Win-Back
by
Lapsed customer decisions for ecommerce teams: value each lapsed cohort by what it is actually worth, then trigger a win-back, a sunset sequence, or suppression instead of mailing everyone the same offer.
Lapsed customers get treated as one group and mailed one offer, usually a discount, usually to everybody who has not bought in ninety days. Two problems follow. Most of that list was never going to come back and its engagement decay is now dragging your deliverability down with it. And the part that would have come back gets handed a discount it did not need. This skill segments lapsed customers by lifetime value, purchase pattern and engagement decay, estimates what each cohort is worth reviving, and triggers the right treatment per cohort: a win-back with an offer sized to the value, a sunset sequence, or straight suppression.
What It Does
- Segments lapsed customers by lifetime value, purchase cadence and engagement decay.
- Estimates the revenue realistically recoverable from each cohort.
- Sizes the offer against cohort value rather than applying one discount to all.
- Separates genuinely lapsed customers from ones simply outside their normal cycle.
- Routes each cohort to a win-back, a sunset sequence, or suppression.
- Protects sending reputation by keeping dead segments out of the send.
Use Cases
- Stop discounting customers who were going to reorder on their own cycle anyway.
- Recover deliverability by suppressing the segment that has not opened in a year.
- Give a high value lapsed cohort an offer that matches what it is worth.
- Run the lapsed review quarterly on the same definition each time.
Example Session
Who in our lapsed list is worth a win-back, and what should each group get?
Your lapsed list is 41,000 contacts. Treating it as one group is costing you on both ends.
- Cohort | Size | Avg LTV | Last open | Recoverable | Treatment
- High value, still opening | 2,900 | $340 | 26 days | $71k | Win-back, no discount
- Mid value, cycle overdue | 6,400 | $120 | 68 days | $48k | Win-back, 10%
- Low value, decayed | 19,800 | $45 | 210 days | $11k | Sunset sequence
- No open in 12 months | 11,900 | $38 | Never | $0 | Suppress
The first cohort is still opening and buys at $340: send a win-back with no discount at all, since the reason they stopped is not price. Suppressing the last 11,900 costs you nothing in recoverable revenue and should lift inbox placement for every send that follows, which is worth more than the segment ever was.
Why This Beats Prompting It Yourself
One List vs Cohorts With A Value: The Hidden Risk
A generic prompt writes a win-back email. Ask how to re-engage lapsed customers and you get a campaign aimed at a single undifferentiated list, with a discount attached because a discount is what win-back emails have. It cannot tell you that a fifth of that list is still opening and does not need the offer, or that half of it should never be mailed again. Those are the two decisions that matter.
When You Still Need The Value Behind The Segment
Reviving a customer is worth doing only up to what that customer is worth. This skill states $340 of lifetime value against a cohort still opening at 26 days, and $38 against one that has never opened, and treats them completely differently. Suppression is treated as a decision with a number behind it rather than as list hygiene, because deliverability lost on dead contacts is paid for by every send afterwards.
Known Limitations
Lifetime value needs order history deep enough to be meaningful, so a store under a year old will produce cohorts the skill labels as low confidence. Cycle detection assumes a repeatable purchase cadence, which works for consumables and understates lapse for considered purchases bought once a year. It decides the treatment per cohort and writes the segment rules, but the sequences themselves are built in your ESP.
Install It
- Download the ZIP.
It is free and there is no account to create. - Unzip it into your agent's skills folder.
Claude Code reads ~/.claude/skills/, which is hidden by default: the command in the folder block opens it. Other agents scan their own directory, so drop the same folder there instead. - Ask your agent to use it.
Restart the agent if it was already running, then it picks the skill up with no config.
skills folder
Copy
~/.claude/skills/win-back/
SKILL.md
references/cohort-rules.md
segments/
# segment rules, one file per review
# macOS: create the folder and open it in Finder
mkdir -p ~/.claude/skills && open ~/.claude/skills
# Windows: paste in the Explorer address bar
%USERPROFILE%\.claude\skills\Frequently asked questions
How is this different from a win-back campaign template?
A template gives you the email. This decides who should receive one at all. It values each lapsed cohort, sizes the offer against that value, and sends a large part of most lists to suppression rather than to a campaign. The email is the easy half of the problem.
Which agents does the skill run in?
Any agent that supports the open SKILL.md format: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Manus, Grok and others load it unmodified. The format is portable, the location is not. Each agent scans its own skills directory, so you drop the same folder into whichever one yours uses. There is no config file to edit and no API key to provision.
What does it need connected to work?
Order history for lifetime value and cadence, plus engagement data from your ESP for decay. With orders alone it still segments on value and says that the decay side is missing, which usually means it will recommend fewer suppressions than it should.
Why would it recommend a win-back with no discount?
Because a cohort that is still opening your email and buys at a high lifetime value did not stop for price. Handing that group a discount trains a margin habit and recovers revenue you were likely to get anyway. The offer is sized to the reason for lapsing, not to the fact of it.
Is suppression not just losing customers?
It is losing contacts who have not opened in a year and are actively suppressing your inbox placement for everyone else. The skill puts a recoverable revenue figure against each cohort, and where that figure is near zero the cost of continuing to mail them is not.
What is included with the skill?
The skill itself, the cohort rule reference, the segment definitions, and the suppression thresholds. It is free and security scanned. We re-publish the ZIP when the cohort model changes, so download it again if your segment sizes shift without a reason.
Win-back, sunset or suppress is a cost per contact decision.
Polar joins lapsed cohorts to lifetime value in one semantic layer, so you spend on the ones worth reviving.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from a win-back campaign template?","acceptedAnswer":{"@type":"Answer","text":"A template gives you the email. This decides who should receive one at all. It values each lapsed cohort, sizes the offer against that value, and sends a large part of most lists to suppression rather than to a campaign. The email is the easy half of the problem."}},{"@type":"Question","name":"Which agents does the skill run in?","acceptedAnswer":{"@type":"Answer","text":"Any agent that supports the open SKILL.md format: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Manus, Grok and others load it unmodified. The format is portable, the location is not. Each agent scans its own skills directory, so you drop the same folder into whichever one yours uses. There is no config file to edit and no API key to provision."}},{"@type":"Question","name":"What does it need connected to work?","acceptedAnswer":{"@type":"Answer","text":"Order history for lifetime value and cadence, plus engagement data from your ESP for decay. With orders alone it still segments on value and says that the decay side is missing, which usually means it will recommend fewer suppressions than it should."}},{"@type":"Question","name":"Why would it recommend a win-back with no discount?","acceptedAnswer":{"@type":"Answer","text":"Because a cohort that is still opening your email and buys at a high lifetime value did not stop for price. Handing that group a discount trains a margin habit and recovers revenue you were likely to get anyway. The offer is sized to the reason for lapsing, not to the fact of it."}},{"@type":"Question","name":"Is suppression not just losing customers?","acceptedAnswer":{"@type":"Answer","text":"It is losing contacts who have not opened in a year and are actively suppressing your inbox placement for everyone else. The skill puts a recoverable revenue figure against each cohort, and where that figure is near zero the cost of continuing to mail them is not."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the cohort rule reference, the segment definitions, and the suppression thresholds. It is free and security scanned. We re-publish the ZIP when the cohort model changes, so download it again if your segment sizes shift without a reason."}}]}</script>