Confirm what stops, when it stops, and what happens to their things. Then get out of the way. The message someone reads on their way out is the one they repeat to other people, and it is written once for thousands of readers who cannot reply to it.
That last part is what makes this copy strange to write. Every other message in this job goes to somebody you can answer. This one goes to a person you will never speak to, at the exact moment they have decided you were not worth it.
Most companies use that moment to make one last attempt. It is the wrong instinct and it is expensive.
The exit is your most-quoted screen
People describe how a cancellation went far more often than they describe how a signup went.
Think about the last time someone told you about a subscription. The story was almost certainly about leaving: the hidden button, the phone call, the four screens of pleading, the charge that appeared anyway. Nobody tells a story about a smooth onboarding.
So the exit copy is doing marketing work whether you designed it to or not. It is being read out loud, in a pub, to people who might have signed up.
Which reframes the whole page. You are not writing to retain this person. You are writing to be described accurately by them later.
Answer the four questions first
Every cancelling user has the same four questions, and if your page does not answer them they will write to support and cost you a real conversation.
- When does it actually stop, as a date, not as "at the end of your current period".
- What happens to the things they made, and how long they have to get them out.
- Whether anything else will be taken from them, stated plainly.
- How to come back, and whether coming back restores their old state.
Give those in the first screenful, before anything persuasive. A page that leads with an offer and buries the date is telling the reader that you would rather they were confused.
Dates beat descriptions here more than almost anywhere. "You have access until the 14th of next month" is a fact somebody can plan around. "Until the end of your billing period" makes them go and work out what that means, and they resent doing arithmetic on your behalf while leaving.
Do not put an obstacle in the sentence
Every extra step you add buys you a small number of retentions and a large number of people who now warn others about you.
The obstacles are well known: the confirmation that shames, the buried link, the support call, the offer that must be declined twice. They do work, in the narrow sense that the numbers move.
What they also do is convert a neutral departure into a grievance. Someone who left easily might come back next year or recommend you to a colleague with different needs. Someone who had to fight will tell that story for a decade, and the story is more memorable than any feature you shipped.
There is a simple test. If your cancellation flow would embarrass you in a screenshot next to your marketing copy, the flow is the truth and the marketing copy is decoration.
One offer, labelled, then silence
You are allowed one alternative, once, clearly marked as optional.
A pause instead of a cancellation is genuinely useful to some people and costs nothing to offer. So is a smaller version of the thing. Put it after the confirmation, not in front of it, and make declining it a single click.
The word that ruins it is "wait". Any copy that opens with "wait", "hold on", or "before you go" reads as a hand on the sleeve. State the option and let it sit there.
And take the answer. An offer that reappears on the next screen is not an offer, it is a negotiation nobody agreed to enter.
Ask one question or none
An exit survey is a favour requested from somebody you have just disappointed, so ask for a small one.
This is where I disagree with common practice. The eight-option radio list plus a free-text box is designed for the analytics team, not for the person filling it in, and it collects the answer people click to make the page go away.
One optional question, placed after the cancellation is already done, gets fewer responses and better ones. "What were you hoping this would do?" beats "reason for leaving" by a distance, because the first invites a sentence and the second invites a category.
Never make it mandatory. A required survey on the way out is a toll, and everyone can tell.
When it was your fault, say so in the message
If they are leaving because something broke, name it on the page rather than in a follow-up email nobody opens.
Most cancellations are quiet and structural: budget, a change of job, a project ending. Some are not, and those people know exactly why they are leaving.
A single honest sentence does more than a discount. "If this is because of the outage in the spring, that is a fair reason and I am sorry" costs nothing and is the only thing on the page that sounds like it came from a person.
Do not attach it to a request. An apology that arrives with a retention offer stapled to it reads as a technique.
Short is allowed
The extension I make has an odd version of this problem, because there are no drafts to hand back when somebody leaves. They are never stored anywhere, so there is nothing to export, nothing to delete, and nothing to apologize about.
My first instinct was that the exit message therefore needed to be padded out into something more substantial. It did not. It needed to say the true thing, which was short, and then stop.
Padding at the exit reads as anxiety, and the reader can smell it. Somebody who has decided to go is scanning for facts, and every sentence that is not a fact is a sentence delaying the one they want.
Write the version you would want to read on the day you cancelled something. Then delete the paragraph you added to make yourself feel better about it, which is the one starting "we would love to have you back".
Write the message you would want to get
Unvent rewrites what you are typing, in place, in the tone you pick, so copy that has to be honest and brief still sounds human. Useful for the messages people read when they are already annoyed.
Add Unvent to Chrome