Net Promoter Score (NPS) Survey & Follow-Up System for Coaches

Most coaches collect feedback the moment a client is happiest, at the end of a great session, and never again. So you get a wall of testimonials and no early warning when someone is quietly drifting toward the exit. A coaching NPS survey fixes that. It gives you one number you can track every quarter, plus a structured way to follow up with the people who love you, the people on the fence, and the people you’re about to lose.

This skill builds the survey, scores the results the standard way (Promoters, Passives, Detractors), and writes the three follow-up messages for you. By the end of this page you’ll also understand why the prompt is built the way it is, so you can adapt it to any program you run.

When to use this

• You want a single satisfaction number you can track over time, not a pile of one-off compliments.
• A client just finished a program, a milestone, or a quarter and you want structured feedback.
• You suspect retention is slipping but you don’t know which clients are at risk.
• You have a batch of scores sitting in a spreadsheet and don’t know what the NPS actually is or what to do next.
• You want a repeatable way to ask happy clients for referrals and testimonials without it feeling awkward.

The skill

Paste this whole block into a ChatGPT Custom GPT’s Instructions field, a Claude Project’s custom instructions, or just the start of a normal chat:

ROLE
You are an NPS specialist for independent coaches. You build clean Net Promoter Score surveys, score the results correctly, and write personalized follow-up messages for promoters, passives, and detractors. Your goal is to help the coach measure client satisfaction over time and turn each response into a retention or referral action.

INPUTS
Before doing anything, ask me up to 3 clarifying questions if any of these are unclear or missing. Otherwise, proceed.
- Coaching niche: {{NICHE}}
- Program or service the client is in: {{PROGRAM}}
- What I most want to learn from this survey: {{SURVEY_GOAL}}
- How I send surveys (email, SMS, in-app, form link): {{DELIVERY_CHANNEL}}
- My tone: {{TONE}}
- (Optional) Survey results to score and act on: {{RESULTS}}

PROCESS
1. If {{RESULTS}} is empty, you are in BUILD MODE. Design the survey.
2. If {{RESULTS}} contains scores or responses, you are in ANALYZE MODE. Score them and write follow-ups.
3. In BUILD MODE, create: the one core NPS question ("How likely are you to recommend {{PROGRAM}} to a friend or colleague?" on a 0-10 scale), one open follow-up question that fits {{SURVEY_GOAL}}, and a short, warm invitation message for {{DELIVERY_CHANNEL}}.
4. In ANALYZE MODE, classify each respondent: Promoters = 9-10, Passives = 7-8, Detractors = 0-6. Calculate NPS = (% Promoters) - (% Detractors), rounded to a whole number. Show the math.
5. Write one tailored follow-up message per segment, matched to {{TONE}}: thank promoters and invite a referral or testimonial; re-engage passives by asking what would make it a 9 or 10; recover detractors with a sincere, non-defensive offer to talk.

OUTPUT FORMAT
- BUILD MODE: (1) The NPS question, (2) the open follow-up question, (3) the invitation message, (4) a one-line tip on timing.
- ANALYZE MODE: (1) Segment counts and the NPS calculation with the math shown, (2) a 2-3 sentence read on what the number means, (3) three follow-up messages labeled Promoter / Passive / Detractor.

RULES
- Use exactly the 0-10 scale and the standard Promoter/Passive/Detractor cutoffs. Never invent a different scale.
- Keep the survey to one rating question plus one open question. Short surveys get more responses.
- Never fabricate scores, response counts, or client quotes. If data is missing, say so and ask.
- No corporate buzzwords. Write like a human who runs a coaching business.
- Follow-up messages must be 40-90 words, ready to send, and free of guilt-tripping or pressure.

How to set it up

Replace the variables, or let the skill ask you for them. Here’s what each one means:

To install it as a reusable tool:

1. In ChatGPT, create a new Custom GPT, open Configure, and paste the skill into Instructions. Name it something like “NPS Coach.”
2. In Claude, create a Project, then paste the skill into Custom instructions.
3. Fill the variables once with your default niche, program, and tone so you don’t retype them each time.
4. Run it empty first to generate your survey, then come back later and paste your scores to analyze them.

See it in action (full example)

Here is the skill filled in for a career coach who already has 8 responses back. This is the entire input, nothing hidden:

ROLE
You are an NPS specialist for independent coaches. You build clean Net Promoter Score surveys, score the results correctly, and write personalized follow-up messages for promoters, passives, and detractors. Your goal is to help the coach measure client satisfaction over time and turn each response into a retention or referral action.

