Portfolio demonstration · Senior Mobile Product Manager & UX Tester
A working audit lab for Dangdi, a social-commerce platform for Chinese communities outside China. It shows how I go from a user’s intent to a verified, prioritized, measurable fix — and keeps every claim traceable to the screen it came from.
支付方式
收货地址
社区发现
Community discovery
Lin, 27 · Berlin
Find a trustworthy housing post and join a local community group within one session.
Analyze journey
本地生活团购
Local commerce
Mrs. Chen, 46 · Toronto
Buy a hot-pot voucher for Saturday and redeem it in store without help.
Analyze journey
旅行灵感
Travel discovery
Zhou, 33 · Sydney
Save a creator itinerary, share it with family and use it offline on the trip.
Analyze journey
How I work
Six stages, each with a concrete artifact in the lab. The stages feed one another: a journey step produces findings, findings produce tickets and test cases, and tickets carry the metric that tells us whether the fix worked.
Map the jobs diaspora users hire the app for. Personas, markets and step-level emotion, friction and drop-off — so the audit starts from intent, not screens.
Annotate real screens against heuristics, accessibility and Chinese-localization rules. Every finding is pinned to pixels, a device and a build.
Turn findings into tickets and rank them with RICE. Reach and confidence are explicit and editable, so the ranking can be challenged in the open.
Write reproducible test cases across a device matrix that includes HarmonyOS and HyperOS phones without Google services. Record every run.
Hand engineering tickets with acceptance criteria, the original evidence and the linked tests — nothing gets lost between design, QA and dev.
Close the loop with a baseline, a target and an owner per ticket. The report separates verified results from illustrative examples.
Traceability
Stakeholders shouldn’t have to trust a summary. In the lab, any finding opens a trace: the annotated screenshot, the observation and impact, the recommendation, the test cases that prove it, the backlog ticket that fixes it, and the metric that measures it.
Follow F-08 through the labChinese localization, abroad
Diaspora users live between two sets of conventions. The details that break trust are rarely the words — they’re currencies, wallets, addresses, fonts and the phones people actually carry.
简体 / 繁體
Script is a reader preference, not a region. HK, TW and MY diaspora read Traditional first.
Pinyin search
“zufang” and “zf” should find 租房 — many users type on German or English keyboards.
¥ is ambiguous
CNY and JPY share a symbol. Always render ISO-aware currency: CA$128 · 12,000日元.
Wallet eligibility
WeChat Pay and Alipay often need a mainland bank card. Check before the PIN, not after.
Addresses & names
Mainland 省/市/区 forms reject “M5V 2T6”. Use per-country schemas and split 姓 / 名.
CJK typography
Han unification makes 关 render Japanese without lang tags. Apply 避头尾 line-breaking.
No-GMS devices
Huawei (HarmonyOS) and CN-ROM Xiaomi phones lack Google push and maps. Test them.
WeChat sharing
Links shared to family in China need preview cards and an H5 page that loads there.
Evidence policy
Seed content that demonstrates the method. The screens in this demo are drawn wireframes, not captures of the live Dangdi app, and the funnel numbers are invented to make prioritization concrete.
A finding captured from a screenshot you upload. It is a real observation, but it hasn’t yet been reproduced on a named device and store build.
Reproduced by the auditor on a named device and build, with a reproduction note and, ideally, a recorded test-case run. Only these count as audit results in the report.
Try it with your own screens.
Upload a screenshot, draw a box, rate severity, and watch it flow to a test case and a ranked ticket. Everything is saved in your browser.