Healthcare
B2B
The problem
NewU is a workplace wellbeing platform. Before a single employee can use it, an HR admin has to set it up. They create initiatives, launch services and invite their team.
That first session was where we lost people. A new admin landed on a dashboard that showed everything at once: modules, settings, analytics and content tools. There was no starting point, no sense of progress and no signal for what done looked like.
34% of new admins finished their first setup session
62% hovered over Skip before leaving entirely
Two thirds of new admins walked away before the product had a chance to help them.
What I found
I wanted to know why admins were leaving, so I started with our growth funnels and then spoke to admins directly.
The funnels showed that admins weren't lazy. They were overwhelmed. They didn't skip because they didn't care. They skipped because they didn't know where to start.
The conversations confirmed it. Admins didn't want more dashboard features. They wanted guidance, and they wanted to understand why each step mattered.
A cohort analysis by company size showed where the pain was sharpest. Enterprise admins had dedicated onboarding support. Small teams moved quickly because there was less to set up. Mid-size companies, with 50 to 200 employees, had the highest drop-off and cared most about how quickly they saw value. That middle group became my focus.
How might we make an admin's first tasks feel like the start of a helpful routine, one that builds confidence and brings them back?
What I believed
Before designing anything, I wrote down three beliefs about how admins would behave, so I could test each one.
1. Fewer choices will lead to more completions
The dashboard gave admins more than 12 possible actions on first load. I believed that limiting each step to two or three options would lower the mental load enough to raise completion. The product wouldn't get simpler. The first encounter with it would.
2. Order matters more than features
The three core tasks formed a natural chain. An initiative gives a service its purpose, and a live service gives employees something to join. If setup followed that chain, each step would build momentum for the next.
3. AI can remove the blank page
Even admins who knew what to do got stuck at empty fields. Initiative descriptions, surveys and wellbeing content all needed words. I believed a generated starting point, not a finished answer, would be enough to get them moving.
What I designed
I replaced the dashboard with a guided track. Admins now move through setup one clear step at a time, with a progress bar that shows how far they have come and what is left.
One decision per screen. Each step asks for one thing and offers two or three options at most.
Setup in its natural order. Create an initiative, launch a service, then invite employees. Every step gives the next one its context.
Examples instead of instructions. Steps like uploading a logo show what good looks like, so admins can follow rather than guess.
AI starters. Admins can let NewU generate initiatives and content, then edit them. They stay in control and never face a blank page.
Room to come back later. Optional steps can be skipped without losing progress.
Every screen was a decision, not a deliverable.
What worked and what didn't
Guided first beat dashboard first
Validated. With fewer choices at each step, more admins finished. The idea of "just a few steps to value" shaped every screen.
Ordering the tasks unlocked 3x more setup actions
Validated. When setup followed the natural chain, admins completed three times more setup steps in the same amount of time. Each finished step proved the next one was worth doing.
AI starters got things moving
Validated. The time it took to launch an initiative dropped sharply. Admins didn't need a perfect draft. They needed something to edit.
Tooltips built confidence but didn't change behaviour
Failed. My first idea was tooltips. We built a tooltip tour with Usertesting and found that admins felt better informed but acted exactly the same. Tooltips answered "what is this?" when admins were really asking "what should I do next?" That failure is what pushed us toward the guided track.
The results
Each result traces back to one of the beliefs I set out to test.
34% to 58% first-session completion, driven by the guided start and AI starters
40% faster time to first value, driven by fewer choices in a better order
25% more admins activated, driven by ordering tasks around what each one depends on
What I learned
Most product teams treat onboarding as a setup task, something to get through before the real product begins. The shift that mattered here was treating the first session as the product itself.
The failed tooltip test taught me the difference between informing people and guiding them. Admins didn't need more explanation. They needed a clear next step. Every first action should feel like value, not overhead.