Zilch acquisition funnel optimisation
Zilch is a UK-based buy now, pay later (BNPL) platform serving over 5 million customers, letting shoppers spread the cost of purchases across interest-free instalments.
I was the lead designer on the Growth and Acquisition team, working alongside a PM, engineers, and a data analyst. I also closely collaborated with other teams accross other verticals that touched onboarding (Decisioning, Marketing, KYC team).
The business objective was to grow the number of users reaching M1 status (first spend) by fixing the leaks in an onboarding funnel that was turning away motivated customers and revamp the UI of the app to make the product more marketable.
Company type
Fintech, Scale-up
Time frame
18 months
Contribution
UX Research, A/B Testing, Product Design & Strategy, Analytics
THE PROBLEM
The funnel treated people who'd applied before, changed their details, or came back to try again as new customers. We handled these scenarios very poorly, so the experience broke for them, predictably, every time.
A PIVOT IN PRODUCT STRATEGY
Before I could dig deeper into the core problems, our strategy was focused on stuffing more users into a broken funnel and being very reactive to problems that appeared along the way.
The PM at the time was focused on reaching web and mobile feature and visual parity, but I soon realised that by moving users straight into the app would remove the need to work on web entirely, saving us a lot of dev time and money.
We soon had a change in the team PM which resulted in a better collaboration and prioritizing the key problems I was eagerly awaiting to tackle.
RESEARCH
While the new PM settled in, I was already looking into the end to end experience, and trying to collate as much information and data as I could while working on how the upcoming brand refresh would be applied to the funnel. We then started looking into the top of the funnel (account creation) first due to the credential stuffing and increased fraud attempts we were experiencing.
MAIN FINDINGS
With the help of the data team, myself and the Product Manager started doing deep dives into the quant data as well as going deeper into the qual side, where I spoke with users, tried competitor apps and gathered insights from other teams such as Customer Service and Marketing.
Login loops and abandoned applications block future attempts
Users endure an average of 4 logins before becoming M1 customers, with 76% of these authentication loops occurring during onboarding. There was also no mechanism to "reset" their journey after a longer time away.
35% of users fail at phone verification step
This was happening because phone numbers were permanently linked to unverified emails with no recovery path available.
40% of users don't complete IDV
This was the third main drop off area, where we were letting users explore the app before they were verified with no education added prior to them seeing the app for the first time.
STRATEGIC DIRECTION
Most of the roadmap was owned by product but the design team and myself were continuously spotting gaps in the product and thinking about the future vision, so we pushed to close them, with the help of the leadership team and other departments. Some of the opportunities I uncovered are shown below.
VISION · FUTURE ONBOARDING
Pitched the future of onboarding to CPO and VP of Growth
Moving the credit agreement and the push opt in page early in the funnel were adopted into the roadmap.
OPERATIONS · COMMS AUDIT
Mapped every message Zilch was sending, because nobody else had
Gaps closed, obsolete messages removed, future endeavours simplified. Marketing used it as a reference throughout.
IDV PROVIDER · ONFIDO
Took an IDV failure pattern straight to our vendor
24% of users went to manual review due to low lighting. Pitched a solution to Onfido for a potential improvement.
EXPERIMENTATION
While larger infrastructure changes were being scoped, I identified opportunities for immediate impact through a few key A/B tests. These would test some of my assumptions and deliver business value while I did the discovery on the rest of the funnel.
AREA 1 OF 3 · ACCOUNT CREATION
30% of phone numbers were already linked to another account, so OTPs silently failed. Email first let us detect existing accounts immediately and route people correctly. It also closed a credential stuffing vulnerability, since we could no longer confirm whether an email had an account.
OTP always sends
If a phone number was linked to another account, users were told clearly and given a path to their existing account.
Email collected upfront
Existing accounts detected at step one rather than being checked in the middle of the flow. Returning users routed to login and verified before hitting any dead ends.
Device-aware experience
Trusted devices get a streamlined login. New devices get full verification once, then trusted from that point forward.
AREA 2 OF 3 · PLATFORM STRATEGY
The initial plan discussed with my PM at the time was achieving full web/app parity. A competitor analysis changed my view on this: almost no BNPL company runs a serious web sign-up flow anymore. I proposed using web as a handoff point instead, moving users to app via a QR code.
Web as entry point, not product
Users get a compelling prompt to continue on app via QR code rather than a degraded web experience.
Cleaner security posture
Removing web sign-ups eliminated the credential stuffing attack surface that had been a persistent vulnerability.
Parity debt gone
One codebase, one journey. Every future improvement applied everywhere rather than being built twice.
AREA 3 OF 3 · FIXING TASK COMPLETION RATE
40% of users who signed the credit agreement never completed IDV. While their ID was checked, they'd land in a shopfront they couldn't use, with no idea why. I redesigned the pending states, moved notifications earlier so drop-offs could be retargeted, and shifted the RCA to after IDV.
"Your credit limit is waiting"
Users who returned after dropping off IDV were met with a targeted screen. Those past the expiry window had their application reset cleanly.
Clear pending state
Users knew exactly where they were and what to expect, instead of landing in a shopfront they couldn't use.
Credit Agreement moved post-IDV
M1 status only set after full completion. No more phantom M1s from users who signed the agreement but never finished IDV.
OUTCOMES
Every conversion improvement here is a motivated user recovered, not a new one acquired.
WHAT I LEARNED
