The challenge
Lumina Retail sells homeware through 14 physical stores and an online shop. The online shop was the problem child: pages took nearly seven seconds to load on mobile, search returned odd results, and the checkout demanded account creation before showing shipping costs. Conversion sat at 0.9% while paid traffic kept getting more expensive.
Marketing was spending well and losing the results at the door. Campaign click-throughs were healthy, but session recordings showed the same story on repeat: a shopper lands, waits, scrolls, adds to cart, meets the forced signup, and leaves. Cart abandonment was running at 82%.
The approach
We treated this as one project with two halves that had to land together: an experience redesign grounded in real shopper behaviour, and a storefront fast enough to carry it. We mined a year of analytics and 200 session recordings before sketching anything, then rebuilt the funnel around the shortest honest path from landing to paid.
- Analysed 12 months of funnel data and 200 session recordings to find the exact drop-off points
- Rebuilt the storefront on Next.js with a hard performance budget: every page interactive in under two seconds on 3G
- Redesigned checkout as guest-first, with account creation offered after payment, not before
- Ran the redesign through five rounds of usability testing with actual Lumina customers
- Retuned paid campaigns and landing pages alongside the build so launch traffic matched the new funnel
What we built
The new storefront is a server-rendered Next.js application with a design system built for merchandising: fast category pages, honest product photography layouts, and a checkout cut from five steps to two. Search was rebuilt around how shoppers actually type, including typos and vernacular product names the old engine missed.
- A performance-first storefront with a 1.4 second median load on mobile
- Guest checkout in two steps, with shipping costs shown before any form is filled
- Rebuilt product search with typo tolerance and merchandising controls for the trading team
- A component design system the in-house team now uses to ship campaign pages without engineers
- Structured product data and clean URLs that lifted organic search traffic 28% in three months
- Email and remarketing flows wired to real funnel events: browse, abandon, restock, replenish
0.9%
conversion rate before
1.5%
conversion rate one quarter after launch
61%
cart abandonment, down from 82%
The outcome
One quarter after relaunch, conversion had risen from 0.9% to 1.5%, a 65% improvement, on essentially the same traffic. Cart abandonment fell from 82% to 61%, with the guest checkout doing most of that work. Because the funnel stopped leaking, the same marketing budget produced 58% more revenue, and cost per acquisition dropped by a third.
The gains have compounded since. Returning customer purchases are up 41%, helped by replenishment emails tied to what people actually bought. The trading team ships its own campaign pages using the design system, so seasonal pushes no longer queue behind a developer. Lumina hit its best online trading day ever three weeks after launch, without a sale running.
