30 Claude Prompts for NPS Surveys

You are a lifecycle email writer who sends NPS invites that get opened and completed, not ignored as research spam. <context> I need an invite email for our next NPS wave. Past invites sounded like internal research requests and completion was weak. </context> <inputs> - Product name: [PRODUCT NAME] - From name and address: [FROM NAME AND ADDRESS] - Time to complete: [MINUTES TO COMPLETE] - What we do with answers: [HOW FEEDBACK IS USED] - Audience: [AUDIENCE SEGMENT] </inputs> <task> Write a subject line, preview text, and body that states why we are asking, how long it takes, and what happens to the answer. Include one plain-text version and one short HTML-ready version with a single button label. </task> <constraints> Body under 90 words. One link only. Do not mention the 0 to 10 scale in the subject. Do not promise a gift or raffle. Do not use more than one question mark in the subject. </constraints> <format> Sections: Subject, Preview Text, Plain-Text Body, Button Label, HTML Body. </format>

You are a survey operations lead writing a single reminder to people who have not completed the NPS survey. <context> The first invite went out [DAYS SINCE FIRST INVITE] days ago. I can send one reminder, then we close the wave. </context> <inputs> - Days since first invite: [DAYS SINCE FIRST INVITE] - Product name: [PRODUCT NAME] - Remaining window: [DAYS LEFT IN WAVE] - First invite subject we used: [FIRST SUBJECT LINE] - Tone we used last wave that felt pushy: [PUSHY PHRASE TO AVOID] </inputs> <task> Write one reminder email that is clearly a follow-up, not a new campaign. Change the subject so it does not look like a duplicate of the first invite. </task> <constraints> Under 70 words. Do not guilt people for not responding. Do not say last chance unless the remaining window is 2 days or fewer. Do not reuse the pushy phrase. Keep one link. </constraints> <format> Subject, Preview Text, Body. Add a one-line note on which day of the wave this should send. </format>

You are a messaging specialist writing an SMS NPS invite that stays inside carrier length limits and still names the brand. <context> We want to offer SMS as a channel for [AUDIENCE SEGMENT] who rarely open email. Legal requires an opt-out line. </context> <inputs> - Product name: [PRODUCT NAME] - Audience segment: [AUDIENCE SEGMENT] - Short survey URL: [SHORT URL] - Opt-out keyword: [OPTOUT KEYWORD] - Character limit we must stay under: [CHARACTER LIMIT] </inputs> <task> Write 3 SMS variants under the character limit, each including the brand, the ask, the link, and the opt-out. Rank them by expected completion versus risk of looking like spam. </task> <constraints> Stay under [CHARACTER LIMIT] including spaces. No emojis. No urgency words like act now. Include the opt-out keyword in every variant. Do not put the score scale in the SMS itself. </constraints> <format> Three numbered SMS texts with character counts, then a 3-row table: Variant, Completion Bet, Spam Risk. </format>

You are a product copywriter writing the in-app NPS banner and the dismiss state so it does not feel like an intercept ad. <context> We will show an NPS banner after [IN APP TRIGGER]. Design and legal already capped how long it can stay on screen. </context> <inputs> - Product name: [PRODUCT NAME] - In-app trigger: [IN APP TRIGGER] - Max words on the banner: [MAX WORDS] - Dismiss label we can use: [DISMISS LABEL] - Where the full survey opens: [SURVEY DESTINATION] </inputs> <task> Write banner headline, one-line supporting text, primary button label, and dismiss label. Add a version for mobile where space is tighter. </task> <constraints> Respect the max word count on the banner headline plus supporting text combined. Do not say it only takes a second if the survey is more than one screen. Do not block the core task the user just finished. No exclamation marks. </constraints> <format> Desktop banner block, Mobile banner block, Button labels, When not to show this banner (3 bullets). </format>

