When to send it
Show the link on the page that confirms the cancellation, and add it to the cancellation email. Make it clear that the cancellation is already done and the survey is optional. A survey that stands between a subscriber and the cancel button feels like a trap, and answers given by people trying to escape aren't worth much. Asked afterwards, with nothing at stake, people tend to be more candid.
Add the customer's first name and plan to the link, for example ?name=Asha&plan=monthly-large. The name greets them; the plan lets you see in analytics whether one plan loses more subscribers than another.
Why these questions
The first page asks for the main reason, with a single answer allowed. Most people have more than one reason, but naming the main one gives you a clean count to follow month by month.
The second page changes with that reason. Subscribers who say the price is too high are asked what would have made it work: a smaller plan, less frequent deliveries, paying ahead, or nothing at all. Those who found an alternative are asked what it was. Anyone who disliked recent boxes, had delivery problems or picked "something else" gets an open question about what you could have done better. Each person sees only the follow-up that fits their answer.
The last question asks whether they would consider coming back. It splits former subscribers into those worth contacting later and those to leave alone.
The coupon
The reward is a unique code for one free box or month if they restart. It costs nothing unless the customer returns, and then it removes the biggest barrier to returning. Unique codes matter here, because a free box posted on a deals forum would get expensive quickly. Both end screens show the code, including the one for people who say they won't be back.
What to do with the answers
Count the main reasons every month. If "I wasn't using it enough" leads, the problem may be the delivery schedule rather than the product, and the before you pause survey can help reach those subscribers earlier. Price answers tell you whether a smaller plan would keep people. Note the share who said "Yes, probably" and get in touch when you have something new, or send the former subscriber win-back survey a few months on.