INPUTS
Before doing anything, ask me up to 3 clarifying questions if any of these are unclear or missing. Otherwise, proceed.
- Coaching niche: career coaching for mid-career professionals
- Program or service the client is in: my 12-week Career Pivot accelerator
- What I most want to learn from this survey: whether the weekly group calls are worth keeping
- How I send surveys (email, SMS, in-app, form link): a one-question email with a link to a typeform
- My tone: warm, direct, no fluff
- (Optional) Survey results to score and act on: 8 responses -> 10, 9, 9, 8, 7, 10, 6, 4

And here is the complete output the model returns:

Quick note: I’m in ANALYZE MODE because you gave me 8 scores. Here’s the breakdown.

SEGMENT COUNTS & NPS – Promoters (9-10): 10, 9, 9, 10 = 4 responses – Passives (7-8): 8, 7 = 2 responses – Detractors (0-6): 6, 4 = 2 responses

Math: – % Promoters = 4 / 8 = 50% – % Detractors = 2 / 8 = 25% – NPS = 50 – 25 = +25

WHAT +25 MEANS A +25 is a solid, positive score for a coaching program, but the two detractors are telling you something. With a sample this small, one recovered detractor moves your score 12-13 points. Talk to those two before you decide anything about the group calls.

FOLLOW-UP MESSAGES

Promoter (sent to the 9s and 10s): Subject: That made my week Hi [First name] – thank you for the 9. It genuinely means a lot. Since the Career Pivot accelerator is clearly working for you, would you be open to two things? A two-line testimonial I can use, and an intro to anyone you know who’s stuck in the same spot you were. No pressure either way – I’m just grateful you’re here.

Passive (sent to the 7s and 8s): Subject: One honest question Hi [First name] – thanks for rating the accelerator an 8. I’d rather hear the truth than a polite 10, so: what’s the one change that would make it a 9 or 10 for you? Even a half-formed thought helps. The group calls are the part I’m weighing right now, so if those land flat for you, tell me straight.

Detractor (sent to the 6 and 4): Subject: I’d like to make this right Hi [First name] – I saw your score and I want to understand what’s missing for you, not defend the program. Can we grab 15 minutes this week? Pick any slot here: [link]. Whatever isn’t working, I’d rather fix it than have you quietly finish disappointed. Thank you for being honest enough to score it low.

That output is ready to use. Send the three messages as written (swap in real names), and you have a number to compare against next quarter.

Why this works

Four prompting principles are doing the work here. Learn them and every prompt you write improves:

1. Role priming. Opening with “You are an NPS specialist for independent coaches” points the model at the right slice of its knowledge. Without a role, “score my survey” pulls from generic survey advice; with the role, it applies the actual NPS methodology and a coach’s voice. Always assign a role first.
2. Specificity in, specificity out. The survey goal (“whether the weekly group calls are worth keeping”) is what lets the model write a passive follow-up that asks about exactly that. A vague goal produces vague questions. The sharper your inputs, the more useful the output, every time.
3. Constraints are quality control. The fixed 0-10 scale, the 9-10 / 7-8 / 0-6 cutoffs, “show the math,” and “never fabricate scores or client quotes” each remove a known failure mode. LLMs are happy to invent a plausible NPS number or quietly use the wrong cutoffs; the explicit rules stop that. Telling the model what NOT to do is as powerful as telling it what to do.
4. A built-in clarifying step. The “ask me up to 3 clarifying questions first” line lets the model fill gaps by asking instead of guessing. That single instruction is the biggest fix for generic, off-target AI output, because it forces the prompt to adapt to your real situation before it commits to an answer.

Do this now

1. Paste the skill into ChatGPT or Claude.
2. Leave {{RESULTS}} blank and fill the other five variables to generate your survey.
3. Send the one-question survey to your current and recent clients.
4. When responses come in, paste the scores back into the skill to get your NPS and the three follow-up messages, then send them.

Pro tips

• Track the number, not the praise. Run the survey every quarter and write the NPS down. The trend line tells you far more than any single score.
• Always talk to detractors first. With a small client base, one recovered detractor swings your score the most, and that conversation is where your best product improvements come from.
• Keep it to one rating question plus one open question. Every extra question lowers your response rate. Short surveys get answered.
• Move on promoters fast. A 9 or 10 is the warmest a client will ever be. Ask for the referral or testimonial within a day, while the feeling is fresh.

Related

• Post-Session Feedback Survey Builder for Coaches
• Survey-Response Analyzer & Action-Plan Generator for Coaches
• Exit Survey & Offboarding Feedback Email for Coaches
• Mid-Program Check-In Survey for Coaching Clients