Rebuilding checkout for a global marketplace
A research- and data-heavy checkout redesign for a high-traffic e-commerce company, co-led with Thoughtworks. I joined at the start of the project and led the product design strategy.
Results
The challenge
Modanisa was moving from selling its own stock to a marketplace model, with suppliers shipping directly and new warehouses opening abroad. The goals were faster delivery, easier returns and a wider range. At the same time, the legacy platform was being rebuilt, which made the project much larger.
Deliveries and operations
- Supplier items arrived late, hurting satisfaction and pushing delivery times below industry standard.
- Suppliers couldn’t fulfil orders from their own warehouses.
- Delivery and return costs to Europe were high, so returned items often missed the season.
Checkout
- Slow, outdated infrastructure and high drop-off on the checkout page.
- Technical dependencies blocked experiments and new payment methods.
- Address forms with little guidance produced many wrong addresses, failed deliveries and extra work for customer care.
How I worked
-
Research and strategy
I organized and led lean inception workshops with the team and stakeholders to agree on user needs and the MVP. The roadmap went from a pilot supplier shipping directly in Turkey, to a new warehouse in Poland for faster European delivery, to a new worldwide checkout and a warehouse in the MENA region.
I ran one-hour moderated interviews with 10 high-frequency customers across countries, and involved stakeholders in writing the questions so business and UX were aligned from the start. We also interviewed suppliers, benchmarked 15+ leading e-commerce checkouts (plus Baymard research), studied Hotjar recordings and heatmaps, and worked with Business Intelligence on address, payment and order data to set a baseline for every decision.
Customer interview notes, grouped on an affinity wall to find patterns. -
Design and test
We redesigned checkout feature by feature across four web breakpoints, iOS and Android, while I contributed checkout components to the new design system. To stop approvals becoming a bottleneck, I introduced daily 15-minute design reviews, plus weekly feasibility sessions with developers.
From one crowded page to two clear steps: address and shipment, then payment. Checkout components I contributed to the new design system. -
Implementation and review
I joined every story’s analysis and dev-box testing, owned the UX/UI backlog and did design QA. I also ran a workshop teaching QA testers to spot UI and UX bugs, which clearly reduced errors in the final product.
Key design decisions
Address section
Data showed 93% of customers had only one saved address. When asked how they recognize an address, 60% said the receiver’s name, 30% the city or street and 10% the address title.
We showed one address by default with a separate interface for changing it, highlighted the receiver instead of the title, and made the title optional.
Localized address forms
With Logistics, we found the fields where people made the most mistakes. Address formats also differed by region.
We cut every field we could, and localized fields, helper text and placeholders for each zone, reducing failed deliveries and return costs.
Split orders and a two-step checkout
For the first time in ten years, one order could arrive in several packages, and the cart was out of scope because of legacy dependencies.
After many drafts, we split checkout into two pages, address and order review followed by payment, so shipment details had room without pushing payment below the fold.
Payments
The average customer had 1.09 saved cards, Klarna required an email that bloated the address form, and payment messages looked like errors.
We showed a single saved card, moved Klarna’s email field into the payment step and redirected to Klarna’s own interface, made messages neutral, added a PayPal checkout button, and improved the cash-on-delivery flow with OTP confirmation.
The hardest trade-off: multi-warehouse delivery. For the Poland warehouse, leadership wanted a temporary two-stage delivery choice (one package or several, then the delivery speed). BI forecast it as the most profitable model, but it was complex for users and expensive to build for a planned six months.
I designed a version that kept development cost low while meeting both the team’s and the stakeholders’ needs. I consider it one of my biggest contributions to the project.
Usability testing
Prototype testing before launch was cancelled, so we tested live with a low-traffic pilot supplier. I ran moderated, think-aloud sessions remotely on desktop and mobile, covering the end-to-end journey from search to payment, after refining the script through three internal pilot sessions.
- Shortened and reworded the address form again, cutting drop-off by a further 12%.
- Found and fixed a large share of 3D Secure payment failures.
- Caught alignment, button and text-size bugs QA had missed.
- Reworked font sizes, weights and colors for readability and accessibility.
What I’d do differently
- Test with users earlier and more often while designs are still in progress.
- Put accessibility on the agenda from day one.
- Push for A/B testing infrastructure sooner, and be more proactive about data gaps.
- Give more attention to micro-interactions.