Stop guessing which onboarding works.
Run two versions of a flow at the same time - shorter or longer, a different opening line, one less step. We split the traffic, watch the completion rate, and tell you when the difference is real rather than noise. Keep the version that won.
Set it up in a switch, and let it call itself.
One switch, and version B is ready.
Turn the test on and we start version B as a copy of the flow you already have. Change the one thing you want to learn about, and leave everything else alone. No second flow to build, no traffic rules to wire up, nothing for your engineers to deploy.
- Version B starts as a copy, not a blank page
- Edit it with the same editor you already use
- Nothing reaches your users until you publish
We tell you when to stop looking.
The trap in testing is calling a winner too early: check often enough and random noise will eventually look like a result. We set the sample up front, hold the verdict until there is enough evidence, and say plainly when two versions are simply the same. A clear difference is still called sooner.
- Sample fixed before the test starts
- "No real difference" is a valid answer we will give you
- Promote the winner into your draft when you are ready
We tell you when the difference is real.
Splitting traffic is the easy part. Knowing whether a difference means anything is the part that decides if you learned something, and it is the part we do for you.
Every flow type
Tours, tooltips, checklists, announcements, and your resource center. Test the wording, the number of steps, the order, or which help links you offer.
Completion, per version
Each version reports its own starts and completion rate, so the comparison is like for like rather than a before-and-after guess.
Called when it is real
A difference only counts once there is enough evidence for it. We hold the result until then, so a lucky first afternoon never reads as a win.
The same audience
Both versions face identical targeting rules, so the only thing that differs is what you changed. That is what makes the answer usable.
Signed-out visitors too
Flows that run before someone signs up can still be tested, so your very first impression is not the one thing you cannot measure.
AI writes version B
Out of ideas is the usual reason a test never runs. Ask for suggestions grounded in your own drop-off, pick one, and get a draft to review.
Included from Growth, at a flat monthly price.
A/B testing is the sort of thing that usually sits behind a sales call and a five-figure contract. Here it is a line on a self-serve plan you can start today and cancel yourself. Your bill is the same whether you run one test or twenty.
Explore the rest of the toolkit.
Start turning signups into active users.
Free forever to start, your flows carry our badge until you upgrade. Live in one line, no sales call.
Every plan includes OnbiAI - draft flows, targeting and audiences with AI