Refund Pattern
by
Refund analysis for ecommerce teams: find the reason that is growing rather than the one that is largest, then route it to the team that can fix it.
Refund reporting stops at a rate, and a rate that has been stable for a year hides everything interesting underneath it. One reason can double while another halves and the total never moves. The reason that is growing is the one worth attention, because it is a new problem with a findable cause: a supplier batch, a description change, a carrier whose service degraded in one region. The largest reason is usually structural and has been there forever. This skill finds the reason that is growing and routes it to the team that can fix it.
What It Does
- Tracks refund reasons by trend rather than by size.
- Finds the reason that is growing inside a stable overall rate.
- Links a growing reason to the change that caused it, where one exists.
- Prices each reason in cash, including both shipping legs.
- Separates a product problem from a description or a delivery problem.
- Routes each pattern to an owner, with the cash and the evidence.
Use Cases
- Find the refund reason doubling inside a flat overall rate.
- Link a spike to the supplier batch or copy change that caused it.
- Route a refund pattern to the team that can actually fix it.
- See what each reason is costing in cash rather than in rate.
Example Session
Our refund rate is flat but something feels off. What is happening?
The rate is flat because two reasons moved in opposite directions.
- Reason | Share now | Share 6mo ago | Cash per quarter | Owner
- Not as described | 14% | 6% | $62,000 | Merchandising
- Damaged in transit | 9% | 18% | $41,000 | None, already fixed
- Changed mind | 41% | 40% | $186,000 | Structural
- Wrong size | 28% | 29% | $124,000 | Merchandising, ongoing
Not as described has more than doubled and traces to a photography change on eleven SKUs in March: the new images are lit differently and the color reads wrong. That is $62,000 a quarter and a photography fix. Damaged in transit halved because of a packaging change, which is worth knowing and needs no action. Changed mind is the largest and has not moved in years, which makes it structural rather than a problem.
Why This Beats Prompting It Yourself
A Stable Refund Rate vs The Reason Moving Underneath: The Hidden Risk
A generic prompt reads the overall rate, sees it flat, and reports that returns are under control. Underneath, one reason has doubled while another halved. The flat line is an average of a new problem and a solved one, and reporting it as stability means the new problem goes unexamined for another quarter.
When You Still Need The Change That Caused It
A growing reason is only actionable once you know what changed. Matching the increase against a photography change on eleven specific SKUs in March turns a return statistic into a work order for a specific team. Without that link, the finding is an observation that everyone agrees is concerning and nobody owns.
Known Limitations
Reason quality depends on what your returns flow captures, and customers frequently select the most convenient option rather than the accurate one, so free text is weighted where available. Linking a spike to a cause is correlation over timing and affected SKUs, strong enough to investigate and not proof. It routes the patterns and does not create tickets in your systems.
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/refund-pattern/
SKILL.md
references/reason-taxonomy.md
patterns/
# one file per run, with the owner and the cash
# 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 returns report?
A report gives you rates by reason. This reads the trend in each reason, finds the ones moving inside a stable total, traces each to the change that plausibly caused it, and assigns an owner. The largest reason is rarely the one to work on.
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?
Returns with reasons and dates, order and product data, shipping cost, and ideally a record of product, copy and supplier changes. That last one is what makes cause-linking possible.
Why ignore the largest reason?
Changed mind at 41% has been 41% for years, which makes it the cost of doing business in this category rather than a defect. Effort spent there moves a structural number very slightly. Effort on a reason that doubled in six months has a findable cause and a fixable one.
How confident is the cause link?
It is correlation over timing and affected SKUs, which is enough to send someone to look and not enough to conclude. The skill names the candidate change and says what would confirm it.
What is included with the skill?
The skill itself, the reason taxonomy reference, the pattern format, and the owner routing. It is free and security scanned. We re-publish the ZIP when the taxonomy changes, so download it again if your patterns start shifting.
A growing refund reason is a margin leak with an owner.
Polar joins refund reasons to SKU margin across 45+ sources, so the pattern you route is the one costing you most.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from a returns report?","acceptedAnswer":{"@type":"Answer","text":"A report gives you rates by reason. This reads the trend in each reason, finds the ones moving inside a stable total, traces each to the change that plausibly caused it, and assigns an owner. The largest reason is rarely the one to work on."}},{"@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":"Returns with reasons and dates, order and product data, shipping cost, and ideally a record of product, copy and supplier changes. That last one is what makes cause-linking possible."}},{"@type":"Question","name":"Why ignore the largest reason?","acceptedAnswer":{"@type":"Answer","text":"Changed mind at 41% has been 41% for years, which makes it the cost of doing business in this category rather than a defect. Effort spent there moves a structural number very slightly. Effort on a reason that doubled in six months has a findable cause and a fixable one."}},{"@type":"Question","name":"How confident is the cause link?","acceptedAnswer":{"@type":"Answer","text":"It is correlation over timing and affected SKUs, which is enough to send someone to look and not enough to conclude. The skill names the candidate change and says what would confirm it."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the reason taxonomy reference, the pattern format, and the owner routing. It is free and security scanned. We re-publish the ZIP when the taxonomy changes, so download it again if your patterns start shifting."}}]}</script>