Handle a Return Request Without Losing the Customer
Writing a nice reply to one message does not solve a return, because a return is not a message, it is a process: first you have to understand what actually happened, then pick the right compensation, then decide whether you are keeping that customer long-term. Most sellers collapse all three steps into one reflex, either refunding instantly or digging in and refusing. But the right question is never 'how much should I spend on this customer,' it is 'what is this customer worth to me going forward.' This prompt is not a one-message answer. It builds an end-to-end protocol that runs from diagnosis to a graduated recovery ladder to a final decision.
You are an operations consultant who specializes in customer experience and return handling. Your expertise is not writing one emotionally sound reply, it is building an end-to-end return protocol that runs from diagnosis to decision. Your job is NOT to produce a nice-sounding message; it is to produce a workable decision system for which compensation gets offered when, and to which customer how much generosity is owed. Most sellers treat a return as a single snap judgment: either refund or dig in and refuse. But the same complaint points to a different decision depending on the customer. Someone returning their first order and someone hitting a problem on their tenth order do not deserve the same treatment. You are the one who sees that difference and turns it into a system. I will give you the following: - The customer's message: [paste it as is] - Product cost: [in dollars] - My real options: [full refund / partial refund / exchange / coupon / free reshipment, whichever apply] - Customer history: [first order or which order number, any past issues] Return the output in exactly this order, with these section headers: 1) DIAGNOSIS: What is the real type of problem (product defect, shipping damage, mismatched expectations, buyer indecision); explain why this diagnosis matters, since different problem types call for different compensation. 2) RECOVERY LADDER: 2-3 sequential steps starting from the cheapest one that will genuinely satisfy the customer (e.g. explanation plus a small gesture first, exchange if that fails, refund if that fails too). 3) WHEN TO REFUND VS. WHEN TO EXCHANGE OR OFFER A COUPON: The decision rule, which diagnosis and which customer profile earns which step. 4) CUSTOMER LIFETIME VALUE NOTE: A short justification for how generous to be, based on this customer's order frequency. The cheapest short-term fix and the solution that keeps the customer long-term are usually different; a small extra cost for a repeat customer is usually cheap once you weigh their future orders. For a one-time or unreliable buyer, the same generosity may not be warranted, draw that line clearly. 5) REPLY DRAFT: A ready-to-send message to the customer that opens by acknowledging the frustration and follows with a clear, timed resolution. 6) INTERNAL NOTE: A short summary of this case for future reference if the same type of request comes in again (is it recurring, is it becoming a pattern). RULES: - Do not use made-up statistics, fake numbers like "X% of customers were satisfied" that you cannot back up. - Do not promise anything you are not authorized to give; work only with the real options provided. - If you suspect fraud, flag it separately, do not default to the most generous option on a good-faith assumption. - The reply draft has to be concrete and timed, avoid vague lines like "we'll look into it." - Write the output in English.
How to use it
- Paste the return or complaint message as is, then give the product cost, your real options (refund, exchange, partial refund, coupon), and how many orders this customer has placed with you before.
- Run the prompt and follow the diagnosis and recovery ladder in the output: confirm what kind of problem it actually is first, then start from the cheapest step that will still satisfy the customer, and only escalate if you have to.
- After you apply the decision, tag that customer (repeat buyer or one-time) and save the outcome as a short note; the next time a similar request comes in, these notes give you a ready-made pattern.
Example / tip
Input: Customer message, the item arrived broken in shipping and this is the second time, product cost $12, options are free reshipment, full refund, or a coupon, customer history is their 4th order. AI output (shortened): Diagnosis, recurring shipping damage, not a product defect; Recovery ladder, start with free reshipment plus a complaint filed with the carrier, if the customer will not accept that move to a full refund; Decision, for a customer on their 4th order, weigh long-term value over short-term cost and be generous.
Usage example
Copy the prompt as is, fill in the bracketed fields for your own business, and paste it into ChatGPT or Claude. The more concrete your input, the more useful the output. Take the first draft, ask for a one-sentence fix on whatever you don't like, then give it a final pass in your own voice.
When not to use it
For high-value returns or ones that risk turning into a legal dispute (bulk orders, business customers, damage claims), do not treat this prompt's recommendation as the final call, take it to your accountant or legal counsel instead. Also, if you suspect fraud (the same customer filing repeated unfounded return claims), this prompt runs on a good-faith assumption, so evaluate that suspicion separately first.
Output quality checklist
- Does the diagnosis step actually isolate the root cause, or does it jump straight to a solution
- Is the recovery ladder ordered from cheapest to most expensive, or does it start with the priciest option for no reason
- Does customer history (repeat buyer or not) actually shape the decision
- Is the reply draft concrete and timed, or does it close with a vague 'we'll look into it'
