The Win Back Sequence
Four emails for the customers who bought once, went quiet, and are still on your bill.
Last week you built the thirty days after the sale. This week you go back further, to the customers who bought before you had any of this and then went quiet.
They are still on your list. You are still paying to keep them there.
The Win Back Sequence is six prompts and one file called win-back.md. Prompt 1 writes nothing. It interviews you until it can answer two questions you have probably never looked up: how often a happy customer actually buys from you again, and what your email tool charges at the tier below the one you are on. Those two answers set the whole campaign. The first one tells you when quiet starts for your business instead of some blog’s 90 day default, and the second one puts a dollar figure on the people you are storing.
Then it builds four emails over sixteen days. Day 0 names the one thing that has changed since they bought. Day 4 asks a single question with no link and no offer anywhere in it. Day 9 makes one offer, to non-responders only, with a deadline that has a real thing happening at the end of it. Day 16 tells them plainly that you are taking them off the list, and gives them one button to stay.
Then it stops. This is a campaign, not a flow that runs forever.
The email with nothing to sell
Day 4, under 80 words, one question, no link in the body and none in the signature.
The question offers them two answers, and the unflattering one has to be the easy one to admit. “What did you think?” gets nothing. “Did it not fit what you needed, or did you just never get around to setting it up?” gets a one-line reply from somebody in a parking lot, because you already wrote their answer for them and all they have to do is pick.
What comes back is a list of the real reasons people stop buying from you, written by the people who stopped, in their words. Prompt 6 reads those replies thirty days later, groups them, counts them, and holds the top reason up against what your sales page promises. When those two do not match, you have found the leak, and you found it in your own inbox instead of paying for a focus group of strangers who have never bought anything from you.
Why the manual way costs you
Retention email is the piece small businesses hand to an agency, and the retainer starts before a single sentence gets written. Typical 2026 retainers start at $1,000 to $3,000 a month for a small business with straightforward needs, which is $12,000 to $36,000 a year out of your pocket, and the agency’s first task is to interview you about why your customers leave. You pay the retainer and you still supply the answer.
Meanwhile you keep paying for the people you are not emailing. Mailchimp’s Standard plan charges you about $20 a month at 500 contacts, checked today on the published pricing, and it climbs with every contact you store, not every contact who reads anything. That is $240 a year at the smallest tier for a list where a chunk of the names have not opened anything since the year they bought, and the number goes up the moment your stored count crosses the next line.
Then there is the figure that should sting the most. Triple Whale’s 2026 benchmark, drawn from more than 40,000 brands, puts the median cost per purchase on Meta at $38.99, with ecommerce closer to $45. You are bidding roughly $39 to introduce yourself to a stranger while a person who already handed you a credit card sits in your own database, ignored, costing you storage.
Running all six prompts end to end moves roughly 30,000 input tokens and 15,000 output tokens. At Claude Sonnet 5 pricing of $2 per million input and $10 per million output, verified today, that is 6 cents in and 15 cents out. Twenty-one cents against a $1,000 retainer, and inside a Claude subscription you already pay for, it is $0 in new tools. The campaign also ends with a list you cut, which makes it the only prompt pack in the Vault that finishes with a bill going down.
The guru version of this is a “we miss you” template with a 20 percent coupon in it, sold as one module inside a course. That template is not a campaign. It is a discount with a sad face on it, and the customers who take it learn to wait for the next one.
How to use it
- Run Prompt 1 today. Twenty minutes, and most of that is you opening two tabs: your order export, to count how long your real repeat cycle is, and your email tool’s pricing page, to find the tier below yours. If the prompt tells you that nothing has actually changed since these people bought, believe it and go fix one thing first. An email that says only “we miss you” is asking a stranger for attention as a favor.
- Run Prompt 2 for the first and last emails, then Prompts 3 and 4 for the two that carry the weight. When Prompt 4 hands you three ranked offers with what each one costs you per redemption, pick by the second column, which is whether it trains people to wait for a discount. Any offer you cannot make again in six months is the wrong offer.
- Run Prompt 5 in a fresh chat the day before it sends, then load the four into whatever already sends your email and place yourself in the segment so you watch all four arrive the way a quiet customer does.
- Put Prompt 6 on your calendar for thirty days after the last send, and keep the promise Email 4 made. Day 17 is when the names come off. A business that threatens to leave and then keeps emailing has taught its list that nothing it sends is real, and your next launch pays for that.
Ten drops built the offer, the page, the launch week, and the thirty days after the sale. This one goes and gets the customers you already earned and then lost quietly, and it is the only one that ends with your monthly bill going down instead of up. Recipes with receipts. Run it this week, and by the end of the month you will know the actual reason people stop buying from you, in their own sentences. Either it makes you money or it costs you nothing.
Next Monday’s Tool Teardown: verdict as always, one real test, keep or replace.
Next Friday’s drop: the Referral Ask Kit. Three prompts and one text message that turn a customer who just got a result into the next customer, without paying Meta for the introduction.