thinkhard.SI

Prompt repository Customer support & success

Follow up on a low satisfaction or NPS score

A customer gave a low rating in a survey and you want to find out what went wrong and fix it, not argue with the score.

Registry no.
SUP-027
Collection
Customer support & success
Laid out for
Any chat model
Score
10 out of 10
Blanks to fill
7
Added

Customer support & successAny chat

The prompt

Score 10 of 10, Excellent

Use it when A customer gave a low rating in a survey and you want to find out what went wrong and fix it, not argue with the score.

  • {{survey}}
  • {{score}}
  • {{comment}}
  • {{records}}
  • {{what_i_can_do}}
  • {{sender_name}}
  • {{root_cause_team}}
Full prompt and why it works
Prompt
Write a follow-up to a customer who gave us a low score in our {{survey}}. Their score was {{score}}. The aim is to understand what went wrong and fix what we can, so that they feel heard; it is not to ask them to change their score.

Their comment, if any:
"""
{{comment}}
"""

What our records show (recent tickets, orders, account notes):
"""
{{records}}
"""

What I can offer or do: {{what_i_can_do}}

Write:
1. A message of 70 to 140 words of plain text that does four things in this order: thanks them for the honest feedback, names the specific problem if their comment or our records show it, says what we have done or will do about it, and invites them to tell us more by reply or a short call. Sign it "{{sender_name}}".
2. Below it, an internal note starting "FOR OUR RECORDS:" with the likely cause from the records, whether the issue is fixed, and anything that should go to {{root_cause_team}} so it does not happen to others.

The message should feel personal and specific, as if written by one person who read their comment. For example, an opening with the right specificity: "You mentioned waiting four days for a reply to your billing question, and you're right, that's too long."

Do not ask them to raise or update their score, do not offer anything not listed in what I can do, and do not explain the problem away or blame another team. Avoid stock phrases such as "we value your feedback". Do not quote their score back unless they mentioned it themselves.

If there is no comment and the records do not show a clear cause, do not guess one: write a shorter message with one open question about what we could have done better. If the records show the problem is still open, lead with what is being done about it and by when.

Variables

{{survey}}
Which survey, for example "post-ticket satisfaction survey" or "quarterly NPS survey".
{{score}}
The score they gave, for example "2 out of 10".
{{comment}}
Their survey comment, word for word, or "none".
{{records}}
Recent tickets, orders or account notes for this customer.
{{what_i_can_do}}
What you can actually offer or fix, or "apology and follow-up only".
{{sender_name}}
The name the message is signed with.
{{root_cause_team}}
Who should fix the cause, for example "the billing team" or "the support team lead".

Why it works

  • Who it's for. The message exists to learn and fix, not to rescue a number, which is why asking the customer to change their score is ruled out.
  • Background facts. It reads the customer's records alongside the comment, so the reply can name the actual problem instead of apologizing in general.
  • If unsure. With no comment and no clear cause in the records, it asks one open question rather than guessing what went wrong.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/low-satisfaction-score-follow-up

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "low-satisfaction-score-follow-up"}, or just ask for it by name.