What sets converters apart
Last updated
Some visitors convert and some don't. Converters shows what the two groups did differently before they decided: the pages they saw, where they came from, what they clicked, what got in their way. Each difference comes with the conversions at stake, so you know which one to look at first.
Open Converters
In your app, select Converters. By default, converting means reaching any goal. Choose one goal to narrow it, or a saved task to use the task's success instead. The date range and filters at the top work here too.
What it compares
Metrickle looks at every trait a visitor had and compares visitors who had it with everyone who didn't:
| Trait | Example |
|---|---|
| Pages | Saw /pricing |
| Entry pages | Landed on /blog/* |
| Sources and campaigns | Came from google.com, or from the spring-sale campaign |
| Country, device, operating system and browser | Visited on a phone |
| Accessibility settings | Used keyboard navigation |
| Custom events | Used compare_plans |
| Links out | Clicked a link to another site, grouped by its domain |
| Friction | Hit dead clicks |
| Slow loads | Had a poor Largest Contentful Paint |
| Survey answers | Answered 0–6 to "How likely are you to recommend us?" |
Only what happened before a visitor's first conversion counts. Seeing the thank-you page after buying doesn't make anyone more likely to buy. Visitors who never converted count everything in the date range.
Slow loads use the worst rating each visitor had for each Core Web Vital. See what slowness costs.
Survey answers are grouped into bands:
| Question | Bands |
|---|---|
| NPS (0–10) | 0–6, 7–8, 9–10 |
| CSAT and star rating (1–5) | 1–2, 3, 4–5 |
| Effort / CES (1–7) | 1–3, 4, 5–7 |
| Multiple choice | Each choice |
Read the results
Each row is one sentence, such as "Visitors who saw /pricing converted 2× as often (40% vs 18% for everyone else)." Next to it are the visitors with and without the trait, their conversion rates and the difference in points.
Traits that go with converting less come first, ordered by what they cost. Then come traits that go with converting more.
At stake
For a trait that goes with converting less, at stake is how many conversions those visitors fell short of everyone else's rate. If your goals carry revenue, it's shown as money too. That's the place to start.
Clear or too close to call
A difference is marked clear when all three hold:
- it's big enough that chance alone rarely produces it (a z-score of at least 1.96),
- at least 10 visitors had the trait, and at least 10 didn't,
- it survives a check for false alarms.
That last check matters. Metrickle tests hundreds of traits at once, and testing that many would flag a few by pure chance. The check holds every trait to a stricter bar, so the clear ones are real patterns.
Traits that aren't clear are listed last. They may become clear with more visitors or a longer date range.
Traits almost no one had, or almost everyone had, are left out. They can't tell anyone apart.
A step on the way
If almost every converter had a trait, such as seeing the checkout page, it's marked a step on the way. It's part of how people convert, not a difference between them.
What it doesn't tell you
This is what went together, not what caused it. People who read the pricing page may already have meant to buy. Before you move a link or redesign a page because of a trait, test the change and check the number moves.
To find out why, ask. Each row has Survey them and Plan a study links that open a draft prefilled from the finding. See ask them why.
With a saved task
When you choose a saved task, success is net: a completion later taken back by a refund, a cancel or a problem report doesn't count. See protected flows.
Ask an AI assistant
Connected AI assistants can read the same results with get_conversion_drivers, and turn a finding into a survey or study draft with suggest_research. See the MCP server.