CRO Specialist
by
Conversion review for ecommerce teams: find where the funnel leaks by source and device, size each leak in dollars, then rank the tests worth running.
Site-wide conversion rate is an average of segments that behave nothing like each other, which makes it almost useless as a diagnosis. Mobile paid social converts at a third of desktop organic, and a site-wide number hides that completely. Teams then run tests on whatever seems interesting, most of which reach no conclusion because the traffic was never sufficient. This skill finds where the funnel actually leaks by source and device, sizes each leak in dollars, and ranks the tests by what they are worth against what they cost to run.
What It Does
- Splits the funnel by traffic source and device rather than reporting one rate.
- Sizes each leak in revenue, not in percentage points.
- Checks whether a segment has the traffic to conclude a test at all.
- Separates a genuine conversion problem from a traffic quality problem.
- Estimates the runtime each test needs before it can be read.
- Ranks the tests worth running, with the value and the runtime.
Use Cases
- Find the segment dragging a site-wide conversion rate down.
- Stop running tests on segments too small to ever conclude.
- Tell a conversion problem apart from a traffic quality problem.
- Rank a test backlog by dollars rather than by opinion.
Example Session
Where are we losing conversions, and what should we test first?
Mobile paid social is the leak. It is 41% of sessions and 14% of revenue.
- Segment | Sessions | Conversion | Revenue lost vs peer | Test | Runtime
- Mobile, paid social | 41% | 0.7% | $184,000 per quarter | Landing page speed | 12 days
- Mobile, organic | 22% | 2.1% | Baseline | None needed | n/a
- Desktop, paid search | 14% | 3.4% | Baseline | Checkout fields | 26 days
- Desktop, email | 9% | 4.8% | Baseline | Too small to test | n/a
Mobile paid social converts at a third of mobile organic on the same devices, which points at the landing experience rather than at the traffic. The first test is page speed on those landing pages and it concludes in twelve days at this volume. Desktop email converts best and is too small to test: leave it alone rather than spending a month on an inconclusive result.
Why This Beats Prompting It Yourself
A Site-Wide Conversion Rate vs The Segment Behind It: The Hidden Risk
A generic prompt reads one conversion number and suggests the standard remedies: shorter checkout, more trust badges, faster pages. Those may all be right somewhere on the site. The average cannot tell you that mobile paid social converts at 0.7% while mobile organic on the same devices converts at 2.1%, which is the entire problem and points somewhere specific.
When You Still Need To Know The Test Will Conclude
A test on a segment without the traffic to reach significance is a month spent to learn nothing, and most backlogs are full of them. This skill states the runtime up front and marks desktop email as untestable at 9% of sessions, which is more useful than a test idea. Knowing what not to test is half of a working program.
Known Limitations
Runtime estimates assume the effect size you specify and current traffic levels, so a seasonal swing changes them. Separating a conversion problem from a traffic quality problem is inference rather than proof: it compares segments on the same devices and intents, which is strong evidence and not a controlled test. It ranks the tests and does not build or run 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/cro-specialist/
SKILL.md
references/segment-map.md
weekly/
# one file per week, with the ranked backlog
# 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 my analytics dashboard?
A dashboard shows the rates. This sizes each gap in revenue against a comparable peer segment, checks whether a test on it could ever conclude, and ranks the backlog on that basis. The output is an ordered list with runtimes rather than a set of charts.
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?
Session data with source, device and funnel steps, plus order data to size the revenue. It does not need a testing tool connected: it ranks what to test, and you run it wherever you already do.
How do you know it is the landing page and not the traffic?
By comparing segments on the same device with similar intent. Mobile paid social at 0.7% against mobile organic at 2.1% is evidence about the experience rather than the audience. It is inference, stated as such, and a holdout test is what would prove it.
Why mark a segment as not worth testing?
Because at 9% of sessions a test needs months to reach significance, and a test that cannot conclude costs a month and teaches nothing. Saying so up front is more valuable than adding another idea to a backlog.
What is included with the skill?
The skill itself, the segment map reference, the weekly backlog format, and the runtime calculator inputs. It is free and security scanned. We re-publish the ZIP when the segment map changes, so download it again if your rankings start shifting.
A conversion leak is worth whatever flows through it.
Polar ties sessions and segments to net revenue in one semantic layer, so the tests you rank are ordered by the amount at stake.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from my analytics dashboard?","acceptedAnswer":{"@type":"Answer","text":"A dashboard shows the rates. This sizes each gap in revenue against a comparable peer segment, checks whether a test on it could ever conclude, and ranks the backlog on that basis. The output is an ordered list with runtimes rather than a set of charts."}},{"@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":"Session data with source, device and funnel steps, plus order data to size the revenue. It does not need a testing tool connected: it ranks what to test, and you run it wherever you already do."}},{"@type":"Question","name":"How do you know it is the landing page and not the traffic?","acceptedAnswer":{"@type":"Answer","text":"By comparing segments on the same device with similar intent. Mobile paid social at 0.7% against mobile organic at 2.1% is evidence about the experience rather than the audience. It is inference, stated as such, and a holdout test is what would prove it."}},{"@type":"Question","name":"Why mark a segment as not worth testing?","acceptedAnswer":{"@type":"Answer","text":"Because at 9% of sessions a test needs months to reach significance, and a test that cannot conclude costs a month and teaches nothing. Saying so up front is more valuable than adding another idea to a backlog."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the segment map reference, the weekly backlog format, and the runtime calculator inputs. It is free and security scanned. We re-publish the ZIP when the segment map changes, so download it again if your rankings start shifting."}}]}</script>