You are an NPS program manager choosing how to send the next wave across email, in-app, and SMS without double-surveying the same person. <context> Last wave we blasted everyone on email and a slice also saw the in-app prompt, so some people got two surveys in a week. </context> <inputs> - Audience size: [AUDIENCE SIZE] - Channels we can use: [AVAILABLE CHANNELS] - Overlap we saw last wave: [OVERLAP NOTES] - Target completed responses: [TARGET COMPLETES] - Quiet hours or send rules: [SEND RULES] </inputs> <task> Recommend a channel mix and a suppression rule so one person gets one survey. Estimate how many invites to send on each channel to hit the target completes, using conservative completion rates if I did not give exact ones. </task> <constraints> Do not recommend a channel that is not in the available list. State any assumption you make about completion rate. Keep the plan to one wave, not a year-long calendar. Call out the double-survey risk in one sentence. </constraints> <format> A table: Channel, Invite Volume, Suppression Rule, Assumed Completion Rate. Then a 5-line send order. </format>

You are a VoC analyst building a coding taxonomy from raw NPS comments before the team starts tagging. <context> We have a pile of open-end comments and no shared theme list. Analysts have been inventing tags, so quarter-over-quarter charts do not match. </context> <inputs> - Product name: [PRODUCT NAME] - Sample of raw comments: [RAW COMMENTS] - Existing tags if any: [EXISTING TAGS OR NONE] - Number of parent themes we can maintain: [MAX PARENT THEMES] - What leadership asks about every month: [LEADERSHIP QUESTIONS] </inputs> <task> Propose a two-level taxonomy: parent themes and child codes. Map each sample comment to a child code. Note comments that need a new code or are not codeable. </task> <constraints> Stay at or under the max parent theme count. Reuse an existing tag when it still fits, do not rename for style. Do not create a theme called Other larger than 15 percent of the sample. Do not invent product bugs that the comments do not state. </constraints> <format> Sections: Parent Themes, Child Codes with definitions, Comment-to-Code map, Codes to drop or merge. </format>

You are a research operations analyst coding a batch of NPS verbatims against a fixed taxonomy. <context> I need this batch coded today so we can close the monthly report. The taxonomy is frozen. I do not want new themes invented mid-batch. </context> <inputs> - Frozen taxonomy: [THEME TAXONOMY] - Comments to code: [COMMENT BATCH] - Score next to each comment if available: [SCORES PER COMMENT] - Multi-code rule: [MAX CODES PER COMMENT] - Language of comments: [COMMENT LANGUAGE] </inputs> <task> Code each comment with up to the max number of child codes. Mark sentiment as negative, mixed, or positive. Flag any comment that is too thin to code or is not in the stated language. </task> <constraints> Do not add new theme names. If nothing fits, use Uncoded and quote the reason in 8 words or fewer. Keep codes at the child level, not the parent. Do not change the respondent score. </constraints> <format> A table with columns: Comment ID or excerpt, Score, Codes, Sentiment, Flag. End with a count per child code. </format>

You are a CX analyst turning coded NPS themes into a driver view that product and success can act on. <context> The monthly theme counts are done. Leadership wants to know which themes actually move the score, not which ones are merely common. </context> <inputs> - Theme counts by score band: [THEME COUNTS BY BAND] - Total responses this period: [TOTAL RESPONSES] - Prior period theme counts: [PRIOR PERIOD COUNTS] - Product areas we can change this quarter: [CHANGEABLE PRODUCT AREAS] - Account segments in the file: [SEGMENTS] </inputs> <task> Rank themes by how strongly they sit with detractors versus promoters. Call out themes that grew versus last period. Name the 3 themes worth a fix this quarter given what we can actually change. </task> <constraints> Do not treat a popular promoter theme as a problem. Do not recommend work outside the changeable product areas unless you label it as a process or success fix. If a count is too small to trust, say so and skip the rank. </constraints> <format> A ranked table: Theme, Detractor Share, Promoter Share, Change vs Prior, Action Owner. Then 3 recommended bets with one sentence each. </format>

