The Challenge

Travel has a trust problem and a discovery problem, and they're the same problem.

Travelers don't trust generic OTA listings. Reviews are anonymous, itineraries are vague, and there's no accountability after the booking. Younger travelers have mostly moved their discovery to TikTok and Instagram anyway, but those platforms don't connect to anything bookable. You see a great tour in Kyoto, you close the app, you forget about it.

On the supply side, experience hosts are good at running experiences, not building audiences. They need to fill bookings but mostly hate doing their own marketing.

Content creators can build audiences and drive discovery, but travel content doesn't convert to income without a booking layer underneath it. Three parties with complementary needs and no mechanism connecting them.

My Role

I led product and design across the MVP, working with distributed squads across Canada, Pakistan, and Brazil. The work covered defining the marketplace mechanics, translating them into a coherent product system, and creating the frameworks that let the team move fast and stay aligned.

That included product strategy and roadmap sequencing, design leadership across the full app, cross-functional orchestration across three time zones, and restructuring how the team worked in Plane after migrating from Jira.

The Marketplace Design

Before any UX work, we had to solve a classic chicken-and-egg problem: no experiences without hosts, no hosts without travelers, no travelers without content, no content without creators who have a financial reason to produce it.

The answer was a profit-sharing model that gives every participant a reason to do the platform's distribution work voluntarily.

Hosts are good at running experiences, not marketing them. By partnering with content creators, they outsource discovery without paying upfront. Creators earn on bookings, so the incentive to perform is built in.

Content creators post travel content to their existing audiences. When someone books an experience through their content, they earn a cut. That gives them a financial reason to push their content harder across every channel they own.

Users who complete an experience can post their own content and earn the same way creators do. Satisfied customers become a distribution layer. They invite friends, share content, and grow the platform's supply of both content and travelers at the same time.

Each participant has a financial reason to grow the platform. The consumer UX is the interface through which they do it.

MVP Pillars

With the marketplace mechanics defined, we built the foundations of the consumer app around three functional pillars:

1. Videos & Profiles: gave us the social foundation, TikTok-style feed, profile pages, engagement.
2. Search & Navigation: built discovery flows, filters, and search.
3. Booking & Post-Tours: tied it together with trust: reviews, cancellation flows, refund clarity.

Solution Pillars

1. Trust & Transparency

The experience detail page is where discovery converts to commitment. It needed to establish host credibility, communicate what travelers are actually getting, and remove the ambiguity that kills bookings on generic OTAs.

Rich itineraries, host profiles, transparent pricing, and video reviews all live here as the foundation for that trust.

2. Authentic Content & Discovery

The video feed is the entry point for most users and had to feel native to how Gen Z already discovers travel. Short-form, scrollable, social. City, experience, and following filters let users narrow without friction. Deep links from each video connect directly to bookable experiences, closing the gap between inspiration and action.

3. Scalable IA & Booking

Global search with filters for city, date, price, and accessibility. An Experiences hub with tabs for all, booked, and saved. Booking confirmation with shareable links, guest management, and explicit refund policies.

The shareable booking link is a distribution mechanic. When a traveler shares their booking, they're doing acquisition for the platform.

4. Operational Foundations

We extended BOAT, an existing moderation tool, to cover TripShepherd content, users, and experiences. We also built an analytics framework across growth, sales, and engagement to give the team visibility into how the flywheel was actually performing.

Reviews as the Keystone

Reviews became the growth loop, more than just feedback:

• Guests gain trust from authentic video and text reviews.
• Hosts build reputation and credibility.
• Platform collects structured data for better recommendations.
• Content from reviews feeds back into discovery.

Low-Friction Onboarding

Social commerce is still novel in Western markets, which created an onboarding problem a standard tutorial couldn't solve. Front-loading information kills activation before users understand the value.

Instead we used behavior-specific onboarding. Guidance appeared when a user attempted a relevant action for the first time. Sign-up prompts only triggered when an action required an account, like saving an experience. By that point the user had already signaled intent, which meant activation happened before we asked them to commit.

Results

TripShepherd now hosts hundreds of experiences across Niagara Falls, Toronto, and New York City, with content posted from users across the Americas. Behavior-specific onboarding reversed early retention problems caused by platform novelty. The profit-sharing mechanic has since evolved into a Creator Marketplace, currently in closed beta, allowing content creators and experience hosts to partner directly.

A Note on our AI Approach

We used ChatGPT throughout the project as a structured thinking partner. The most valuable application was documentation: generating a canonical MVP Spine and Core PRD designed from the start to be ingested by AI coding tools like Cursor downstream. That created a single source of truth that kept design and engineering aligned across three continents without constant syncs.

We also used it to pressure-test edge cases before they hit design or engineering. Booking conflicts, host cancellations, missing location permissions, refund logic. Running those scenarios early in a draft, critique, refine loop surfaced failure modes that would have been expensive to find later. For a lean product team working with distributed squads, AI acted as the team members we didn't have.