+413% signup conversion, from one field removed.
A single required field was quietly costing a B2B procurement platform 98% of the people who reached it. Removing it took signup completion from 1.6% to 8.2%, and led the design system that shipped the new flow.
A construction-fintech platform raised funding to bring embedded finance to Saudi mega-projects. It had almost no one signing up.
The product is a B2B procurement and embedded-finance platform built for vendors and contractors on large-scale Saudi construction projects. As Product Lead, I owned the full registration experience, from the first landing page to the moment a vendor could submit a financing request.
Early on, the signup flow asked for a "unified number," a national business identifier, immediately after the landing page and before the user had seen any value. It felt like a reasonable verification step. The data said otherwise.
Diagnosed
Cross-referenced GA4 funnel data with Microsoft Clarity session recordings to find where users were actually leaving, not where we assumed.
Redesigned
Led the redesign of the signup flow, moving identity verification to after value delivery instead of before it.
Built the system
Built and led the design system the new flow shipped on, so every screen after signup inherited the same components.
One page was absorbing a third of all signup traffic, and sending most of it away.
Site visit to completed signup, the quarter before the unified-number field was removed.
Of the users who reached the unified-number step, almost all of them left without continuing.
A parallel performance issue compounding the same funnel. Half of all traffic was on mobile.
We hesitated before removing the field. It felt necessary for company verification, and rethinking the flow meant investing time in a change that might not move anything. What made the decision easy in the end wasn't intuition, it was watching the session recordings: people landed on the page, paused, and hit back. Nobody was leaving to go look up their number. They were leaving.
The interface, before and after.
Same registration goal, same audience. The unified-number screen removed, verification moved to after the user had an account.
The new landing page shipped alongside the signup redesign, two months after I joined as design lead.
Two independent sources confirmed the same thing.
A leak in the numbers is not the same as knowing why. It took both data sources together to make the call with confidence.
Google Analytics
The quantitative · what happened 1.6% end-to-end signup conversion, with the unified-number step in placeThe funnel showed a sharp drop at one specific step. Of everyone who reached the signup pages, only a small fraction ever reached the completion page. The unified-number step was the single largest leak in the entire journey.
Microsoft Clarity
The qualitative · why it happened Rage backs a recurring pattern across session replaysRecordings showed the same behaviour again and again: a user lands on the unified-number screen, pauses for a few seconds, then hits back or closes the tab. No one was leaving to go look up their number. The field read as a barrier, not a filter.
Same journey, one step removed, a different outcome.
Total traffic dropped between the two periods. Completed signups still rose. The signal: the field wasn't filtering, it was repelling.
A second funnel, a harder question: how does our most complex flow compare to the rest of B2B finance?
The pay-later request is the platform's core financing product: a five-step flow covering credit qualification, document upload, and signature. I ran the same two-source analysis on it.
| Flow type | Typical completion |
|---|---|
| Simple contact form | 60 – 80% |
| Auto insurance application | 35 – 50% |
| This platform's pay-later request | 52% ← we're here |
| Business bank account opening | 20 – 35% |
| B2B / SME loan request | 15 – 30% |
| Mortgage application | 8 – 15% |
At 52% end-to-end completion, the flow sits roughly twice the typical range for a comparable B2B loan request.
The wizard itself converts well. Only 14% of people who view the pay-later page ever start the request, that's the narrow entrance, not the flow. The biggest lever isn't optimising the five steps further, it's getting more qualified users to start in the first place.
| Transition | Drop-off | B2B benchmark | Verdict |
|---|---|---|---|
| Start → Basic info | 22.6% | 5 – 15% | Needs review |
| Basic info → Documents | 16.7% | 15 – 25% | Within range |
| Documents → Sign | 20.0% | 10 – 20% | Within range |
| Sign → Submit | 20.0% | 10 – 15% | Needs review |
Sample: 31 users, 1 Apr – 7 May 2026. A small sample, the direction is consistent, the exact rate needs a longer window before any major decision.
This was a product design investigation from end to end, engineering executed the fix.
Identified the friction
Cross-referenced Microsoft Clarity recordings, GA4 funnel data, and PageSpeed Insights to surface the unified-number field and mobile load time as the two compounding bottlenecks.
Built the design system
Led and built the design system the new signup flow and landing page shipped on, so every screen after launch inherited the same components and tokens.
Owned the conversation
Authored a formal performance and funnel audit, presented it to engineering, and led the prioritisation discussion across three independent data sources.
Shipped and validated
Within two months of joining, led the launch of the new landing page and signup flow, then re-ran the same analysis 30 days later to confirm the lift held.
Three lessons this work taught me.
Data beats assumption
We believed the field was necessary for verification. The numbers showed it was a barrier first, a filter second. The right question isn't "does this protect us," it's "does this repel users."
Two sources beat one
Analytics says a leak exists. Session recordings say why. Big calls don't get made on either alone, they get made on what both confirm together.
Removal is a design decision
Not every improvement is an addition. The biggest gain here came from removing something everyone assumed was required.
Building a product that needs the same kind of look?
I work with GCC fintechs and regulated platforms that need someone to own the full picture, from the data to the system that ships.
Start a conversation →