Featured on
Kit 01
Everything you need to get up and running. The Feedback Synthesiser takes raw customer feedback from any source and returns a structured, prioritised analysis in under 30 seconds.
Open the tool in Claude →A Claude account - Pro or higher is recommended; large feedback sets can hit limits on free plans
Feedback data as a CSV file, or text you can paste directly
Optionally: contextual columns in your CSV like MRR, plan tier, account age, or customer segment
Open the Feedback Synthesiser in Claude
Add optional business context - e.g. "Solo founder, 40 paying customers, early-stage invoicing tool"
Upload your CSV or paste your feedback text
If using CSV: select which column contains the feedback, then tick any context columns
Click Analyse feedback and wait ~10-20 seconds
Review the output, then export or copy as needed
If you'd rather work directly in Claude chat, set up a project with the system prompt below. Paste your feedback into any conversation in that project and Claude will return the same structured analysis as text - without the visual interface. This also works well if you want the analysis baked into a broader project that holds your roadmap or customer context.
Go to claude.ai/projects and create a new project
Name it something like Feedback Synthesiser: [Your Company]
Copy the system prompt below and paste it into the Project instructions field, then fill in the bracketed sections
Open a new conversation inside the project and paste your feedback directly into the chat
Claude will return the same structured analysis - themes, priorities, recommendations - as text
Copy the prompt below and paste it into the Project instructions field when setting up your Claude Project. Fill in every section marked with [brackets] before saving - the more specific you are, the more useful the segment and recommendation outputs will be.
You are a customer feedback analyst embedded in [Company Name]'s feedback review workflow. Your job is to help the team make fast, confident decisions based on what customers are actually saying. **About this business:** [Company Name] is a [B2B SaaS / B2C / marketplace - delete as appropriate] product serving [brief description of customer base, e.g. "small creative agencies using project management software"]. The team is [solo founder / small team of N] and reviews feedback [weekly / after each release / ad hoc]. **What matters most here:** - [Top priority, e.g. "Retention - we care most about signals from paying customers, especially high-MRR accounts"] - [Second priority, e.g. "Onboarding friction - new users under 30 days are a key watch area"] - [Third priority if relevant, e.g. "Feature gaps - we're pre-roadmap, so specific missing features are high signal"] **How to handle this feedback:** When analysing feedback, weight your findings toward [e.g. "customers on paid plans over free users"] where contextual data allows. If segment data is present, surface patterns that would change what the team works on - don't mention segments just to mention them. Be direct and specific. Avoid generic recommendations like "improve communication" - push toward actionable next steps the team could act on this week. If the evidence doesn't support a strong conclusion, say so rather than overstating confidence. **Output guidance:** Return structured JSON as specified by the tool. Prioritise clarity over comprehensiveness - 3 sharp themes beat 6 vague ones. The output will be read by [a founder making roadmap decisions / a small team in a weekly review - adjust as needed], so write for that audience.
You can have multiple versions of this project - one per product line or customer tier - if you analyse meaningfully different feedback pools
The more specific your priorities and customer description, the sharper the segment signal output will be
The tool will auto-detect common column names, but it works best when your columns match these patterns:
| Column type | Examples it recognises |
|---|---|
| Feedback text | feedback, comment, review, response, note |
| MRR / revenue | mrr, revenue, arr, ltv |
| Plan / tier | plan, tier, status |
| Account age | account_age, tenure, days, months |
| Segment | segment, cohort, type, size |
If your column names don't match, you can manually select them in the column picker step.
One row per feedback item works best
You don't need to clean the text - the model handles messy punctuation and formatting fine
10-100 items is the sweet spot; very small sets (under 5) produce weak patterns, very large sets (500+) may hit token limits
Top signal
The single most important thing the feedback reveals. Good for sharing in a Slack update or kicking off a team discussion.
Recommended focus
What to do first, based on frequency and impact. Treat this as a starting point, not a directive.
Theme cards
Each card shows the problem customers are experiencing, evidence from the feedback, and a specific recommendation. High-priority themes appear first.
Segment signal (on a theme card)
Only appears when the contextual data shows a meaningful pattern for that theme - e.g. "This issue is concentrated among accounts under 60 days old." Omitted if no strong pattern exists.
Segment insights (at the bottom)
A standalone section summarising the most important cross-cutting patterns across your customer segments. Only appears when the data supports it.
| Problem | Likely cause | Fix |
|---|---|---|
| JSON parse error | Response cut off mid-output | Reduce feedback volume, or split into batches |
| No segment insights | Not enough contextual data, or patterns aren't strong enough | Add more context columns, or accept that this batch doesn't have clear segment patterns |
| Wrong column auto-selected | Column name not recognised | Manually select the correct column in the picker |
| Analysis feels generic | No business context provided | Add context in the text field, or set up a Claude Project with the system prompt |
Questions or feedback? Get in touch.
Featured on