Overview
Project Helix redesigned the cart and checkout of Coppel's web and app while the store was migrating to Salesforce Commerce Cloud. We designed in Q4 2024, built during the migration (Q2 – Q3 2025) and launched in September 2025.
Role and team
We were a team of seven: three UI designers, two UX designers, one UX writer and me as Design Lead for Purchase & Payments, working with Product Owners and developers. I orchestrated the experience:
- I kept the purchase rules intact from the cart to the end of checkout while the platform underneath was replaced.
- I pushed accessibility into the components themselves (accessible properties, not patches) and reworked the layout of Cart, Delivery details, address forms, payment method selection and the Thank You page.
- I worked on how we worked: helping the team embrace uncertainty, get better at understanding problems before designing, and build a shared understanding with POs and developers.
The challenge
Customers liked Coppel, but buying online was harder than it should have been:
- Stagnant conversion. Experience metrics were stable, but web conversion stayed low.
- An accessibility gap. The app's accessibility score lagged far behind the web, leaving out part of our customers.
- Intent without a path. Customer satisfaction showed people wanted to buy; the checkout itself was the barrier.
The values behind these metrics are confidential, so this case study reports changes only.
Process and the hardest decision
The hardest decision was to change the experience during the platform migration at all. Any change had to be millimeter-precise: one broken rule could stop purchases. We negotiated each change with product and development, and when information was incomplete we chose the best decision we could defend with what we knew, instead of waiting for certainty.
What we changed
1. One cart for every category
Problem: Customers couldn't combine categories like motorcycles, clothing and furniture, so one purchase intent was split into several.
- Unified cart: one cart that processes several categories in a single transaction.
- Save for later: when a customer pays for one group of products, the rest stay in the cart.
- Auto-recovery: the cart is restored after interruptions or connection errors.
2. Information architecture and content
Problem: Content wasn't structured for customers with basic digital skills.
- Scannable writing: short copy that lowers cognitive load and guides each step.
- Conversational tone: instructional, natural language that teaches as it goes and narrows the digital literacy gap.
3. Coppel Credit and card payments
Problem: Frequent bank rejections and friction in the final steps.
- Logical sequencing: earlier choices simplify the steps that follow.
- No more Order Review: we removed the redundant review step (7 steps → 6). That let us move card payments to a secure iFrame, which cut rejections caused by expired dynamic CVVs.
Results
Measured in January 2026, four months after launch. Usability is reported as SUM, the Single Usability Metric (Sauro and Kindlund), which combines task completion, time on task, errors and satisfaction into one score.
+2 percentage points
Web conversion
down 60 percent
Card rejections from expired dynamic CVVs
+30 points
App accessibility score
+7 percentage points
Usability (SUM)
| Metric | Change | Source |
|---|---|---|
| Web conversion | +2 pp | Digital analytics |
| App conversion | Held steady (+0.1 pp) despite higher traffic | Digital analytics |
| Dynamic CVV rejections | −60% | Operations |
| App accessibility | +30 pts | Lighthouse, axe DevTools |
| Web accessibility | +4 pts | Lighthouse, axe DevTools |
| Usability (SUM) | +7 pp | CX & Research team |
| E-commerce NPS | Held steady (+0.13 pts) | Medallia and manual measurement |
The redesign improved web conversion, app accessibility and usability, and kept app conversion and NPS steady while the platform underneath changed.
What I learned
- No big bangs during structural change. Redesigning the experience and the process in the middle of a platform migration multiplied the risk. Next time I'd separate them.
- Empower the team to decide. I acted as the team's experience filter. I'd rather build a high-performing team that makes those calls without going through me.
