Kiwi.com: redesigning a product that couldn't stop moving.
Kiwi.com kept shipping across web, iOS and Android. The redesign had to be delivered incrementally.
The short version
The challenge
Improve visual quality and consistency across a mature, high-density product without slowing down the teams shipping it.
My role
I led the design-system work with another designer and dedicated engineers, then joined Search & Booking to apply it in the product.
The result
We rebuilt Orbit's token and component architecture, then applied it in active product work across Search & Booking.
Orbit was already a mature design system.
After several years of work, Kiwi.com's design system had more than 50 components across web, iOS and Android, dedicated engineering support and established processes for requests, feedback and reviews.
Product designers knew how to bring new needs into the shared library. I regularly joined teams to review work, explore new patterns and decide what belonged in Orbit. The system reduced inconsistency and made recurring work faster, but screen-level quality was still uneven.
A mature system still produced uneven screens.
The components were consistent, but hierarchy, density and layout often varied. As Kiwi.com expanded, new flows, edge cases and states appeared faster than Orbit could define a clear visual standard for them.
Orbit helped unify the product, yet screen quality still varied. Parts of its visual language had also begun to date.
The first move was an incremental refresh.
I kept the direction close to the existing product so teams could adopt it without waiting for a full redesign. The refresh updated the icon set, introduced rounder corners and revised the typography, colors and shadows to make the interface feel more current as a whole.
I also translated Kiwi.com's updated “Hack the Travel” brand into the product through binary details, grid motifs and a more technical visual character, while keeping those cues secondary in transactional screens where clarity mattered more than expression.
The company reset widened the brief.
The work had received positive feedback and rollout planning was underway. A wider company reset then changed the product strategy. The new brief removed continuity with the existing product and asked us to explore Kiwi.com without its previous constraints.
I used that freedom to push color and gradients further, test lighter iconography and explore new interaction patterns and the object-collage language from the brand. Applying the direction to complete screens and new product areas, including an AI travel companion, gave later iterations a broader set of tested options.
The clean-slate direction had to become shippable.
Replacing Kiwi.com in one move was unrealistic, so I narrowed the focus to Kiwi.com Guarantee and the booking funnel, where the changes could have the most impact. The key question was which concepts improved comprehension and cohesion enough to justify their migration cost.
Together with a content designer and researcher, we built prototypes around unresolved questions and tested them in user interviews. I brought the findings back to stakeholders, then helped split the direction into increments for A/B tests and session recordings. That gave teams a basis for choosing what to ship, adapt or drop.
The redesign went deeper than the visual layer.
Together with another designer and dedicated engineers, I reworked the token structure and simplified components using newer Figma capabilities. The new approach reduced the number of variants, made the architecture more modular and made common changes easier for designers to manage.
We consolidated shared foundations into one main library. Web, iOS and Android kept the system-specific components they needed, while the rest no longer had to be maintained three times.
Engineering shaped the migration.
Every Figma change needed a production equivalent. I worked with dedicated engineers to map the new architecture to production components, check compatibility with existing screens and plan a migration that would not pause team roadmaps.
That collaboration brought the design-system team into product work earlier, so components were tested against real data, business goals and edge cases inside the teams adopting them.
I brought Orbit into Search & Booking.
My product work included a new navigation model, fare selection, AI-assisted search and a wider redesign of the booking flow. These flows combined dense information, commercial targets and technical constraints with existing behaviours that could not simply disappear.
Fare selection had to explain meaningful differences between ticket types while supporting upsell. I led the design iterations on hierarchy, packaging and messaging, then worked with the product team to run repeated A/B tests. The gains were incremental, which is common in an already optimised part of a mature funnel.
Product work became the pressure test.
Designers began using the new library in their own domains while the production component migration was being planned. Working inside Search & Booking showed me what the foundations handled well and what still needed to change.
I fed those findings back into Orbit and aligned the system work with the teams adopting it. The visual language, components and product screens developed in parallel.
What changed
I led Orbit's move to a new token and component architecture with one shared cross-platform library. Designers adopted the foundations while engineers prepared matching production components and the migration plan.
The refreshed language moved into navigation, fare selection, AI-assisted search and the wider booking flow. Those releases gave the rest of Kiwi.com a practical path for adopting the system incrementally.