Maya

← back to work

KYC

Rebuilding eToro's identity verification flow. Two parallel tracks: A/B testing the live production flow to smooth out the friction points, while designing the next-generation flow on the new design system.

Product Designer · eToro · 2025 — ongoing

Client
eToro
Industry
Fintech · Online trading
Role
User research, Moderated interviews, Prototypes, User testing, Onboarding flow, High fidelity design, Dev handoff, Design QA
Team
Product designer (me), Product manager, Engineering team, Compliance partner
Duration
September 2025 — ongoing
Methods
  • User interviews
  • Mixpanel
  • Fullstory
  • A/B testing

Problem

Describe the context. What was the problem space? Who were the users? What constraints existed?

My role & team

I was the lead designer working with a team of [N] engineers, [N] PMs, and a researcher. I owned [scope].

Process

Outline the steps you took: research, sketches, wireframes, testing, iteration. What were the key decisions?

User research

Both tracks — the live A/B tests and the new design system flow — were anchored by the same research foundation. We talked to users who finished KYC, users who got stuck, and users who quit halfway. We watched Fullstory sessions to see exactly where people hesitated, and we mined support tickets for the patterns that never make it into interviews.

  • [40+] User interviews
  • [3] User segments
  • [6] Markets
  • [200+] Fullstory sessions reviewed
User interviews tracker — names and emails blurred for privacy. Columns include date, goals, experience level, education, intent, and conversion outcome.
A slice of the user interviews tracker. Each row is one conversation, tagged by experience, intent, education, and whether the user eventually converted. Names, emails, and avatars blurred for privacy.

Three patterns kept coming back across segments — and shaped every decision that followed.

  • Is there any monthly subscription fee or trading fee included with my eToro account?
    New user · onboarding interview
  • I signed up to invest in XXX but I don't think it is available on eToro.
    New user · onboarding interview · asset name anonymised
  • When I got on board with eToro, what I found is that I could only use my credit card.
    New user · onboarding interview

Part 01 · In parallel

A/B testing the live production flow

The existing KYC was already live in production. User interviews and data from Mixpanel and Fullstory kept surfacing specific places we wanted to improve — so we ran A/B tests directly on the live flow, addressing those screens one at a time in parallel with the broader redesign.

A/B tests

Three tests had the biggest impact on completion. For each one we set the funnel goal in Mixpanel, paired it with Fullstory sessions to see exactly what users were doing on screen, and shipped the winning variant.

  • Test 01

    AI-suggested alternative assets

    🥈 2nd place · eToro internal hackathon

    Before

    Original empty-state search screen showing 'No results found' with no further action.

    After

    Redesigned search empty state — 'We don't offer POET yet, but…' with four alternative assets suggested, each with a Trade button.

    "I signed up to invest in XXX but I don't think it is available on eToro."

    Real user feedback · asset name anonymised

    Problem

    This one didn't start as a KYC-flow test — it surfaced in user interviews. Users were closing their accounts because they couldn't find the assets they were looking for. Once we dug into the data, the scale was bigger than we expected:

    • 300.9K Monthly users hitting "no results"
    • 12.4% Close their account due to a missing asset

    What we tried

    An AI suggestion layer that proposes alternative assets — closely correlated or category-similar — whenever a user hits a dead end. The POC used GPT to generate real-time suggestions in two scenarios: empty search results, and KYC-blocked asset types. Competitive analysis confirmed Kraken, Binance, and Trading 212 weren't solving this well, which made the opportunity clearer.

    Result

    This test shipped and ran in production. Measured impact across users who'd previously hit a dead end:

    • −5% Fewer account closures from users who didn't find what they wanted
    • +12% Took the suggested alternative asset
    • +6% Additional deposits from users who took the alternative

    How we prototyped

    Before building anything, we validated the alternative-assets logic by prompting GPT directly: given an asset the user searched for but couldn't trade, return a short list of closely-correlated alternatives — same sector, same theme, same risk profile — that we could offer them. The POC ran as a shared ChatGPT conversation; we tested it with real search terms from the data and rated the suggestions for quality and relevance. Once the patterns held up, we wired the prompt into the production search and KYC-block flows.

  • Test 02

    Real-time document verification

    Problem

    Verification was the slowest step in onboarding — 12 minutes on average, and up to 47 hours when manual review was triggered. 38% of POA documents were getting rejected (49% in France), and the rejection notice was sent by email — by which point most users had already closed the app. On top of that, 25% of users completed KYC stuck in a pending status, never sure whether they were verified, often leaving the funnel before they could deposit.

