Churn Signal
by
Churn prevention for ecommerce teams: find the customers drifting away while they are still reachable, then pick an offer sized to what they are worth.
Churn in ecommerce is silent. Nobody cancels; they simply stop, and the business finds out ninety days later when a report shows a cohort that did not come back. By then the relationship has gone cold and a winback costs three times what an intervention would have cost at the right moment. The signals arrive earlier than that: a lengthening gap between orders, a basket that shrinks, engagement that stops before purchasing does. This skill finds the customers drifting while they are still reachable and picks an offer sized to what each one is worth.
What It Does
- Reads the gap between orders against each customer's own rhythm.
- Detects the shrinking basket that precedes stopping entirely.
- Watches engagement, which falls before purchasing does.
- Separates a customer drifting from one whose cycle is simply long.
- Sizes the offer against lifetime value, not against the last order.
- Picks the save offer per customer, with the value it is protecting.
Use Cases
- Reach a customer before the gap becomes a churn.
- Stop sending winbacks to customers who were never going to lapse.
- Size a save offer against what the relationship is worth.
- Find the engagement drop that arrives before the order stops.
Example Session
Who is about to churn, and what should we send them?
Four hundred and ten customers are drifting. Only 90 need a discount.
- Segment | Customers | Signal | Offer | Value protected
- High value, gap 2x normal | 90 | Order gap | 20% plus early access | $184,000
- High value, basket shrinking | 140 | Basket down 40% | Replenishment reminder | $206,000
- Mid value, engagement stopped | 180 | No opens, still buying | Preference reset | $88,000
- Long cycle, not drifting | 620 | None, normal rhythm | Nothing | n/a
The 620 are the important line: their gap looks alarming against the average and is completely normal for them, so sending a discount would give margin away to people who were always coming back. The shrinking-basket group needs a reminder rather than a discount, because they are still buying and the basket is falling for a reason a reminder usually fixes.
Why This Beats Prompting It Yourself
A Fixed Days-Since-Order Rule vs Each Customer's Own Rhythm: The Hidden Risk
A generic prompt flags everyone who has not ordered in ninety days, which mixes genuinely drifting customers with people whose cycle has always been four months. Discounting the second group costs margin on purchases that were already coming, and it is usually the larger group by some distance.
When You Still Need The Customer Who Needs No Discount
Two of the four segments here get something other than a discount, and one gets nothing at all. A shrinking basket usually responds to a reminder, and an engagement drop among people still purchasing is a preference problem rather than a price one. Treating every churn signal as a discount opportunity is how retention programs erode the margin they are meant to protect.
Known Limitations
The individual rhythm model needs several orders per customer, so first and second-time buyers are handled with cohort-level expectations that the output labels. Engagement signals depend on email and site tracking, which privacy features have made partial. It picks the offers and does not send them.
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/churn-signal/
SKILL.md
references/rhythm-model.md
segments/
# one file per run, with the offer and the value
# 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 lapsed customer segment?
A lapsed segment is a date filter applied to everyone. This compares each customer against their own purchase rhythm, which is the only way to tell a drifting customer from one on a long natural cycle. Those two need opposite treatment.
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 with customer identifiers and dates, product margin, and email or site engagement where available. Engagement improves the early warning and the order rhythm alone does most of the work.
Why send some customers nothing?
Because they are not drifting. Six hundred and twenty customers on this list have a four month cycle and are behaving exactly as they always have. A discount to them is pure margin given away on orders that were coming anyway.
How early is early enough?
The order gap signal fires at roughly twice a customer's normal interval, which is early enough that a light touch still works. Engagement signals fire earlier still and are less reliable, so the skill uses them to prioritize rather than to trigger.
What is included with the skill?
The skill itself, the rhythm model reference, the segment format, and the offer sizing rules. It is free and security scanned. We re-publish the ZIP when the model changes, so download it again if your segments start shifting.
A save offer should cost less than the account it saves.
Polar computes lifetime value on one definition, so the offer you pick is sized against the revenue actually at risk.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from a lapsed customer segment?","acceptedAnswer":{"@type":"Answer","text":"A lapsed segment is a date filter applied to everyone. This compares each customer against their own purchase rhythm, which is the only way to tell a drifting customer from one on a long natural cycle. Those two need opposite treatment."}},{"@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 with customer identifiers and dates, product margin, and email or site engagement where available. Engagement improves the early warning and the order rhythm alone does most of the work."}},{"@type":"Question","name":"Why send some customers nothing?","acceptedAnswer":{"@type":"Answer","text":"Because they are not drifting. Six hundred and twenty customers on this list have a four month cycle and are behaving exactly as they always have. A discount to them is pure margin given away on orders that were coming anyway."}},{"@type":"Question","name":"How early is early enough?","acceptedAnswer":{"@type":"Answer","text":"The order gap signal fires at roughly twice a customer's normal interval, which is early enough that a light touch still works. Engagement signals fire earlier still and are less reliable, so the skill uses them to prioritize rather than to trigger."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the rhythm model reference, the segment format, and the offer sizing rules. It is free and security scanned. We re-publish the ZIP when the model changes, so download it again if your segments start shifting."}}]}</script>