You are a customer insight lead writing a one-page brief of what detractors said this period, for the weekly ops meeting. <context> Product and support leads will not read 200 comments. They will read one page if it is specific and quotes real words. </context> <inputs> - Detractor comments: [DETRACTOR COMMENTS] - Period: [REPORTING PERIOD] - Product name: [PRODUCT NAME] - Known outages or incidents in the period: [INCIDENTS IN PERIOD] - Segments we care about: [SEGMENTS] </inputs> <task> Summarize detractor comments into 4 to 6 themes. For each theme include one short verbatim quote and a count. Separate incident-driven comments from ongoing product or service issues. </task> <constraints> Use only quotes that appear in the pasted comments. Do not clean up grammar so much that the voice is lost. If an incident is listed, do not bury those comments inside a generic reliability theme without naming the incident. Keep the brief to one page. </constraints> <format> Header with period and detractor count, then theme blocks: Theme, Count, Quote, Incident or Ongoing. </format>

You are a customer marketing partner pulling promoter comments that sales and success can reuse without sounding like a testimonial mill. <context> We have promoter verbatims and I need a short reel for QBR slides and recruit-a-reference outreach, with anything confidential stripped. </context> <inputs> - Promoter comments: [PROMOTER COMMENTS] - Industries we sell into: [INDUSTRIES] - Words we cannot publish: [WORDS TO STRIP] - Use cases we want proof for: [TARGET USE CASES] - Format needed: [SLIDE OR EMAIL] </inputs> <task> Select 8 comments that name a concrete outcome. Rewrite each into a one-line proof point and keep the original next to it. Flag any comment that should stay internal. </task> <constraints> Do not invent metrics the comment does not contain. Strip the listed words and any company or person names that look identifying. Prefer comments that match the target use cases. Do not turn a vague love the product into a proof point. </constraints> <format> A table: Original, One-line Proof, Use Case, Internal or Sharable. End with 3 comments to ask for a reference call. </format>

You are a customer success manager writing the first reply to an NPS detractor within the SLA. <context> A customer scored us [NPS SCORE] and left a comment. I need a reply that owns the issue, names the next step, and does not sound like a template. </context> <inputs> - Contact name: [CONTACT NAME] - Account name: [ACCOUNT NAME] - NPS score: [NPS SCORE] - Their comment: [VERBATIM COMMENT] - What we can do next: [NEXT STEP WE CAN OFFER] - Owner who will follow through: [OWNER NAME AND ROLE] </inputs> <task> Write a short email that quotes the part of their comment we are acting on, states the next step, names the owner, and asks one question if we still need a fact. </task> <constraints> Under 130 words. Do not say we apologize for any inconvenience. Do not ask them to raise the score. Do not promise a fix we did not list. If the comment is about billing and the next step is a product call, flag the mismatch instead of sending it. </constraints> <format> Subject line and body, ready to send. Add a one-line internal note if the next step does not match the comment. </format>

You are a CX lead writing outreach to NPS passives whose comments point to a fixable gap, not indifference. <context> Passives are most of our sample. I want a short note for people who scored 7 or 8 and named a specific friction, not a mass thank-you. </context> <inputs> - Contact name: [CONTACT NAME] - Score: [NPS SCORE] - Their comment: [VERBATIM COMMENT] - Fix or help we can offer: [FIX OR HELP] - Product name: [PRODUCT NAME] </inputs> <task> Write an email that thanks them without flattery, restates the friction in their words, and offers the fix or help. End with one yes/no question. </task> <constraints> Under 110 words. Do not call them a passive. Do not ask them to become a promoter. If the comment is empty, write a 4-line version that asks what almost made this a 9. Do not attach a calendar link unless the help requires a meeting. </constraints> <format> Subject, Body, and a fallback Body for empty comments. </format>

You are a customer marketing manager writing a promoter follow-up that asks for a referral or review only when the comment supports it. <context> A customer scored 9 or 10. I want a follow-up that thanks them and, if the comment is specific, asks for one concrete action. </context> <inputs> - Contact name: [CONTACT NAME] - Score: [NPS SCORE] - Their comment: [VERBATIM COMMENT] - Ask we prefer: [REFERRAL, REVIEW, OR CASE STUDY] - Link or next step for the ask: [ASK DESTINATION] </inputs> <task> Write a thank-you email. If the comment is specific enough to quote, include a short quote and make one ask. If the comment is thin, thank them and skip the ask. </task> <constraints> Under 120 words. One ask only. Do not stack a review and a referral. Do not offer payment. Do not rewrite their comment into marketing copy inside the email they will read. </constraints> <format> Subject and Body. Label the top of the draft ASK or THANK ONLY so I do not send the wrong one. </format>

