Find the Themes in Customer Feedback and CSAT Comments with Real Counts

Returns a theme table with exact counts, a verbatim quote from your data, likely root cause and suggested owner, plus top actions and data gaps.

You getFirst line, table with columns and top 3 actions

Prompt

Prompt
You are a customer support analyst who turns raw feedback into clear action items for product, support and operations teams.

I will paste customer feedback or CSAT comments right after this prompt. Group them into themes and show what is driving them.

Rules:
1. Count only what is in the pasted data. Each count is the exact number of comments you assigned to that theme. If a comment fits two themes, count it in both and note it.
2. Quote examples word for word from the data, in quotation marks. Do not paraphrase, merge or correct quotes.
3. Label root causes as likely, and base each one on wording in the comments, not on guesses about internal systems.
4. Suggest one owner per theme: Product, Engineering, Support, Billing, Logistics, Marketing, or Other (name it).
5. Do not add percentages, trends or benchmarks unless they can be calculated directly from the pasted data.

Output format:
- First line: total comments analyzed, and how many were too vague to classify.
- Table with columns: Theme | Count | Verbatim example quote | Likely root cause | Suggested owner. Sort by count, highest first.
- Top 3 actions: numbered, one sentence each, each tied to a theme.
- Positive themes: up to 3 bullets with counts and one verbatim quote each.

End with one line on what this data cannot tell us, such as small sample size, missing scores, dates or customer segments.

Example output

At a glanceComplaint themes in 42 CSAT commentsExact comment counts per theme; 6 comments counted in two themes
  • Late deliveries: 11
  • Missing or damaged items: 8
  • Hard to skip or cancel: 7
  • Slow support replies: 6
  • Refund delays: 4

Example for: 42 CSAT comments from a meal-kit subscription service, pasted from a monthly survey export.

42 comments analyzed. 3 were too vague to classify (for example, "meh" and "ok I guess"). 6 comments were counted in two themes.

ThemeCountVerbatim example quoteLikely root causeSuggested owner
Late deliveries11"Box showed up Thursday night, I needed it Tuesday. Food was warm."Carrier delays in some areas; delivery window not metLogistics
Missing or damaged items8"No limes AGAIN and the cream leaked all over the bag."Packing errors and weak packaging for liquidsLogistics
Hard to skip or cancel7"Took me 20 minutes to find where to skip a week."Skip and cancel options hard to find in the account areaProduct
Slow support replies6"Emailed twice about a missing box, heard back 4 days later."Email queue backlog; no chat option mentionedSupport
Refund delays4"Was told credit in 48 hours, still waiting a week later."Refund timing promised by agents does not match billing processBilling

Top 3 actions

  1. Late deliveries: ask Logistics to pull delivery-time data for the postcodes in these 11 comments and check whether one carrier or region is behind most of them.
  2. Missing or damaged items: add a packing check for liquids and small produce, since cream, limes and herbs appear repeatedly.
  3. Hard to skip or cancel: have Product test how many clicks it takes to skip a week and put the skip button on the account home screen.

Positive themes

  • Recipe quality (9): "The miso salmon was better than takeout, my kids ate all of it."
  • Friendly agents (5): "Dana on chat sorted my refund in five minutes, super kind."
  • Portion size (3): "Portions are actually filling, no snacking after."

What this data cannot tell us: comments have no dates, scores or locations, so we cannot see trends over time, link themes to low ratings, or confirm which regions have delivery problems.

Written by AI from this prompt. Your results will vary.

Tips and common edits

Why this works

  • Requiring exact counts from the pasted data stops the model from inventing percentages or survey-style statistics.
  • Verbatim quotes in quotation marks let you check every theme against the source text quickly.
  • A named owner per theme turns the analysis into a list you can route to teams the same day.

Use after a CSAT survey, NPS export or a week of ticket comments. Paste the raw text right after the prompt in ChatGPT, Claude or Gemini. Remove names and emails first.

Common edits

  • Add a column for average CSAT score per theme if your pasted data includes scores.
  • Limit the table to the top 5 themes for a short slide or weekly update.

Prompt details

By
Prompts-GPT editors
Published

Did this prompt work?

Keep exploring

More in Customer Support