    What we tried

    Real-time verification, moved to the start of the flow. Documents are checked instantly as they're uploaded; rejections show in-app with the specific reason rather than landing in an inbox users had already left. Paired with personalised getting-started components so the next step is always clear.

    Result

    Forecast impact, based on the baseline data:

    • 38% POA rejection rate today — target cut to near-zero with real-time verify
    • 49% France market rejection rate — worst-case baseline
    • 25% Users stuck in pending KYC status after submission
    • 96% Lower cost per verification ($0.75 → $0.03)

    Baselines from internal data, June 2025. Will update with measured cuts once shipped.

    What users were telling us

    • I already submitted the requested documents multiple times, and it never got accepted.
      Real user · support email
    • I can't deposit funds into my account, because I've not been verified as yet, this is so frustrating.
      Real user · support email · Jan 2024
    • My experiences are non-existing because I can't deposit until I am verified and that, apparently, takes forever (still pending).
      Real user · support email · Feb 2024
    • My onboarding is horrible. I'm waiting to get my account approved for a week now!
      Real user · support email
    • It's only showing pending verification. Now it's almost 1 week but still not verified — please help.
      Real user · support email · Feb 2024
    • I'd like to invest but it's been 4 days and my verification is still pending so I can't deposit.
      Real user · support email · Jan 2024

    1. Step 01

      Real-time document verification — step 1

    2. Step 02

      Real-time document verification — step 2

    3. Step 03

      Real-time document verification — step 3

    4. Step 04

      Real-time document verification — step 4

    5. Step 05

      Real-time document verification — step 5

    6. Step 06

      Real-time document verification — step 6

  • Test 03

    Guided first-time deposit wizard

    Problem

    First-time deposits sit at the heart of activation — but the deposit card was a single dense screen that asked everything at once: amount, currency, payment method, card details. About half of registered users have no trading experience, and many were dropping out before completing. User interviews surfaced concerns about withdrawals, commissions, and how the process actually worked — but the deposit card never addressed any of them. Two patterns dominated:

    • 18% First-time users who opened the payment-method dropdown
    • 77% First-time deposits made with credit card

    Most users assumed CC was the only option — non-CC payment methods were effectively invisible, even when they were the better fit for the user's region or amount.

    What we tried

    We redesigned the deposit card as a step-by-step wizard, built on four principles: one decision per page to reduce cognitive load, progressive disclosure so we never showed everything at once, push for completion via a progress bar and the Zeigarnik effect, and focus — no FAQ links or distractions that pulled the user out of context. The flow splits into Welcome → Amount → Method of payment → Payment → Confirmation → Success, with stateful behaviour: if a user closes the wizard mid-flow, it resumes exactly where they left off. MoP became a first-class screen with each option labeled by speed and fee, surfacing alternatives to CC that were previously buried. We shipped in priority order — the five markets representing 49% of FTDs first (UK, France, Italy, Spain, UAE), mobile-first.

    Result

    Running as an A/B test through Optimizely against a control of the legacy deposit card. Funnel goals set in Mixpanel with Fullstory sessions tracking exactly where users hesitate inside the wizard.

    • FTD Primary metric
    • FTT · Activation Secondary metrics
    • MoP distribution Diagnostic — did non-CC options become visible?

    Experiment in flight. Will update with measured numbers once it concludes.

    What users were telling us

    • I could only use my credit card it seemed, or a debit card. There's no other options like wiring — no other options to work with eToro.
      Real user · onboarding interview
    • I can't attach my card. I don't know what's going on. I'm not sure what's going on, but I can't — and there's no other options like wiring.
      Real user · onboarding interview
    • That's why I haven't done anything with it. I would have put money on eToro already, but I can't seem to do anything — I'm stuck.
      Real user · onboarding interview

Part 02 · In parallel

Redesigning the flow on the new design system

While the tests on the live flow were running, we designed the next-generation KYC from the ground up on eToro's new design system. Every step rebuilt mobile-first — name input through proof-of-identity upload — to feel less like compliance and more like getting started.

The redesigned flow

Welcome screen — name input
Welcome & name
Date of birth input with numeric keyboard
Date of birth
Gender selection — three tile options
Gender
Country and city selection with search
Location
User's investment goals — multi-select list
Investment goals
Risk profile — selected option highlighted
Risk profile
Source of funds — chip selection
Source of funds
Phone verification — 5-digit code input
Phone verification
Proof of identity — document type selection
Proof of identity
Proof of identity — uploading state
Uploading state