The Context
4–14 Day Manual Loop
The Legacy Bottleneck
The manual backend process blocked immediate user activation, turning a quick action into a multi-week delay.
The Product Fix
Structured Parsing
Automated Generation
Replacing the manual queue with an automated parsing interface instantly turned raw text into structured, bookable digital itineraries.
The Business Outcome
Decoupled Headcount
Operational Scale
Automating ingestion broke the link between volume and operational overhead, unlocking seamless backend scalability.
Plan a trip. Pick a country first. Fill out a static form. Hit a dead end.
What looked like a product was actually a facade for an offline operational nightmare.
Behind the scenes, every request dropped into a black hole of manual, offline operations, forcing high-intent trip planners to wait 4–14 days for a single draft itinerary.
The Architectural Pivot
-
The Architectural Pivot -
Static form ➔ Processing scattered trip ideas ➔ Structured backend parameters
How We Got Here
Our first visual attempt fell flat, forcing us to rethink the user journey.
🔴 V1: Jigsaw Map (Retired)
The Concept: A gamified flow where travel parameters (purpose, style/vibes, who, etc.) acted as interlocking puzzle pieces.
The Flaw: Users ignored it. Testing proved the map added no functional value; it merely took up screen space and distracted from the task.
The Lesson: Form was overshadowing function. I scrapped the map to reclaim screen real estate and prioritize user funnel momentum.
🟢 Latest iteration: Guided AI
The Pivot: A streamlined chat interface that translates complex travel parameters into a familiar, step-by-step dialogue.
The Execution: Hybrid inputs. Blending guided selections with open text gave flexibility while feeding our algorithm exact structured data.
The Outcome: Accelerated completion. Stripping away visual friction kept users moving rapidly, capturing exact data density for the backend.
Designing for Faster Conversion
Rigid, interdependent constraints create upfront friction when users lack the full context to make final decisions.
Restructuring the sequence to lead with high-certainty preferences like “trip vibes” builds immediate investment, capturing intent early to minimize drop-offs and accelerate the path to an instant draft itinerary.
09. Trip Profile
Before revealing the final itinerary, the system reflects the user's travel identity back to them as a crucial psychological finish line that validates their effort and builds anticipation for their match.
Deep Dive
06. Country Logic
Country is just one window into our backend architecture. Example map below to see how a single interaction drives the entire logic tree.
Path 1
User already mentioned a country early on ➔ Might input more countries ➔ System checks support ➔ Shows multimodal option cards comparing mentioned countries ➔ User selects 1 ➔ Moves on to Season & Timing question
Path 2
User did not mention any countries prior ➔ Inputs multiple countries at the prompt ➔ Shows option cards comparing mentioned countries ➔ User selects 1 ➔ Moves on to Season & Timing question
Path 3
User did not mention any countries prior ➔ Inputs an unsupported country ➔ System checks support and offers closest matches ➔ User avoids answering/insists on selection ➔ System captures the intent and moves on to Season & Timing question
Trip Profile (Desktop View)
The End of the Manual Loop
Data & Adoption Outcomes
Automated intake parsing successfully replaced a high-friction manual operational loop with real-time platform scale.
3.5k+
Queue Deflection
The platform successfully absorbed 3,509 unique trip requests, instantly triggering 938 automated itinerary matches to completely eliminate the legacy 4–14 day manual bottleneck.
85%
Onboarding Completion
Out of the specific test cohort users who actively initiated a new conversational trip request, 85.2% successfully completed the full onboarding journey to reach the matching draft itinerary.
95%
Recommendation Completion
Contextual country recommendations were served to 66% of users initiating a trip request, and 95% of that group successfully completed the onboarding flow using a suggested country.