An onboarding that halves in time.

78% of new users never finished.

Trainiac is a digital personal-fitness platform inside Gympass. Onboarding is where it tries to learn enough about a user to recommend a coach, a plan, and a starting workout.

The existing flow was a long, single-page intake form. Drop-off sat at nearly 78%. The product team had tried trimming questions and rewording copy without meaningful improvement. They asked me to take a wider look.

The form was shaped wrong.

I ran user interviews, surveys, and usability sessions across age groups, fitness levels, and geographies. I synthesised the data through an affinity diagram tied to our primary personas. Two pain points dominated:

• Length & openness. A long list of open-ended questions on a single page felt daunting and tedious, especially to users already unsure about their fitness identity.
• No sense of progress. A scroll with no visible structure read as “forever.” Users couldn’t tell where they were, how long was left, or what was optional.

Progressive Disclosure plus a map.

The redesign was anchored on a single UX principle, Progressive Disclosure, supported by a small set of patterns that took the “forever” out of the flow.

One question at a time, but always with the map.

I split the long form into a step-by-step flow, with one or two related questions per screen. The stepper at the top shows the overall arc; the time-to-completion estimate sets a concrete ceiling. The skip option, paired with a few clearly-optional questions, takes the pressure off without compromising what the system needs to recommend a starting point.

Contextual help icons next to ambiguous questions surface a one-line explanation in place, with no page change or modal. The user gets clarity without losing momentum.

Multi-variable testing, not guessing.

Once the components were in place, I proposed multi-variable testing on three dimensions (quantity of questions, content copy, and section order) rather than choosing by intuition.

• Quantity. We tested four lengths. Reducing the question count by 40% dropped abandonment dramatically without hurting recommendation quality downstream.
• Content. Several copy variations went head-to-head; the version that wove brief educational content between questions held attention measurably better than question-only flows.
• Order. We tested permutations of the three macro-sections. The winning order (Behavioral → Goals → Profile) lined up with how users wanted to introduce themselves: here’s what I do, here’s what I want, here’s who I am.

The combined effect of those three winning variables is what produced the headline numbers.

Adapted for Brazil.

Gympass has a large Brazilian user base. After the US rollout I worked with the local team to adapt the flow to cultural and linguistic context: warmer copy, shifted defaults, and a couple of restructured questions that didn’t translate one-to-one. The completion rate in Brazil came in at 75%, higher than the US baseline.

That part of the project taught me to treat localisation as a design problem, not a translation problem. Same UX principles, different cultural defaults.

What I came away with.

This project reaffirmed that good design isn’t one-size-fits-all. The same UX principles needed different defaults in different markets, and the willingness to test rather than assume is what produced the result. Multi-variable testing on real users beat any version of “the right way to do onboarding” we could have argued about internally.

The bigger lesson was about the relationship between completion rate and trust. Cutting questions, showing progress, and giving a skip option told the user the product respected their time. Conversion was the side effect.