Site experience
Why did someone visit, and could they do what they came for?
Experience design / Vista Musical Instruments
Swee Lee
Helping customers find their next step.
Find the right instrument. Discover an alternative. Continue the experience in-store. A body of work connecting customer research, shared systems, and the everyday decisions that make those journeys possible.
- Role
- UX/UI Design Lead
- Period covered
- Jan 2023 - Oct 2026
- Scope
- Sitewide storefront audience: 200,000+ monthly users
01 / Customer research
A research programme built to inform decisions.
The existing surveys had overlapping purposes, inconsistent questions, and different triggers. I facilitated sessions with business leads to audit and prioritise the questions, then defined three instruments with distinct jobs. Existing questions were retained where useful to preserve continuity.
- Generate
Question ideation
Crazy 8s brought business leads into generating and challenging the questions, rather than commissioning another disconnected survey.
- Audit
Start, pause, continue
Existing questions were reviewed for purpose, duplication, clarity, and mobile usability.
- Prioritise
Tier the questions
Prioritisation translated the workshop into three instruments, with each question tied to a specific research purpose.
Post order
What shaped the purchase decision and the route to it?
Post purchase
How did delivery, the product, and support meet expectations?
Low navigation scores opened a follow-up question about browsing difficulties, so the score could arrive with an explanation. Master templates supported market-specific links; English and Malay interview scripts extended the research to stores.
The Site Experience trigger changed to six minutes on desktop and four on mobile, without requiring items in the cart. Post-purchase feedback moved to a ten-day survey and a fourteen-day product review, retiring the thirty-day flow.
A separate intervention / May 2026
A more accessible feedback entry point.
A checkout survey entry point rolled out on 14 May 2026. Quarterly participation across the two live instruments rose from 399 to 568 (+42%). Site Experience responses increased from 44 in Q1 to 222 in Q3.
The participation gain followed the 2026 entry-point change, not the earlier question redesign; response volume does not measure customer satisfaction.
Q3 202531 31
Q4 202535 35
Q1 202644 44
Q2 2026149 149
Q3 2026222 222
02 / 2023 foundation
Build the research, measurement, and design foundations.
Within my first month, I developed online personas for Swee Lee, Teisco, MONO, Harmony, and Heritage, alongside a survey insights deck and research into customers' primary online objectives.
When Google sunset Optimize in February 2023, I sourced its A/B testing replacement. During the Universal Analytics migration in May, I learned and adopted GA4. I then established a plan, launch, and review rhythm for VWO testing.
Design the Dawsons storefront
In June 2023, I designed eleven page templates: homepage, product, contact, brands, departments, cart, about, login, account creation, account details, and navigation. The delivery also covered key screens, all email templates, homepage and USP banners, and a text styles kit.
Rebuild the shared design system
I moved component libraries from Sketch to Figma, named undocumented Shopify components, reconciled them with the UI kit, and retired archive libraries. The Master UI directory standardised fields, labels, dark and light button states, chat widget colours and icons across brands.
Revamp WPG
The brand update covered its colour palette, UI kit, secondary pages, and collections navigation within the wider shared-component system.
Train and direct other designers
I delivered Figma and UI training to colleagues and directed Master UI directory work across multiple contributors, building shared ownership of the system.
Apply AI to creative operations
The work progressed from Midjourney image generation in January 2024 to AI-generated product descriptions in 2025 and AI-assisted brief screening, which flags missing information before work enters delivery.
03 / Product-page continuation
Give every product page a next step.
A product page was a dead end for shoppers who were not ready to buy. I treated discovery, relevant alternatives, and store availability as one continuation strategy: help people keep browsing, compare options, or find the product in-store. The work shipped in stages, not as one experiment.
- 2 April 2024 / Design began
Make browsing easier.
Customer-shaped navigation and contextual search quicklinks created more routes through the catalogue.
- 19 November 2024 / Live
Offer a relevant alternative.
Customers Also Bought used existing co-purchase data across Singapore, Malaysia, and Indonesia.
- 5 May 2026 / Live
Connect shoppers to stores.
Product-page stock availability linked online product discovery to physical store inventory across the same markets.
01
Make browsing easier.
Navigation and search quicklinks
Card sorting shaped the category structure around how customers group products. Desktop wireframes and a mobile full-screen navigation concept began in April 2024. Quicklinks surfaced popular searches in the search bar, giving shoppers another route when they had nothing obvious to click.
Dedicated navigation events captured each interaction and its label. In a 14-day Singapore review, 192 users generated 241 quicklink events; specific gear names drew more use than generic terms in that snapshot. Customer surveys also show navigation satisfaction improving over time.
Interaction counts describe use; they do not establish task completion.
