ONB/PFG-02
ONB/PFG-02
Elevate by Principal
Elevate by Principal
Fintech
Fintech
B2B Onboarding & Activation
Users were vanishing between sign-up and contract activation. The ones who made it through landed on a dashboard offering nine things and pointing at none of them. Took me a while to see those were the same problem. The product was asking for more than it had earned and handing back less than it promised.
Users were vanishing between sign-up and contract activation. The ones who made it through landed on a dashboard offering nine things and pointing at none of them. Took me a while to see those were the same problem. The product was asking for more than it had earned and handing back less than it promised.
Earn the right to ask.
Earn the right to ask.
Funnel research
Funnel research
Adaptive forms
Adaptive forms
Design system
Design system
ONB/PFG-02
ONB/PFG-02
Role
Role
Senior Product Designer
Senior Product Designer
Deployed
Deployed
2023 to 2025
2023 to 2025
Domains
Domains
Registration / Activation / Savings
Registration / Activation / Savings
Team
Team
PM, ENG, Data, CS, Marketing
PM, ENG, Data, CS, Marketing
Constraint
Constraint
Fields the sales team depended on
Fields the sales team depended on
Tools
Figma, Claude Code, Cursor
Measured
FullStory, Google Analytics, Salesforce Marketing Cloud
50%
50%
50%
50%
50%
More completed registrations, year over year
More completed registrations, year over year
2% → 10%
2% → 10%
2% → 10%
2% → 10%
2% → 10%
Email engagement, measured in Salesforce Marketing Cloud
Email engagement, measured in Salesforce Marketing Cloud
01
01
The artifact
Three classification systems. Same practice underneath all of them.
Every item gets a rule, the rule gets written down, and the rule is the deliverable. Not the screen it produces.
Same skeleton, three step counts. Confidence decides what gets inserted, not which flow you’re in.
The automated business lookup returns a confidence level based on how many attributes it can validate: location, email, phone, industry, employee count.
That level doesn’t pick between three flows. It decides how many verification steps get inserted into one.
High confidence, five steps. You confirm what we already found and you’re done. Lookup finds nothing, seven steps, with disambiguation and manual entry filled in. Every tier lands on the same confirmation screen and the same finish.
Building it as insertion instead of branching is the only reason it outlived the feature it was built for. Three parallel flows would have been three things to maintain and nobody maintains three things.
Every field in the old flow, one row each, sorted into required-to-qualify-a-lead or nice-to-have, and for the second column, where it went instead.
| Field | Classification | Where it went |
|---|---|---|
| Revenue generation | Nice to have | Collected later, once the user has a reason to answer |
| Multiple phone numbers | Nice to have | Collected later, once the user has a reason to answer |
| Title | Nice to have | Collected later, once the user has a reason to answer |
| Ownership structure | Nice to have | Collected later, once the user has a reason to answer |
02
02
The work
Two surfaces. The intake, and the page they land on.
Four steps became three. Password creation left onboarding entirely. The email verification gate came off the front. Business name gained a lookup, which is what feeds the confidence tiers.
Before: four steps
After: three steps
The old flow ended with homework. The new one ended with the thing they came for.
The ending changed more than the middle did.
Old flow finished on “Almost there! Please verify your email.” You’ve just filled out four screens and the reward is in your inbox somewhere, maybe. New one finishes on “Instant savings unlocked,” with something to actually click.
Same funnel. Completely different feeling walking away from it.
The first version offered business benchmarks, focus content, a business profile, talk-to-an-owner, a concierge, a library, and strategic partners. Everything was available. Nothing was directed. The second added a value proposition at the top and kept nine cards underneath it.
The third gave one block the top of the page, with one job and one call to action, and moved everything else below the fold.
01
Version 1
02
Version 2
03
Version 3
A dashboard that offers everything directs you to nothing.
To show somebody their savings you need their spend data. Spend data is also the last thing they want to hand a product they haven’t seen work yet. Chicken, egg.
So: four ways in, sized to how much they’re willing to give up. Connect QuickBooks. Upload a spreadsheet. Type in a few vendors by hand. Or see instant savings with no data at all.
123- 1Connect QuickBooks: costs trust, for the user ready to integrate
- 2Step-by-step entry: costs effort, for the user who wants to see value first
- 3Upload a spreadsheet: costs medium, for the user with data but not access
The manual path isn’t a fallback for people who couldn’t manage the real one. It’s the cheap look around that earns you the integration later.
03
03
How the work actually moved
Requirements came from concierge and sales. They had the relationships, they knew what the business actually needed answered, which also meant they owned every field I was about to propose cutting.
So I talked to SMB owners first. Not as a formality. Without that, removing fields is a designer’s opinion against a sales team’s requirements, and I know how that one ends.
Business owners, customer managers, and marketing: three groups with three different definitions of a qualified lead.
04
04
What I didn't get to do
I wanted those questions gone.
What I got was a classification. Required to qualify a lead, or nice to have. And the nice-to-haves didn’t disappear. Revenue, ownership structure, title, they all still get collected. Just later, once somebody has a reason to answer.
Smaller win than deleting them. It was also the one on the table. Ask a sales team to give up data they rely on and you’re not having a conversation about the user anymore, you’re having one about whose requirements count. Moving a field is a scheduling question. Deleting it is a territory question. I’d rather win the scheduling one.
The dashboard cost me more. Two versions shipped before the third one worked, and I designed both of the ones that didn’t. Both went live. Both failed to point anybody at savings.
What finally fixed it was taking things away. That’s not where I’d been looking.
It took two shipped versions to learn the page needed one job.
05
05
Outcome
It was the first time a business owner said, ‘I didn’t know I could save this much. Where do I sign?’
50% more completed registrations, year over year.
Email engagement 2% → 10%, measured in Salesforce Marketing Cloud.
06
What I wanted to show you
Most onboarding work is about removing steps. This was about asking fewer questions.
Which is harder, because someone has to decide what you’re confident enough not to ask, and then go defend it to the people who wanted the answers.
ケビン・チャップマン
No. 26 / Lot 01 / SS26