You are a customer success manager preparing a 15 minute close-the-loop call after a low NPS score. <context> Email was not enough. I have a call booked with the person who scored us and I want a script I can glance at, not a monologue. </context> <inputs> - Contact name and role: [CONTACT NAME AND ROLE] - Score and comment: [SCORE AND COMMENT] - Account health notes: [ACCOUNT HEALTH NOTES] - What we already tried: [ACTIONS ALREADY TAKEN] - Call length: [CALL LENGTH IN MINUTES] </inputs> <task> Write a time-blocked call plan: open, diagnose, commit, close. Include 5 questions that get past the survey comment into the real issue, and a closing line that restates the commitment. </task> <constraints> Fit the call length. Leave more than half the time for them to talk. Do not open by reading their score back like a reprimand. Do not promise a roadmap item that is not in actions already taken or in a new commitment I type later. </constraints> <format> Time-blocked agenda, 5 questions, Commitment restatement line, After-call CRM note template. </format>

You are a VoC operations owner turning closed loops into a tracker row your weekly NPS standup can trust. <context> We say we close the loop but the sheet is a mess of half notes. I need one clean row per response so we can see aging and owners. </context> <inputs> - Response record: [RESPONSE ID, SCORE, COMMENT, DATE] - Current owner: [CURRENT OWNER] - Actions taken so far: [ACTIONS TAKEN] - Customer reply if any: [CUSTOMER REPLY OR NONE] - SLA in business days: [SLA DAYS] </inputs> <task> Write one tracker row and a status of Open, Waiting on Customer, Waiting on Us, or Closed. If closed, write a 2-sentence outcome. If open, write the next action and the aging versus SLA. </task> <constraints> Do not mark Closed unless the customer reply or the actions taken show the issue was addressed or the customer declined further contact. Do not invent a date. If the record is missing a field, list the missing field instead of guessing. </constraints> <format> A single table row: Response ID, Status, Owner, Aging vs SLA, Next Action or Outcome, Risk if we slip. </format>

You are a VoC lead writing the monthly NPS note for the executive staff meeting. <context> I have 10 minutes on the agenda. They want the score, what moved, and what we are doing, not a tour of the dashboard. </context> <inputs> - Current NPS and response count: [CURRENT NPS AND N] - Prior period NPS and N: [PRIOR NPS AND N] - Top detractor themes: [DETRACTOR THEMES] - Top promoter themes: [PROMOTER THEMES] - Actions already in flight: [ACTIONS IN FLIGHT] </inputs> <task> Write a one-page exec summary: headline score movement, why it moved, who is affected, and the 3 actions in flight with owners if I provided them. </task> <constraints> Lead with the number and the N. Do not hide a drop. If N is too small to treat the move as real, say that in the first paragraph. No appendix. No request for more headcount unless an action already lists it. </constraints> <format> Sections: Score, What Moved, Who Is Affected, Actions in Flight, Decision Needed (or None). </format>

You are a CX analyst writing the trend narrative that sits next to a 6 to 12 month NPS chart. <context> The chart is already built. I need the paragraph that stops people from inventing their own story in the meeting. </context> <inputs> - Monthly scores and N: [MONTHLY SCORES AND N] - Major product or service events: [EVENTS BY MONTH] - Sampling or method changes: [METHOD CHANGES] - Segment mix shift if any: [SEGMENT MIX SHIFT] - Audience: [AUDIENCE FOR THIS NARRATIVE] </inputs> <task> Write a trend narrative that ties score moves to events, method changes, or mix shifts. Separate a real change in sentiment from a change in who answered. </task> <constraints> Do not claim a causal story the events list cannot support. If a method change lands in the same month as a score jump, say both and do not pick a winner. Keep it to 180 words. No forecast unless asked. </constraints> <format> Three short paragraphs: What the line did, What else changed in the data, How to read next month. </format>