02
Offer a relevant alternative.
Customers Also Bought
Working with developers, I explored a co-purchase carousel using available customer purchase data, without adding a paid recommendation plugin. It launched across Singapore, Malaysia, and Indonesia in November 2024.
Observed: 14,488 recorded interactions from 3,712 users in the lifetime review, approximately 3.9 interactions per recorded user.
These are interactions over the reporting period, not per visit. They do not establish additional purchases or prevented abandonment.
03
Connect shoppers to stores.
In-store stock at the point of decision
Store availability appeared above the fold on the product page, outside an accordion, with a route to store-by-store details. It offered an answer when an item was unavailable online, without asking customers to call or guess. Public launch followed QA on 21 April, on 5 May 2026; April testing is excluded from the launch baseline.
Observed: Singapore's monthly interactions per user stayed within 1.98-2.10 from May to September and remained similar five months after launch.
Stable interaction frequency shows sustained use among feature users, not proof of improved conversion or store visits.
Stock availability / Singapore
Five months later,
a similar pattern.
Equal 14-day windows, five months apart.
Cumulative feature engagement: 17,883 users across Singapore, Malaysia, and Indonesia through 2 October 2026.
| Measure | 5-18 May | 19 Sep-2 Oct |
|---|---|---|
| Users | 1,096 | 1,166 |
| Interactions | 2,271 | 2,313 |
| Per user | 2.07 | 1.98 |
These features were delivered at different times. They are grouped here by the customer problem they address, not presented as a single experiment or a proven combined conversion lift.
04 / Retail & kiosk
A connected in-store experience.
Research, launch touchpoints, and the appointment journey at Swee Lee Clarke Quay.
Retail research extended beyond the kiosk: surveys in Malaysia and Indonesia, in-person interviews in Singapore and Malaysia, and qualitative testing in Swee Lee stores brought local context into the design work.
The Clarke Quay launch included digital and printed store assets, cafe templates, and iPad screen content. Related retail work covered signage, Jakarta murals, and pillar design in Malaysia.
My responsibility
I designed the kiosk experience and proposed direct appointment booking on the kiosk, with phone-number confirmation. That proposal did not launch.
The business constraint
The wider business prioritised adoption of VMI Connect, its membership app. I contributed to that initiative but did not own the app or the strategic requirement to route visitors through it.
Proposed / not launched
Complete the task at the kiosk.
- Choose an appointment at the kiosk
- Book directly
- Receive phone-number confirmation
Implemented / kiosk entry
Continue through the membership app.
- Discover the appointment at the kiosk
- Scan the QR code and download VMI Connect
- Continue to the Swee Lee Experience appointment form
- Submit a request; staff follow up to arrange it
Customers who discovered the appointment page through the online store could reach it without the kiosk's required app-download step. The same service therefore had different entry requirements depending on where the journey began.
Swee Lee Experience appointment page ↗The kiosk measurement boundary.
Kiosk sessions declined during the reporting period. The app-first direction provides context, but the kiosk data alone cannot tell us whether visitors abandoned the task or completed it elsewhere. Evaluating this journey requires visibility across the kiosk, app, appointment request, and eventual visit.
05 / Measurement
A measurement system for connected journeys.
Navigation and quicklinks supported discovery; recommendations and store stock opened onward paths. The next challenge was connecting those interactions to what happened downstream.
Weekly feature reviews combined event analysis with store feedback and shopper conversations. Experimentation covered search prominence, recommendations, price presentation, and add-to-cart variants, alongside qualitative research.
Connecting Shoplift variant IDs to Microsoft Clarity tags made it possible to filter recordings and heatmaps by test variant. The aim was to understand behaviour behind the numbers, not just report a winning variation.
Set the question before launch.
For gift with purchase, I defined basket-size and redemption measures before delivery, designed the threshold states, and asked how tracking could be automated. The later AOV data did not establish a lift attributable to the feature.
Check the computation, not just the interface.
A repeated co-purchase percentage led to investigating the underlying calculation. A display-only fix risked leaving the same data problem in another surface.
Find the missing connection.
Feature reviews exposed events that identified a product but not the originating component. The reporting available could not establish whether a stock check or recommendation led to a cart add. I specified graded tracking options suited to the shared custom-component system.
What this work changed in my practice
Workshops establish the question. Shared systems make delivery sustainable. Store research exposes what a dashboard misses. Post-launch evaluation determines which conclusions are justified and which decisions need better evidence.
My contribution is connecting those stages, while making customer needs, business priorities, and the limits of the evidence visible to the people making the next decision.
Selected work and reporting through 2 October 2026. Feature usage and response counts are descriptive observations, not attributed revenue outcomes. Earlier archive images are shown for design context.