Publishing a tour is the easy part. The Results tab inside a tour tells you whether it actually helped anyone, and where it is losing people.
Pick the right version and date range
Results are kept per version of the tour. Every time you change the steps and save, you start a new version, because numbers from an old set of steps cannot fairly be compared with a new one.
Use the Version picker to switch between them, and the Date range next to it to narrow the window. If a version is younger than the range you picked, the panel says so rather than pretending you are looking at a full month.
The four numbers at the top
Started — how many identified users began the tour.
Completion rate — what share of them reached the end.
Median time to complete — how long it takes in practice. A tour that takes far longer than you expected is usually asking people to do something harder than you thought.
Goal — how many went on to do the thing the tour was about. If you have not set a goal, this reads No goal set, and you are only measuring whether people clicked to the end.
Completion and goal are different questions, and the gap between them is the interesting part. A tour everyone finishes but nobody acts on has a content problem, not a delivery problem.
The five outcomes
Under the bar, every run is sorted into one of five buckets: All, Completed, Skipped, Snoozed and Stuck.
Treat them differently. Skipped means people chose to leave — usually a relevance or timing problem. Stuck means the tour stopped working for them, which is yours to fix.
Where people drop off
This is the part worth looking at first. Each step shows how many people reached it and how many were never seen on the next one.
One step with a large drop tells you more than the overall completion rate does. Read the step's copy and ask whether you are explaining something, or asking someone to do something they are not ready to do yet.
Issues: steps that broke
A step can fail on a visitor's page rather than being abandoned, and those are listed separately under Issues.
Each entry names the step, how many events and users it affected, what went wrong, and when it was first and last seen. The usual causes are:
The element was not found. The page changed after the tour was built and the selector no longer matches anything.
The element was ambiguous. The selector matched more than one thing on the page, so we skip the step rather than guess which one you meant.
Both are fixed the same way: open the step and pin it to something stable, such as [data-testid="inbox"], rather than a selector that describes where the element currently sits in the page.
Check this panel after any front-end release. A step that silently stopped matching will not show up as a complaint from customers — it just quietly drags your completion rate down.
People
Below the issues, the People table lists the visitors behind the numbers, with the furthest step each reached, how many runs they had, and when they were last active. Click an outcome above to filter the list to it.
Only identified users appear here. If you are not passing user details to the widget, your tours still run, but this table will be empty and Started will undercount.
What to do with what you find
High completion, low goal — the tour is being read but not acted on. Rewrite the step that asks for the action, or set the goal closer to what the tour actually teaches.
A cliff at one step — look at that step alone. It is usually too long, or it asks for a decision the visitor cannot make yet.
Anything under Issues — fix it before reading anything else on this page. Broken steps make every other number meaningless.