To respect client confidentiality in a competitive setting, the company name, branding, and data presented have been changed.
Wanderloom | Weaving Connections Through Travel
AI / ML Product Strategy for User Engagement & Growth
Role: Product & Strategy Consultant (Pro Bono)
Duration: Ongoing 2024
Contributor: Tyler Gustafson (Gustani)
Overview
This consulting engagement explored how data-driven insights and AI integration transformed a social travel platform from MVP to a scalable product focused on meaningful connections.
Key Highlights
-
Developed comprehensive product roadmap that prioritizes data-driven features
-
Led development of AI/ML integration strategy leveraging Vertex AI & Gemini
-
Derived actionable insights from user behavior, trip activity, and app usage to guide high-impact platform enhancements
The Challenge
How to harness data-driven insights to transform the platform, shape a product roadmap, and build proof-of-concept features that would fuel Wanderloom’s next phase of growth.
The Approach
The scope of this work focused on two key objectives. First, we examined what the data was telling us—looking at user behavior, trip activity, and app usage—to diagnose the problem. Then, once we understood the core issue, we developed more targeted solutions to address it.
What were seeing
in the data...
(Users, Trips, App Usuage)
What we can do
about it...
(Features, UX, Generative AI/ ML)
&
The Data...
We examined a variety of factors when analyzing user behavior, but for confidentiality reasons we’ll keep our focus narrow. One particular analysis I want to highlight is our work in Neo4j (graph neural networks), where we explored user network behavior. Below is an initial overview of the Wanderloom user network and their connections:
Wanderloom
User Connection Network
(Subset)
The beauty of this approach is that by closely analyzing user behavior patterns and network formation—using centrality and other advanced algorithms—we can identify key patterns and groups. These insights can then inform targeted improvements. For example, we can see distinct networks or friend groups to build focused recommendation systems around. We also notice isolated users we can engage further to prevent churn, as well as ghost users who appear disconnected and may not be active. Each of these findings represents an opportunity to enhance the platform through strategic AI integration and UX improvements, keeping every interaction effortless and engaging. Below is what some of these groups look like:
Lets Run
Graph Clustering Alogrithims!
As I mentioned, this was just the tip of the iceberg in our analysis, and there were several areas we won’t cover in this write-up. Below are a few more categories we explored to inform our strategic approach:
1
User Behavior
Examined user growth by region and location to understand adoption patterns and identify new opportunities for engagement. Additionally, we analyzed how network effects—such as friend connections and user clusters—impact overall engagement and retention.
2
Trip Activity
Analyzed factors like trip size, duration, and scheduling patterns to uncover potential ad opportunities and better understand user travel habits. We also tracked how many trips each user had taken—especially those with fewer than one trip—to identify where we could encourage more participation.
3
App Usuage
Explored how much time users spent on various features, along with their interaction patterns, to gauge which parts of the app were most engaging. Assessing search function performance also helped us pinpoint areas for optimization, ensuring a smoother user journey.
The additional analysis ultimately identified three key issues, highlighted below.
Recap of Data Driven Problems (TLDR)
What we can do about it...
Having identified our key challenges, we’re now focused on making it effortless for users to expand their networks within the app. The main goal is to find effective ways to deepen their connections with others.
Due to client confidentiality, we’ll keep things fairly high-level and only showcase two examples and avoid getting into some of the reasons why we focused on this items. Below are some of the areas we covered.
How do we get users more connected?
-
Make Connections Exciting: Transform the profile page with engaging visuals, interactive friend counts, and features that inspire users to grow their network.
-
Make the “Share” Button Accessible: A primary goal right now is to grow and to do that we need to ensure it’s effortless for users to share with friends.
-
Add a Post-Trip Feature: After a trip, encourage users to add missed connections and provide opportunities to plan their next adventure.
-
Simplify Adding Connections in Trips: Clearly differentiate between connections and non-connections within the interface.
Feature Roadmap Example 1: Trip Recap
To prevent ‘one-and-done’ sojourns, we want to provide an engaging post-trip nudge that keeps users coming back. It’s also an opportunity for them to quickly expand their networks and give us valuable feedback, strengthening our database for future recommendation systems. This is a key strategic move to capture data, and improve future user experiences, while also addressing some of the opportunity areas that we identified.
After Trip is Complete, Prompt User...
Feature Roadmap Example 2: AI Itinerary Planner
We then introduced an AI trip planning feature to make the process more effortless and truly ‘wow’ our users, letting them focus on building connections rather than getting bogged down in logistics. This also creates an additional revenue opportunity through integrated ad placements, making the platform more valuable for both users and partners. Below is an overview of the application feature.
Overview
The trip planning feature dynamically recommends itinerary items based on the current plan, group size, and location. These recommendations adapt as activities are added, ensuring they match the vibe of the existing itinerary. With a click, users can add activities to the plan or create a poll - reducing the stress of trip planning and creating exciting plans.
How it works:
-
Dynamic Recommendations: Tailored to the group size, location, and existing itinerary and update in real-time as plans evolve.
-
Quick Add or Poll Options: Users can instantly add suggestions to the itinerary or generate a poll for group decisions.
-
Intuitive Updates: Recommendations adapt to changes, ensuring seamless alignment with the trip’s tone and preferences.
-
Optional Guidance: Users can input preferences or specific guidance (e.g., “Looking for relaxed activities”) to refine recommendations further.
Future Enhancements:
-
Additional Features: A text box can be added to the trip planner to capture user preferences and make suggestions even more personalized or pull in from user profile preferences.
-
Trip Generator: Can be used to suggest entire trip from scratch based off group attributes and past trips.
We have a couple Deployment Options
Homegrown Model
"Why not build our own model?" Building and maintaining a custom LLM requires significant resources and expertise. Without a dedicated Data Science / ML team this would be a significant investment.
Hugging Face
We can use and train different open-source models and train them. There is a benefit of not being locked into GCP, but this will also take significant effort for ML Ops and Deployment for scaling.
Google Cloud (Commerical Use)
Of these three options Google Clouds Vertex AI is ideal for a team with limited technical AI / ML capabilities and offers adaptability, security and efficiencies for development and deployment.
Vertex AI's low-code/no-code solutions with Gemini make it easier to deploy a user-friendly trip planner feature.
Vertex AI's low-code/no-code solutions with Gemini make it easier to deploy a user-friendly trip planner feature.
Summary
This was a brief overview of my work at Wanderloom (with certain details changed for confidentiality). By leveraging data-driven insights rather than relying on gut feelings, we tackled major app challenges and built a product roadmap rooted in real user needs. From identifying key user behaviors to implementing AI-driven solutions, every step was guided by a strategic focus on delivering effortless, engaging experiences.
The methods and mindset behind this project—grounded in analytics, user empathy, and practical execution—show how a thoughtful approach to problem-solving can lead to impactful product outcomes. By iterating on user feedback and applying scalable technology, we were able to craft solutions that resonate deeply with users.
