We Tried to Shorten Our Onboarding. Twice. The Long Version Won Both Times
Open any growth blog and you'll find the same advice: shorten your onboarding. Cut steps. Reduce friction. Every field you remove is conversion you gain.
I believed it. I've even written it. So on a SaaS product I work on, we tested it properly. Twice, months apart, on different devices, with different step counts.
The long onboarding won both times.
This post is about why, and about the distinction we ended up drawing between friction that costs you users and friction that builds the sale. The numbers are roughed out and the product details are deliberately vague, but the shape of the result is real.
The setup
Our onboarding is long by any standard: roughly nine steps between signup and the paywall. Some steps do obvious work (tell us about your business, pick your goals). Some are what I would have called filler: an interstitial that "analyzes" what you just entered, a couple of affirmation-style screens that reflect your answers back at you, a progress moment that shows you a score before you've paid for anything.
The hypothesis wrote itself. Users are dropping off through these steps. The filler steps carry no information we need. Remove them, get people to value faster, watch conversion climb.
Test one: cut the fluff
The first experiment removed the softest steps: the affirmations, a couple of questions we never acted on, the theatrical pauses. Same funnel, several steps shorter. Cleaner, faster, more respectful of the user's time. I was confident enough that I mostly worried about how much we'd win by.
Control won. Not a tie: the shorter flow produced measurably fewer people reaching the paywall and starting a trial. The steps I'd dismissed as filler were carrying weight I couldn't see.
We shipped the long version back and I filed the result under "weird, revisit later."
Test two: mobile, where friction hurts most
Months later we came back with a sharper version of the same idea. Mobile users drop off through onboarding at a higher rate than desktop, and the standard explanation is that long flows punish small screens. So we built a mobile-only variant: five core steps instead of nine, with the removed steps deferred until after the trial started. Nothing was deleted, just moved out of the critical path.
If the shorten-your-funnel advice holds anywhere, it should hold on a phone.
Control won again. The nine-step flow beat the five-step flow on the metric that matters: registration through checkout to trial start. The short flow got more people to the paywall faster, and converted fewer of them once they arrived.
Two experiments, two devices, months apart, same answer.
What I think is actually happening
I don't get to see inside users' heads, so this part is interpretation. But the pattern fits a few well-worn pieces of psychology that the cut-your-funnel advice quietly ignores.
Onboarding is a sales conversation, not a form. We'd been treating step count as pure cost: every screen a chance to leave. But every screen is also a chance to build the case for paying. The "filler" steps were doing the job a good salesperson does: restating the user's problem in their own words, showing that the product understood their situation, building anticipation for the result. Cutting them shortened the pitch.
Effort creates investment. People value what they've worked for. A user who has answered eight questions about their business has sunk something into the process, and walking away at the paywall now means writing that off. The IKEA effect is usually invoked for furniture, but it applies to funnels: a little assembly makes the outcome feel more yours.
Reflection steps convert information into perceived value. The steps that reflected the user's answers back ("here's what we found about your situation") did no data collection at all. That's exactly why I wanted to cut them, and exactly why they mattered. They were the moment the user saw the product thinking about them specifically. Speed-running someone to a paywall skips the part where the product proves it was listening.
The distinction that survived: dead friction vs. productive friction
The lesson is not "long onboarding good." We have also fixed genuinely broken steps (layout overflow on small screens, tap targets too small for thumbs) and those fixes helped. That's the distinction worth keeping:
Dead friction is effort with no payoff for the user: broken layouts, confusing copy, questions that feel pointless, waiting with no explanation. Kill it on sight.
Productive friction is effort that builds commitment or perceived value: questions the product visibly uses, moments where the user sees themselves in the output, small investments that make the outcome feel earned.
The growth-listicle advice collapses both into "steps," counts them, and tells you fewer is better. Counting steps measures the cost side of every screen and ignores the value side entirely.
How to test this on your own funnel
If you want to run the same experiment, a few things we'd do again and a few we learned the hard way:
Measure to the end of the funnel, not to onboarding completion. A shorter flow will almost always show a higher completion rate. That's the trap. Our short variants completed better and converted worse. Judge on trial starts or payments, not on finishing the flow.
Defer, don't delete. Moving steps after the paywall (rather than removing them) lets you separate "this step blocks conversion" from "this data is useless." Ours blocked nothing; it was load-bearing.
Re-test before you believe yourself. One surprising result is a fluke candidate. We only trusted this after it replicated on a different device class months later.
Interrogate each step, not the count. The right question is never "how many steps do we have?" It's "what is this step buying, and who is paying for it?"
The uncomfortable takeaway
The advice to cut onboarding steps isn't wrong because friction doesn't matter. It's wrong because it's a proxy metric dressed up as a principle. Step count is easy to see and easy to reduce, so it became the thing everyone optimizes. What users actually respond to (feeling understood, investing effort, anticipating a result) is harder to see and doesn't fit in a listicle.
We tried to respect our users' time, twice, and both times they told us, through their behavior, that they'd rather be sold to properly. I still wince slightly when I watch someone click through step seven of nine. Then I look at the conversion numbers and leave it alone.