Synthesize Voice of Customer themes from open-text survey responses
advancedClaude OpusCustomer ServiceCx Researchvoice-of-customercx-researchsurvey-analysissynthesisnps
Use case
Use this when you have open-text NPS, CSAT, churn-survey, or onboarding-survey responses that nobody has time to read. The prompt produces a prioritized theme map with verbatims, segment cuts, and recommended next actions — ready to drop into a QBR or roadmap review.
The prompt
You are a senior customer research analyst supporting the VP of Customer Experience at{{company_name}}, a{{industry}}company. You will synthesize{{response_count}}open-text survey responses into an actionable Voice of Customer report. Source data: - Survey type:{{survey_type}}(e.g., post-onboarding NPS, churn exit, quarterly CSAT) - Time window:{{time_window}}- Segments captured per response:{{segments}}(e.g., plan tier, tenure, ICP fit, CSM owner) - Raw responses:{{raw_responses}}Produce a synthesis with: 1. Theme map (5–9 themes). For each theme: - Short name and one-sentence definition - Volume: how many responses map to it (count and % of total) - Sentiment skew: positive / mixed / negative - Two to three verbatim quotes (verbatim — no paraphrasing) with respondent segment tag - Segment concentration: which segments over-index on this theme 2. Cross-cuts: where the same theme shows up differently across plan tier, tenure, or persona. 3. What changed vs. the previous period if{{prior_period_summary}}is provided — call out emerging and fading themes. 4. Three recommended actions, each tied to a specific theme, with the team that owns it (Product, Support, Onboarding, Pricing, etc.) and a rough effort tag (S/M/L). 5. Two questions the data cannot answer that warrant a follow-up interview round. Rules: - Do not invent quotes. If a quote is not in the source, do not include it. - Do not collapse contradictory signals into a single theme — surface the tension. - Prefer specific operational language ("ramp from trial to paid takes too long") over vague labels ("usability"). - Order themes by a composite of volume and severity, not volume alone.
Variables
{{{{company_name}}}}Replace with your {{company name}}{{{{industry}}}}Replace with your {{industry}}{{{{response_count}}}}Replace with your {{response count}}{{{{survey_type}}}}Replace with your {{survey type}}{{{{time_window}}}}Replace with your {{time window}}{{{{segments}}}}Replace with your {{segments}}{{{{raw_responses}}}}Replace with your {{raw responses}}{{{{prior_period_summary}}}}Replace with your {{prior period summary}}
Example
Input
company_name: Lumen Analytics industry: B2B SaaS (revenue analytics) response_count: 184 survey_type: post-onboarding NPS (day 30) time_window: Jan 1 – Mar 31 segments: plan tier (Starter / Growth / Enterprise), CSM-led vs. self-serve, ICP fit score raw_responses: | "Setup docs were fine but the Salesforce sync silently failed for two weeks." — Growth, self-serve, ICP 4 "Our CSM Priya was excellent — she rebuilt our dashboards in one call." — Enterprise, CSM-led, ICP 5 "I still don't know what the difference is between a 'pipeline view' and a 'forecast view'." — Starter, self-serve, ICP 3 "Pricing jumped when we added a third seat — felt like a bait and switch." — Growth, self-serve, ICP 4 "Reports take 40+ seconds to load on our data volume." — Enterprise, CSM-led, ICP 5 [... 179 more responses ...] prior_period_summary: Q4 top themes were (1) integrations breakage, (2) pricing surprise on seat expansion, (3) reporting performance.
Output
# Lumen Analytics — Day-30 NPS Synthesis (Q1, n=184) ## Top themes 1. **Silent integration failures** — 38 responses (21%), strongly negative. Sync jobs (Salesforce, HubSpot, Snowflake) fail without alerting the customer. Concentrated in self-serve Growth (28 of 38). Verbatim: "the Salesforce sync silently failed for two weeks." Verbatim: "found out our HubSpot pipe was stale only because a deal closed-lost in the wrong stage." Tension: Enterprise / CSM-led customers do not surface this — likely because the CSM catches it, not because the failure rate is lower. 2. **Pricing surprise on seat expansion** — 31 responses (17%), negative. Same theme as Q4 — has not faded. Growth tier over-indexes (24 of 31). Verbatim: "Pricing jumped when we added a third seat — felt like a bait and switch." 3. **Reporting performance at scale** — 22 responses (12%), mixed. Enterprise-only. Verbatim: "Reports take 40+ seconds to load on our data volume." 4. **CSM as the product** — 19 responses (10%), strongly positive. Enterprise CSM-led only. Verbatim: "Our CSM Priya was excellent — she rebuilt our dashboards in one call." Cross-cut: this is great, but it means the self-serve path is failing the same job. See Theme 5. 5. **Conceptual confusion in the core UI** — 17 responses (9%), negative. Starter / self-serve. Verbatim: "I still don't know what the difference is between a 'pipeline view' and a 'forecast view'." [Themes 6–8 omitted for brevity] ## What changed vs. Q4 - Emerging: Conceptual UI confusion (new — was not a top theme in Q4). - Persistent: Pricing surprise and integration failures — not improving. - Fading: Onboarding scheduling complaints (down from 14% to 3%, likely from the new self-scheduling tool). ## Recommended actions - **Integration health alerts (Product, M)** — Targets Theme 1. Notify the customer in-app and via email when a sync has been stale > 24h. - **Seat-expansion pricing transparency (Pricing + PMM, S)** — Targets Theme 2. Show the projected MRR delta in-app before a seat is added. - **Self-serve "first dashboard" guided flow (Onboarding, M)** — Targets Themes 4 and 5. Replicate what CSMs do in their first call. ## Questions for follow-up interviews - For Theme 1: how long does a silent sync failure typically run before the customer notices, and what is the downstream revenue impact? - For Theme 4: which 3–4 things does a CSM do in onboarding that we could productize?
Tips for best results
- 1Feed responses with their segment tags inline — segment cuts are where the real insight is, and Claude cannot infer them.
- 2Always include a prior-period summary if you have one. The 'what changed' section is usually the most valuable part of the report for a CX leader.
- 3If you have more than ~300 responses, batch by segment first (e.g., one synthesis per plan tier) and then ask Claude to synthesize the syntheses.
- 4Ban paraphrased quotes explicitly — synthesis prompts otherwise tend to hallucinate plausible-sounding verbatims.
- 5Pair this output with a 30-minute review with Support and CSM leads before sharing it broadly. They'll catch themes the survey missed.
Related prompts
Synthesize customer interview transcripts into themes
advancedAnalyze multiple customer interview transcripts or notes and synthesize them into a structured thematic summary with key insights, patterns, and recommended actions.
Customer Serviceresearchsynthesisqualitative
Write an NPS follow-up survey question set
intermediateGenerate a targeted NPS follow-up survey question set that turns a score into actionable customer insight.
Customer Servicenpssurveycx-research
Identify jobs-to-be-done from customer interview notes
advancedAnalyze customer interview notes to identify the jobs-to-be-done (JTBD) customers are hiring your product to do — and the functional, emotional, and social dimensions of each job.
Customer Servicejtbdjobs-to-be-donecustomer-research
Need help implementing this prompt in your workflow?
Book a call