Skibookers is an AI-powered platform for planning and booking ski trips end-to-end
A seed-stage travel-tech startup competing with the classic OTA model (Booking or Expedia). The business bet was on a smart recommendation system powered by the parent company's data about ski travellers.
This case is about how the core product bet broke under real users, and how I reframed the product without giving up its positioning.
Role
Founding Senior Product Designer
Timeline:
July 2025 → April 2026
Deliverables
Core user flow from 0 to launch, design system, multi-step quiz with gamification, landing page, brand identity. Mobile & desktop
Team
CEO, CTO, frontend and backend engineers, PM (last 3 months only)
Chapter 1. The bet
To become the AI-first player in ski travel through a new interaction model
1.1 Business context
1
Research* showed that users are willing to spend up to 10 minutes on onboarding if the perceived value is high enough**
2
Parent company's data: 1.6M+ customers, 584 ski resorts. Operational data could power an AI-native recommendation layer.
3
The OTA model is saturating. AI-first travel players are emerging (Mindtrip, $22M Series A), and big players are adapting (Expedia in ChatGPT, TripAdvisor AI Planner)
1.2 Team setup
Product team
CTO, 1 front-end engineer, 1 back-end engineer, and myself as the sole designer.
No PM in the first 7 months. product decisions were made within a CEO–engineering–designer triangle. I shaped product hypotheses alongside the team, not just their execution.
Scope to build
Design-system, quiz, offer page, booking flow, brand identity. Mobile + desktop.
Deadline
3 months until MVP launch
1.3 Full design scale. Customer Journey Map for MVP
11
surfaces across discovery, offer, and conversion. This case focuses on the offer page
Chapter 2. First design iteration
Dashboard with one curated recommendation
The bet was that the recommendation engine could be confident enough to skip comparison: one ideal trip, one CTA, one click.
2.1 Results
5%
Conversion to next step > Booking Details
2.2 Why it didn't work
1
The engine missed basic constraints. Quiz recommendations didn't match user inputs
Example: a user with a €1k budget would get a €4k Chamonix offer. The offer became irrelevant before users could trust it.
2
Users wanted to compare, not commit. The bet assumed AI confidence high enough to skip comparison. Research showed otherwise — users described wanting to evaluate options, not accept a single curated trip. Verbatim: "I don't see what else there is. Where are the other options?"
2.3 Design feedback
User feedback on the v1 design:
"Apple-like, clean and accurate."
The diagnosis: rebuild the offer page around the actual product constraint — users want to compare, recommendation engine can't be one-choice confident. Design execution stays, product format changes
2.4 The team decision: postpone the MVP launch
After diagnosing v1, the team agreed the product wasn't ready to scale. The December launch was cancelled.
The company brought in a PM, expanded engineering capacity. February became the new ship date.
Chapter 3. The rebuild
The new design.
Skibookers is on live
Beyond rebuilding the recommendation system, redesigning the offer page was essential
3.1 Alternatives
Place all offers inside a horizontal scroll
Use editable pills for location and dates + horizontal scroll
Standard inputs (location, dates, travelers) + card-based layout, mobile-first, 100vh scrollable area with previews of other offers.
Doesn’t solve the issue on mobile
mobile issue remains, and the dashboard feels unscrollable
Solves mobile constraint. Preserves "we curate, not filter" positioning by limiting choice to 3-5 vetted options, not Booking-style filters.
3.2 What changed compared to v1
Introduced Quiz Edit instead of forcing users to retake the full quiz. Users could directly edit key inputs (dates, location, budget) through simple controls. Together with the PM, we identified which questions actually affected recommendation results.
Mobile-first approach. Rebuilt the entire flow for mobile (74% of traffic). Each card became a standalone offer, creating a unified interaction pattern between the quiz and the offer page.
Made the choice explicit. Users could clearly see the number of offers available and compare them. This addressed feedback like: “I don’t see what else there is.”
System consistency. Offer, flight, hotel, and transfer cards now follow the same visual logic. Previously, the quiz and offer flow felt like separate systems.
Reduced card height so each offer fits within a single screen without scrolling.
Added resort information cards, filling a gap identified in the first version.
3.3 Outcome
5%→22%
Offer page → user details conversion increased
The improvement exposed the next bottleneck: the user details form (593 reached, 59 completed, 90% drop-off)
Chapter 4. Summary
Conclusion & Reflection
4.1 Outcome
2
Launches in 10 months (Dec failed, Feb shipped)
1 major redesign (offer page) for mobile an desktop
89%
Of quiz starters reach the offer page. Quiz worked perfectly according to quality and quantity researches
x4
Growth after Quiz offer page > User details
0 sales
Root cause was outside design: pricing fit, supply availability, audience-channel mismatch.
Holding company paused Skibookers
4.2 What I would do differently
1
Validate willingness to pay before optimizing the funnel. We focused UX research on usability: flow clarity, comparison preference, conversion friction.
2
Delegate system work to keep product focus. My attention split between system-level decisions, product decisions and brand work. The design system could have gone to an external designer earlier.
Appendix: Sources of data
Contact
Paul Schneider © 2024