dated section in the "Customer Feedback Themes" Notion page.

Integrations Used

  • Zendesk — source of support-flagged feedback
  • Typeform — source of structured survey feedback
  • Google Sheets — source of NPS comments captured by CS
  • Slack — where the weekly digest is delivered to the product team
  • Notion — persistent history of themes for week-on-week comparison

Who This Is For

Product managers at scale-up SaaS companies where customer feedback arrives in three or more places and nobody has been formally given the role of reading all of it. Especially useful for teams of one or two PMs covering 5,000+ customers, where prioritisation by anecdote is the failure mode the digest is meant to replace.

Time & Cost Saved

Reading 60-100 feedback fragments across three tools and writing a coherent weekly summary takes a PM roughly two hours, which is why most weeks it does not happen. This workflow takes minutes and the PM's two hours go into the conversations the digest provokes rather than the data assembly.