You are a VoC analyst building an NPS scorecard by segment for a quarterly review. <context> The overall score looks stable, but I suspect one segment is carrying the number and another is quietly sinking. </context> <inputs> - Overall NPS and N: [OVERALL NPS AND N] - Scores by segment: [SEGMENT SCORES AND N] - Segment definitions: [SEGMENT DEFINITIONS] - Period: [REPORTING PERIOD] - Decision this scorecard should inform: [DECISION TO INFORM] </inputs> <task> Build a scorecard that ranks segments by score and by weight in the overall number. Flag any segment where N is too small. End with the one sentence that should drive the stated decision. </task> <constraints> Do not average segments as if they were equal size. Show N next to every score. If a segment definition is fuzzy, say so rather than treating it as clean. No heat-map colors in text, use words like above, in line, below. </constraints> <format> A table: Segment, N, NPS, Weight in Overall, Read. Then one decision sentence. </format>

You are a chief of staff turning NPS into one board slide that a director can defend without a backup appendix. <context> The board packet allows one slide for customer sentiment. Last quarter we crammed four charts and the discussion went nowhere. </context> <inputs> - Current NPS, N, and target: [CURRENT NPS, N, TARGET] - Four-quarter trend: [FOUR QUARTER TREND] - Top 2 drivers this quarter: [TOP TWO DRIVERS] - Action and owner: [ACTION AND OWNER] - Risk the board already knows: [KNOWN BOARD RISK] </inputs> <task> Write the slide: title, one score block, one trend sentence, two driver bullets, one action, and the speaker note the presenter should say if someone asks why they should trust N. </task> <constraints> One slide of content, not a script for five slides. Do not introduce a second metric. If the known board risk overlaps a driver, connect them in one clause, do not hide it. No adjectives that are not backed by the numbers given. </constraints> <format> Slide title, On-slide text in layout order, Speaker note (max 60 words). </format>

You are a CX operations lead breaking company NPS into team-level views without turning it into a blame chart. <context> Support, success, and product each think the score is the other team's problem. I need a breakdown they can use in a shared review. </context> <inputs> - Company NPS and N: [COMPANY NPS AND N] - Theme counts tagged to teams: [THEMES BY TEAM] - Team owners: [TEAM OWNERS] - Period: [REPORTING PERIOD] - Shared goals already agreed: [SHARED GOALS] </inputs> <task> Allocate themes to teams using the tags I provided. For each team write a score-relevant summary, one owned action, and one dependency on another team. </task> <constraints> Do not compute a fake team NPS unless I gave team-level scores. Do not rank teams from best to worst. If a theme is tagged to two teams, split the action, do not pick a winner. Keep each team block under 80 words. </constraints> <format> One block per team: Themes they own, What customers said, Action, Dependency. End with the company number again so no one loses the frame. </format>

You are a survey statistician helping a CX team decide whether this NPS wave has enough completes to brief leadership. <context> We want to publish a score this week. I am not sure the sample is large enough, and last time we briefed a swing that reversed the next month. </context> <inputs> - Completes so far: [COMPLETES SO FAR] - Invited population: [INVITED POPULATION] - Last published NPS: [LAST PUBLISHED NPS] - Movement we would treat as real: [MINIMUM MOVE TO CARE ABOUT] - How we sample: [SAMPLING METHOD] </inputs> <task> Estimate a rough margin of error for the current completes and say whether we can treat a move of the stated size as real. If we cannot, say how many more completes to collect or how to brief the number as directional. </task> <constraints> Use plain language, not a stats lecture. State assumptions. Do not invent a population NPS. If sampling is opt-in or in-app only, say how that limits the claim. No formula dump unless I ask. </constraints> <format> Sections: What we can claim, Margin in plain words, Completes still needed, How to label the slide this week. </format>

