How to measure user onboarding: completion rate and drop-off per step
Views, completions, early closes and a funnel per step tell you where a tour loses people. Here is what each number means and what to do about it.

An onboarding tour that nobody finishes looks exactly like one that everybody finishes, until you measure it. The measuring does not have to be elaborate. Four numbers and one funnel answer most questions.
This post explains those numbers, how to read a funnel per step, how Stepsome gets the data to you, and the limits you should keep in mind before drawing conclusions.
The numbers worth watching
| Number | Meaning |
|---|---|
| Views | Times a tour or widget was shown. |
| Completed | Views that reached the end. |
| Closed early | Times it was dismissed before the end. |
| Monthly active users | People who opened a page with your site file on it, counted once per month. |
From the first two you get the completion rate: completed divided by views. It is a useful single figure for comparing a tour with its own previous version. It is much less useful for comparing two different tours, because a three-step tour and a nine-step tour are not the same ask.
Monthly active users gives you the denominator for reach: of the people who used the app this month, how many were shown the tour at all? A tour with a fine completion rate that only a small share of new users ever sees has a targeting problem, not a content problem.
Read the funnel per step
The completion rate tells you that people leave. The funnel tells you where. For each step it shows how many users reached it, so the difference between two bars is the number who left in between.
Where the drop sits says a lot about the cause:
- A large drop at step 1. The tour started at the wrong moment or for the wrong people. The user came to do something else. Look at when it starts and who sees it before you rewrite any text.
- A drop at one step in the middle. That step asks for something the user cannot or will not do yet: connect an account, enter data they do not have at hand, make a decision. Or it points at something that was not on their screen.
- A slow, even decline. No single step is broken. The tour is longer than the user’s patience.
- A drop just before the end. The user got what they needed and left. That may be fine. Check whether they went on to use the feature.
In Stepsome, open Analytics in the dashboard, choose the period at the top right (7, 30 or 90 days, or 12 months) and click a tour or widget in the table for its step-by-step funnel. Button clicks are listed underneath, so you can also see whether anyone used the “Show me how” button in a popup.
Two ways to get the data
Stepsome never receives your visitors’ data, so the numbers have to live somewhere you control. There are two routes, and you can use both.
Route 1: the analytics tool you already have
Without any setup, Stepsome sends its events to PostHog, Plausible, Google Analytics, Umami or Mixpanel when your app already loads one of them. The events are named tourcast_step, tourcast_complete and tourcast_dismiss, among others, and carry the tour name, the step number, the total number of steps and the step title. That is enough to build the same funnel in the tool where the rest of your product data already is.
You can also listen in your own code and send the events wherever you like:
window.addEventListener('stepsome:event', (e) => console.log(e.detail));
Route 2: a collector you host yourself
For the Analytics and Users pages in the Stepsome dashboard, you run a small service called the collector. It comes with Stepsome in two versions: a Cloudflare Worker with a D1 database, or a Node server (Node 22 or newer, no dependencies) that keeps everything in one SQLite file.
- Put the collector online on Cloudflare or your own server. On your own server, put it behind HTTPS.
- In the dashboard, go to Settings → Analytics, fill in the Collector address and the Read key and click Test connection.
- Download the site file again and deploy it.
The read key stays in your browser and never ends up in the site file; only the collector address does. Why you might want visitor data on your own server in the first place is the subject of Onboarding without third-party scripts.
Count people, not browsers
Without further setup, visitors are counted anonymously, once per browser. One person on a laptop and a phone is two visitors. To count people, tell Stepsome who the user is after login:
Stepsome.identify('user_42', {
email: 'ada@acme.com', plan: 'pro',
account: { id: 'acme', name: 'Acme Inc' },
});
- The first value is your own id for the user. Someone who uses two browsers is then counted once.
accountgroups users per customer, so you can see whether a whole account has been onboarded or only the person who signed up.- Everything else is a trait. Traits show on the Users page and can be used to choose an audience for a tour.
On the Users page you can search by user id, account or trait and click a person to see what they saw and completed, most recent first. That is the fastest way to answer a support question like “did this customer ever get the setup tour?”
What to change when a step leaks
- Wrong moment: change when it starts. Switch from First visit to On event, so the tour begins when the visitor first clicks into the feature it explains.
- Wrong people: narrow On pages and Audience.
- A step that asks too much: move that task out of the tour into a checklist item, where it can be done later and still be tracked.
- Too long: cut the tour to the steps needed for the first result and move the rest to tooltips or a highlight. Which pattern fits which message helps you decide where each piece goes.
- Unclear text: rewrite the step to say what the user gains, then compare the funnel of the following weeks with the one before.
Change one thing at a time and give it enough traffic before you judge. With few new users per week, a difference between two weeks can be chance.
Limits to keep in mind
- Some visitors are not counted, on purpose. Visitors with Do Not Track or Global Privacy Control switched on are never sent to the collector. Your numbers are a floor, not a census.
- “Seen” is per browser. A first-visit tour shows again for the same person on a new device, which adds a view.
- Reports are cached for a minute. Click Refresh when you are testing and want the newest numbers.
- Finishing a tour is not activation. The number that matters is whether new users reach the outcome your onboarding exists for. Put tour completion next to that product metric, not in its place.
If you have not defined that outcome yet, start with User onboarding for SaaS: a practical guide for small teams. For which plan includes the funnel per step, see pricing.


