When Northwind Commerce came to us, their average checkout abandon rate was sitting at 71%—not unusual for multi-region retail, but painful when you know what drives it. Three separate storefronts, each inherited from a different acquisition, each with its own cart logic, form validation quirks, and visual language. Users moved between regions for shipping options and fell off at the seams.
Starting with evidence, not opinions
Before designing a single screen we spent two weeks in analytics. We mapped every step of the purchase funnel across all three storefronts using session recordings, heatmaps, and funnel analysis. The data pointed to three distinct drop-off zones: the shipping options screen, the payment method selection, and the order review page.
Shipping options had 22 permutations across regions, presented as a flat list with no hierarchy. Users from Region B were frequently landing on Region A's checkout and encountering shipping methods that weren't available for their address—only finding out after filling in payment details. This single issue accounted for 31% of total checkout exits.
The most valuable insight: users weren't leaving because checkout was too long. They were leaving because it was unpredictable. Every fix we made was about reducing uncertainty.
The five design decisions that moved the numbers
- 1.Region detection at cart entry, not checkout — users saw only relevant shipping options from the start.
- 2.Shipping tier hierarchy redesigned: recommended first, economy second, express third, with estimated delivery dates shown inline.
- 3.Payment method selection collapsed to a single primary action (the most-used method per region), with alternatives accessible but not dominant.
- 4.Progress indicator redesigned to show estimated minutes remaining, not step count.
- 5.Error messages rewritten to be specific and actionable — "We need a 5-digit US ZIP code" instead of "Invalid postcode".
How the design system made this possible at speed
Doing this across three codebases in six months would have been impossible. We started with the design system: a shared token layer that all three storefronts consumed. Typography, colour, spacing, and interactive states were unified in week two. By week four, new checkout components were being built once and deployed everywhere.
The ops team adopted the same component library mid-project—an outcome we hadn't anticipated. Seeing the storefront components working in production gave their engineers confidence. They migrated their inventory dashboard to the same system in parallel, compressing a planned parallel workstream by six weeks.
Results after 90 days
38%
Conversion lift
61ms
P95 API latency
4×
Faster feature delivery
The 38% conversion lift was measured against the same 90-day window the prior year, controlling for seasonal variation. P95 API latency dropped from 340ms to 61ms as a side effect of consolidating three backends into one. The design system effect compounded: six months post-launch, Northwind's own team had shipped more features than in the previous two years combined.
What other e-commerce teams can apply today
- Audit before designing. Session recordings reveal the actual friction, not the friction you imagine.
- Consolidate before optimising. A design system investment pays for itself in the first project that benefits from it.
- Reduce uncertainty at every step. Users tolerate long checkouts; they abandon unpredictable ones.
- Region-aware UX is not a nice-to-have for multi-market brands—it's table stakes.
“Tachyon didn't just redesign our store—they rebuilt our capability to keep improving it ourselves. Six months in and our own team has shipped more than we did in the previous two years.”
