Subscription Save
by
Subscription cancellation decisions for ecommerce teams: value the remaining lifetime of each subscriber, then offer a skip, a pause, a swap or a discount sized to what is actually at risk.
Cancellation flows offer the same thing to everyone, usually a discount, usually the deepest one the business has approved. That is expensive in two directions. A subscriber canceling because the deliveries are stacking up does not need a discount, they need a skip, and handing them money trains a habit that costs every cycle afterwards. And a subscriber worth twelve more months gets the same offer as one worth two. This skill values the remaining lifetime of each subscriber, reads the reason behind the cancellation where it can, and picks the save: skip, pause, swap, or a discount sized to the value at risk.
What It Does
- Estimates remaining subscription lifetime and value per subscriber.
- Separates a cadence problem from a price problem from a product problem.
- Matches the save to the reason: skip, pause, swap product, or discount.
- Sizes any discount against the remaining value rather than a fixed policy.
- Flags subscribers whose save would cost more than the revenue it retains.
- Tracks which saves held past the next two cycles rather than just the moment.
Use Cases
- Stop discounting subscribers whose real problem is delivery frequency.
- Offer a pause where a cancellation is seasonal rather than final.
- Let go of the subscribers whose save costs more than they are worth.
- See which saves actually held two cycles later.
Example Session
Someone is canceling their monthly subscription. What should we offer?
Four cancellations this morning. Only one of them should be offered a discount.
- Subscriber | Cycles | Remaining value | Signal | Offer
- 14 months | 14 | $310 | Two skips already used | Change cadence to 8 weeks
- 3 months | 3 | $95 | Canceled after price change | 15% for 3 cycles
- 22 months | 22 | $480 | Seasonal, cancels each winter | Pause until March
- 2 months | 2 | $28 | Never opened, no engagement | Let go
The first subscriber has used two skips already, which is a cadence signal rather than a price one: moving them to eight weeks keeps $310 of value and costs nothing. The last is worth $28 and any save offer costs more than that, so let them go cleanly. Only the second is actually leaving over price, and three cycles at 15% is the smallest offer that covers it.
Why This Beats Prompting It Yourself
One Save Offer vs The Reason For Leaving: The Hidden Risk
A generic prompt writes the cancellation flow copy and attaches the discount, because that is what save flows do. It has no way to tell that two skips already used is a cadence problem, that a winter cancellation every year is seasonal, or that a two month subscriber is worth less than the offer. One offer for everyone is expensive at both ends of the list.
When You Still Need The Value At Risk
A save is only worth making up to what the subscription is worth. This skill states $310 of remaining value against a cadence signal, and $28 against a subscriber worth letting go. It also follows the save two cycles out, so a discount that merely delayed the cancellation by a month is recorded as the failure it was rather than counted as a win.
Known Limitations
Remaining lifetime needs enough subscription history to model, so a program under a year old produces wider estimates and the output says so. Reason detection reads behavioral signals and any reason the customer gave, which means an unexplained cancellation from an engaged subscriber will sometimes get a generic recommendation. It picks the offer and does not modify the subscription in your platform.
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/subscription-save/
SKILL.md
references/save-rules.md
saves/
# one file per run, saves tracked two cycles out
# 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 cancellation flow?
A flow offers the same thing to everyone who reaches it. This reads the signal behind the cancellation, values the remaining subscription, and picks between a cadence change, a pause, a product swap and a discount. A large share of cancellations are better served by something that costs nothing.
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.
Why would it let a subscriber go?
Because a subscriber worth $28 of remaining value is not worth a save offer that costs more than that, and keeping them on a discount they will cancel from anyway just moves the loss. Letting go cleanly is a decision with a number behind it here, not a failure.
What does it need connected to work?
Subscription history with cycles, skips, pauses and cancellations, plus order value. With engagement data from your ESP it reads the signals more accurately. Without it, cadence and price signals still work and the output notes the reduced confidence.
Does it track whether the save worked?
Yes, two cycles out. A discount that delays a cancellation by one cycle is not a save, and counting it as one is how save flows keep getting more expensive without retaining anyone. The follow-up is what makes the offer sizing improve over time.
What is included with the skill?
The skill itself, the save rule reference, the offer templates per signal, and the two-cycle tracking format. It is free and security scanned. We re-publish the ZIP when the save model changes, so download it again if your offer mix shifts.
Skip, pause and discount each cost a different amount.
Polar values a subscriber's remaining lifetime across 45+ sources, so the save you offer stays proportionate to what it retains.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from a cancellation flow?","acceptedAnswer":{"@type":"Answer","text":"A flow offers the same thing to everyone who reaches it. This reads the signal behind the cancellation, values the remaining subscription, and picks between a cadence change, a pause, a product swap and a discount. A large share of cancellations are better served by something that costs nothing."}},{"@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":"Why would it let a subscriber go?","acceptedAnswer":{"@type":"Answer","text":"Because a subscriber worth $28 of remaining value is not worth a save offer that costs more than that, and keeping them on a discount they will cancel from anyway just moves the loss. Letting go cleanly is a decision with a number behind it here, not a failure."}},{"@type":"Question","name":"What does it need connected to work?","acceptedAnswer":{"@type":"Answer","text":"Subscription history with cycles, skips, pauses and cancellations, plus order value. With engagement data from your ESP it reads the signals more accurately. Without it, cadence and price signals still work and the output notes the reduced confidence."}},{"@type":"Question","name":"Does it track whether the save worked?","acceptedAnswer":{"@type":"Answer","text":"Yes, two cycles out. A discount that delays a cancellation by one cycle is not a save, and counting it as one is how save flows keep getting more expensive without retaining anyone. The follow-up is what makes the offer sizing improve over time."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the save rule reference, the offer templates per signal, and the two-cycle tracking format. It is free and security scanned. We re-publish the ZIP when the save model changes, so download it again if your offer mix shifts."}}]}</script>