Rapid Feedback Theme Clusterer (Paste → Cluster → Tag Severity)
Quickly turn pasted feedback into clean themes with frequency, severity, and representative quotes so teams can spot the loudest problems fast.
Prompt Overview
Tips For You
- Paste at least 50–200 feedback lines for stable clusters. - Add known modules (e.g., Billing, Onboarding) to sharpen subthemes. - Export the JSON to Google Sheets for quick charts and pivot summaries.
From Operations TeamNexusAi TechnologyProblem It Solves
Manual clustering is slow and inconsistent. This prompt standardizes tagging, reduces noise, and produces a shareable theme table in minutes.
Clean theme naming
Generates user-friendly, concise theme labels.
Evidence-backed severity
Severity grounded in quotes and user outcomes.
Import-ready JSON
Directly loadable into Sheets or Airtable.
Ambiguity log
Tracks unclear items and placement rationale.
AI Prompt Instructions
Act as: A senior product analyst experienced in qualitative research synthesis, customer support triage, and product discovery.
Why this task matters: Teams sit on piles of feedback but struggle to see patterns. Fast, consistent clustering reveals the top pain points and where to focus next.
Important boundaries:
- Use only the provided feedback; do not invent issues.
- Prefer simple, descriptive theme names users would recognize.
- Keep quotes verbatim; lightly trim for clarity only.
User inputs:
- Raw feedback lines (support tickets, survey comments, reviews, notes).
- Optional: known personas, product areas/modules, and any banned categories.
Objectives:
1) Cluster entries into clear themes and subthemes.
2) Score each theme: frequency (count), severity (1–5 based on user impact), recency (% from last 30–60 days if provided), and effort guess (S/M/L if mentioned).
3) Output a clean table plus a JSON block teams can load into a spreadsheet.
Analysis workflow:
1) Normalize text: remove duplicates, merge near-duplicates, flag ambiguous entries.
2) Draft themes using user language; group borderline items and justify placement.
3) Compute metrics: frequency, severity (impact on task completion, revenue risk, or support load), and recency trends.
4) Surface 3–7 highest leverage themes and include 1–2 representative quotes each.
5) List edge cases that didn’t cleanly fit and why.
Required output format:
- Summary paragraph (3–5 sentences).
- Table (Theme | Subtheme | Frequency | Severity 1–5 | Recency % | Effort S/M/L | Top Quotes [1–2]).
- JSON array of objects with the same fields for import.
- Notes: key assumptions, data gaps, and recommended next data pull.
Quality controls:
- Deduplicate aggressively; show dedupe count.
- Avoid theme creep; keep names concise (<5 words).
- Check sentiment bias; severity must be grounded in text evidence.
Verification checklist:
- Are top themes backed by multiple entries and quotes?
- Do severity justifications reference user outcomes?
- Are ambiguous items logged with rationale?
Final instruction:
Return the full deliverable. If inputs are small (<20 lines), still apply the same structure. If severe data gaps exist, add a short request for specific missing inputs.
Expected Outcome
Summary: We identified 6 themes, led by Checkout Failures (24% of entries, Severity 5) and Slow Onboarding (18%, Severity 4). Recency skewed toward the last 30 days for both. Table (excerpt): - Checkout Failures | Payment declines | 28 | 5 | 61% | M | "Card keeps failing at confirm screen." - Slow Onboarding | Confusing steps | 21 | 4 | 54% | S | "I’m lost after email verification." JSON (excerpt): [ {"theme":"Checkout Failures","subtheme":"Payment declines","frequency":28,"severity":5,"recency":0.61,"effort":"M","quotes":["Card keeps failing at confirm screen."]} ]
Implementation Journey
Cluster in ChatGPT
Open ChatGPT and paste 50–200 raw feedback lines with any known personas or modules. Ask for theme clustering with frequency, severity, and quotes using the provided prompt. Expect a summary, a table, and a JSON block.
10-15 minQuantify in Google Sheets
Paste the JSON into Google Sheets using IMPORTJSON or a quick script. Build a pivot to rank themes by frequency and a bar chart for severity distribution. Highlight the top 5 themes.
10-20 minShare and tag in Notion
Create a Notion database for themes and paste the table. Tag related feature requests and attach representative quotes so engineers and designers can see real user language.
10 min