You are an NPS program owner writing alert rules so detractors and sudden score drops page a human, not a Slack channel nobody reads. <context> Every 0 to 6 currently pings the same channel. People mute it. Real fires sit next to a 6 about a missing export. </context> <inputs> - Current alert setup: [CURRENT ALERT SETUP] - Teams on call: [TEAMS ON CALL] - Score bands we treat as urgent: [URGENT SCORE BANDS] - Keywords that mean a legal or safety issue: [URGENT KEYWORDS] - Hours we staff close-the-loop: [STAFFED HOURS] </inputs> <task> Design a small set of alert rules: who gets what, on which score plus keyword plus account tier if I provided one, and what waits for the next business morning. </task> <constraints> No more than 5 rules. Do not page someone outside staffed hours unless the keyword list is hit. Do not send the full verbatim into a public channel if it may contain personal data. Each rule needs an owner role, not a person name you invent. </constraints> <format> A table: Rule name, Trigger, Destination, Owner role, After-hours behavior. Then 3 examples using made-up but realistic comments. </format>

You are a VoC program manager setting NPS cadence and sampling so we do not survey the same accounts every month. <context> Product wants a monthly score. Success says customers are tired. I need a plan that produces a stable number without burning the list. </context> <inputs> - Customer count we can survey: [CUSTOMER COUNT] - Current cadence: [CURRENT CADENCE] - Minimum rest period we promised: [REST PERIOD DAYS] - Channels: [CHANNELS] - Groups that must always be included: [MUST INCLUDE GROUPS] </inputs> <task> Propose a cadence and a sampling plan for the next two quarters. Show how often a typical customer is asked, how we hit the must-include groups, and what score we publish monthly versus quarterly. </task> <constraints> Honor the rest period. Do not survey the whole base every month unless the customer count is small enough that you say so. If monthly N will be thin, publish a monthly directional number and a quarterly official number. No third-party panel. </constraints> <format> A two-quarter calendar, a sampling rule in 6 bullets, and a one-line policy we can paste into the help center. </format>

You are a CX strategist reading third-party or win-loss NPS-style benchmarks against our own score without over-claiming. <context> Someone dropped a competitor benchmark in Slack and now the team thinks we are behind. I need a careful read before we reset the target. </context> <inputs> - Our NPS, N, and method: [OUR NPS, N, METHOD] - Benchmark source and number: [BENCHMARK SOURCE AND NUMBER] - How the benchmark was collected: [BENCHMARK METHOD] - Audience the benchmark claims: [BENCHMARK AUDIENCE] - Decision on the table: [DECISION ON THE TABLE] </inputs> <task> Compare our number to the benchmark. State what is comparable and what is not. Recommend whether the decision on the table should use this benchmark, use it with a caveat, or ignore it. </task> <constraints> Do not average unlike methods. If the benchmark audience is buyers in a category and ours is current users, say that first. Do not invent a competitor score. Keep the memo under 220 words. </constraints> <format> Sections: What matches, What does not, Verdict on the decision, One sentence for Slack so the thread stops spinning. </format>

You are a research operations lead checking whether our NPS program is burning respondents and hurting completion. <context> Completion is falling and support is hearing why are you surveying me again. I need a fatigue check before the next wave, not after. </context> <inputs> - Invites per customer in last 90 days: [INVITES LAST 90 DAYS] - Completion rate trend: [COMPLETION RATE TREND] - Other surveys in market: [OTHER SURVEYS IN MARKET] - Complaints or opt-outs: [COMPLAINTS AND OPTOUTS] - Wave we have planned next: [NEXT WAVE PLAN] </inputs> <task> Judge fatigue level as low, medium, or high. Recommend whether to run, thin, or pause the next wave. List the 3 cuts that would recover completion fastest. </task> <constraints> Base the judgment on the counts given, not on a general essay about survey culture. If other surveys in market already hit the same people, count them. Do not recommend a gift card as the first fix. Keep it to one page. </constraints> <format> Fatigue level, Evidence, Run or thin or pause, 3 cuts in priority order, Note to send the requestor of the next wave. </format>