product designer • Caravelo • web • 2025
subscription management | −27% tickets
Turning a legacy back-office tool into one airline partners can run themselves.
Components reused from Veloair, the design system I built for Caravelo’s consumer products
User interviews & contextual inquiries
External tools (Kaui, Kill Bill) replaced by in-platform workflows
Plan-change and cancellation tickets in the first 6 months
tl;dr
Caravelo’s back-office platform had grown for years without a designer looking after it. Simple tasks like changing or cancelling a subscriber’s plan needed a support ticket: someone at the airline opened it, and someone at Caravelo made the change in two other tools, Kaui and Kill Bill. I was the only designer on the redesign, working with a PM and two web engineers. We reorganised the platform around the tasks airline teams actually do, integrated Kaui and Kill Bill so partners could make those changes themselves, and built it on Veloair, the design system I’d made for Caravelo’s consumer products. Tickets for plan changes and cancellations dropped by 27% in the first 6 months.
overview
Caravelo powers subscription programmes for airlines. Its Subscription Management Platform is the back-office tool used to manage subscribers, plans, payments, and day-to-day operations.
Two groups used it every day: airline teams in sales, marketing and support, and Caravelo’s customer success and support staff. The second group spent much of their time doing things for partners that the platform didn’t let partners do themselves, mainly plan changes, cancellations and chargebacks.
The goal
Let airline teams handle their own day-to-day work in the platform, so Caravelo’s team doesn’t have to do it for them.
problem
I reviewed the platform against usability heuristics, watched people use it during their actual work, and looked at how other SaaS back-office tools handle the same kinds of tasks.
- Too many tools. Core tasks meant switching between the platform, Kaui, Kill Bill and Jira.
- Partners couldn’t act on their own. Many changes needed a support ticket, so partners waited while Caravelo’s team did the work.
- Hard to build on. One-off Vue components made the platform slow to maintain and extend.
- Behind similar products. Comparable SaaS tools let users finish these tasks themselves, in one place.
discovery
I was the only designer, working with a PM and two web engineers. I ran 6 interviews and contextual inquiries: 3 with Caravelo’s customer success and support staff and 3 with airline partners. The same frustrations came up on both sides: too many tools, and too little control over simple changes. I turned the findings into journey maps and a new information architecture.
Why self-service
Every plan change that goes through a ticket costs the airline time and costs Caravelo a support person’s time. If partners could make these changes themselves, Caravelo could take on more airlines without growing its support team at the same rate.
design decisions
Most decisions came down to one question: does this still need someone at Caravelo to step in?
-
Navigation built around tasks
The old navigation mirrored how the legacy system was built, not the tasks people came to do. I reorganised it around those tasks, using the journey maps from the interviews.
-
Plan changes and cancellations without a ticket
Before, a plan change or cancellation took time from both sides: someone at the airline opened a ticket, and someone at Caravelo assessed it and made the change across Kaui and Kill Bill. These were most of what Caravelo’s support team handled for partners, so they were the main target. Kaui and Kill Bill are now integrated in the backend, and airline teams make these changes themselves, from the subscriber’s profile in the platform.
-
A way into support from inside the platform
For anything that still needs Caravelo, I added an entry point in the platform that opens the Jira support portal, so users can reach support without leaving the platform. It was the first piece we shipped.
-
Building on what already existed
Instead of starting from scratch, I extended Veloair, the design system I’d built for Caravelo’s consumer products. 48 of its 70 components worked as they were, and I added the patterns a back-office tool needs that a consumer product doesn’t. With two engineers, reusing what already existed meant more of their time went into the new workflows.
-
Connecting Figma to the code
Figma had just released its MCP server, and I suggested using it here: connecting Figma to the code repositories so engineers could generate components from the designs instead of rebuilding them by hand. Other teams at Caravelo had already cut component build time by nearly half this way. I also designed around the integration constraints from the start, so the designs were ready to build.
validation
I took the designs back to the same people I’d interviewed at Caravelo and at airline partners, and reviewed them regularly with the PM and engineers to check what was feasible.
-
Payments & billing
Support staff needed to see transactions and billing documents side by side when helping a customer. → We merged payments and billing into one section.
challenges & trade-offs
-
Chargebacks stayed in Stripe
Chargebacks were the other big source of support work, but rebuilding that flow inside the platform was too much for the scope. Instead, we embedded Stripe, the tool already used for chargebacks, inside the platform, so users could handle them without leaving it.
impact
-
Fewer tickets
Tickets for plan changes and cancellations dropped by 27% in the first 6 months, because airline teams could now make those changes themselves instead of asking Caravelo.
-
Two fewer tools
Kaui and Kill Bill were no longer needed for day-to-day work. Airline teams do it from the platform.
-
Shared foundations
The back office now uses the same design system as Caravelo’s consumer products, so new screens can reuse existing components instead of being built from scratch.
see other projects
couple onboarding | targeting +16% completion
See project →travel search | +8 pts NPS
See project →wedding vision | AI prototype
Case study coming soon