Back to work
Fintech

TikTok Pay KYC — scaled across markets

KYC across Vietnam, Malaysia, Indonesia, and the US for TikTok Pay, Shop, and Live. Vietnam new‑user completion 70% → 90%. Vendor docking SOP cut integration cycles by ~40%.

Role
Product Designer (Intern)
Period
Jun — Sep 2025
Client
TikTok · PIPO UED
Themes
Fintech · Compliance · Multi-region · Design System
TikTok Pay KYC — scaled across markets
Chapters
Background

What KYC is

Regulatory gate before any payment can run. Break it, users drop off. Loosen it, the regulator does.

  1. Identity verification

    Local ID + selfie + liveness check, per market.

  2. Compliance contract

    Each regulator audits the flow, the data, the SDK.

  3. 4 frameworks

    Same gate, different rulebooks.

    • SBV
    • BNM
    • OJK
    • FinCEN

“When KYC breaks, users drop off. When it’s loose, the regulator does.”

the design constraint, in one sentence

Where the work landed

Four markets. One design system. TikTok Pay · Shop · Live across all.

  • Vietnam (VN): regulator SBV; Pay, Shop, Live
  • Malaysia (MY): regulator BNM; Pay, Shop
  • Indonesia (ID): regulator OJK; Pay, Shop
  • United States (US): regulator BSA · FinCEN · CIP; Pay, Shop, Live

One design system

  1. Vietnam (VN): regulator SBV; Pay, Shop, Live
  2. Malaysia (MY): regulator BNM; Pay, Shop
  3. Indonesia (ID): regulator OJK; Pay, Shop
  4. United States (US): regulator BSA · FinCEN · CIP; Pay, Shop, Live
Strategy

Three layers of the problem

UX · Compliance · UI — three distinct failures, three distinct fixes. None solvable alone.

Problem

Users reusing historical identity data had no fallback when their ID couldn’t be found — flow forked into silent drop‑offs.

Approach

“Not My ID” escape path + Goal Gradient progress visibility — the user always knows where they are.

Problem

OCR fields inconsistent MY ↔ VN. Manual re‑entry. Regulatory risk in the inconsistency itself.

Approach

Standardized field schema, co‑authored with Legal + Risk, regulator‑aligned across all four markets.

Problem

Vendor SDK type, radius, icons, errors fragmented. Trust collapsed at the handoff to vendor UI.

Approach

SDK‑AAI Style Alignment + reusable templates + secondary UI encapsulation for clean country variants.

Three audiences, one flow

Each KYC change had to satisfy all three at once.

One flow

  1. Business

    Lift completion to unlock payment volume across markets.

  2. User

    Onboarding that feels simpler, clearer, more trustworthy.

  3. Design

    Framework that absorbs many SDKs without falling apart.

Execution

The flow, with the escape path

Most flows assume the happy path. KYC’s failure mode is the silent fork — a user who can’t proceed but doesn’t know why. We surfaced the alternate route as a first‑class step.

Vietnam KYC · v2
  1. Open TikTok Pay

    Entry · prefilled identity from history

  2. Choose ID type

    Citizen ID · passport · DL

  3. Verify identity

    OCR + liveness

  4. Branch: Not My ID

    Re-enter manually

    Surfaces the alternate route — drop‑offs collapse from 30% to ~10%.

  5. Confirm match

    Goal Gradient progress · 3 of 4 done

  6. KYC approved

    Pay unlocked

Completion across markets

Vietnam was the trial — once it landed, the same pattern (escape path + Goal Gradient + standardized OCR) ported to MY · ID · US.

KYC funnel · Vietnam · v1 → v2
StepValue
Open KYC100%
Upload ID92%
Pass OCR88%. OCR taxonomy standardized · MY/VN aligned
Liveness90%
Approved90% (v1: 70%). Up from 70% in v1 — “Not My ID” + progress.

Vendor docking SOP

First standardized SOP for TikTok PIPO — workflows, acceptance criteria, UI delivery priorities. The artifact I’m proudest of.

Workflow, Acceptance, Delivery. Vendor integration cycle −40%.

Workflow

Stages, owners, acceptance gates per SDK vendor.

Acceptance

Checklist tied to OCR mapping + UI templates + halts.

Delivery

UI tokens, component library, country‑variant rules.

Vendor integration cycle

−40%

“Cross-market scalability stopped being a function of who was on shift.”

what the SOP changed

Reflection

Compliance‑first UX is its own discipline — regulator language and human language don’t compromise. The right answer was less about fixing one flow, more about making the SDK system modular, configurable, data‑driven.