New to skills? Start with the AI Skills Library for how to add one.
Puts two pieces of published content side by side over a date range you choose, and builds an interactive dashboard comparing how they performed.
What it does
The skill fetches your published content and gives you a picker: choose a content type, then pick two items of that type and a date range. Because both dropdowns are scoped to the type you picked, you can't accidentally compare a checklist against a banner.
What you get back depends on what you're comparing. For guide flows and checklists, that's views, unique viewers, completion rate, missing element errors, a step-by-step funnel for each one showing where people dropped, and a views trend chart. For announcements, it's views, reactions, comments, and engagement per 100 views—no funnel or completion rate, since those don't apply.
After the dashboard, you get a few written observations about what the numbers mean.
What you'll need
The Userflow MCP connected
Two published items of the same type. Drafts have no analytics to compare.
How to use it
Ask for a comparison and pick from the widget it shows you:
"Compare my two onboarding flows."
"Which of these announcements performed better?"
"Benchmark the Trial Signup flow against the Welcome Tour over the last quarter."
You can also name both directly, or paste their Userflow links, and skip the picker.
The date range defaults to the last month. Shorter and longer ranges are available, from a single day up to a year.
What it covers
Guide flows, checklists, announcements, launchers, banners, and resource centers. Event trackers and the Adoption Agent are left out—neither has the kind of session funnel this comparison is built around. The Adoption Agent has its own analytics, and userflow-adoption-agent-topics covers what people asked it.
Things worth knowing
A date range that starts after something launched will hide its launch spike. The skill checks each item's first publication date against your window and warns you when the comparison could mislead because of it.
Zero views isn't necessarily a problem. Some content is designed to start only after something else finishes. The skill looks at the first step and the start conditions before calling anything dormant.
Missing element errors are counted per flow, not per step. If the count is zero, drop-off is about behavior rather than something being broken—which changes what's worth fixing.