Opportunity:
Travel and tourism is a huge sector, with direct contributions of $2.57 trillion to the global economy in 2017 and total contributions of $8.27 trillion [1]. Travel agencies (including companies such as Expedia and Priceline) generated $40.5 billion dollars in 2017, with annual growth of 5.3% from 2012 to 2017 [2]. Given the size of this industry, even small improvements in helping consumers decide on and book travel arrangements would have a huge impact. Many consumers are looking for more tailored and personalized recommendations. Yet as the amount of options and information on travel increases, it becomes ever more time consuming to sift through all of the possibilities to find the unique and perfect fit for any given individual. This creates a large opportunity to go beyond the traditional and old-fashioned ranking systems.
Solution:
Duffle is a response to the increasing demand for unique experiences. It facilitates travel discovery and booking by analyzing personal preferences (stated and based on past reviews), airline, restaurant, car rental, tourist activities, weather, and social media data. AI-powered chatbots consider budget, frequent flyer and other reward membership information, location, and availability. Users are able to swipe left or right trip and activities recommendations. The algorithm learns over time from the user response to recommendations and can improve them when users input personal information such as pet or kids joining for the trip. Fellow travelers can also be added to a trip so Duffle and not the main user is the Q&A point for all people on the trip. For solo travelers, Duffle can connect users to other travelers considering the same destinations and dates.
Once a trip is booked, Duffle compiles a list of recommended attractions or activities and historical information about sites the user can visit. The service includes automatic check-in, 24/7 travel assistance app, and on the ground guide.
Duffle assists users past the initial booking by providing packing list recommendations, local sim card options, or advice on recommended vaccinations. It can also alert to security concerns or unfavorable weather changes and offer quick assistance with rebooking. The platform benefits from network effects on the client side since the algorithm learns from the likes and dislikes of the specific user and those of other similar users and can connect travelers with similar interests, budget, and desired travel dates.
Pilot:
Our pilot would require the initial aggregation of source data needed for the curation algorithm. Thus, we would begin to create the network effects within the university format at the University of Chicago or University of Illinois at Chicago (or any comparable university) that have a population of active social media participants that engage with rich multimedia content which can provide travel preference data. We would initially pool this data to create curated content which our data scientists and travel agent partners would initially evaluate for relevance and feasibility. After 2-3 months of the aggregation process, we would look to merge the data and algorithmic learnings to create a machine learning software that can input new data sets, curate existing ones, and accept continuous inputs that will allow it to learn continuously.
Post this initial data collection and curation period, we will engage our university volunteers in person and present them our recommendations. We will conduct online surveys and in-person interviews to gauge their WTP for the trips recommended to them and their relative interest levels. Additionally, we will compare our recommendations vs those of our competitor set (see below), using the data we collect on the student’s interests, geography, and finances. We would also have them rate their interest levels relative to our recommendations and those of competitors on a blind basis.
While the student population is ideal to begin with, they have limited income levels and travel abilities. We would lastly roll out this final pilot stage to all members of the university (faculty, staff, etc.) in order to maintain a comprehensive set of age and income ranges for our algorithm that encompass all market segments. We will look for our pilot to confirm/negate our hypothesis that our recommendation platform with be 1) more robust and comprehensive than competitors, 2) more thorough in its presentation, 3) provide more value for each budget, 4) have a higher probability of booking than competing platforms.
Competitors/Risks/Feasibility:
Currently, there are various companies offering services that Duffle will offer, but none that have successfully combined the data that we plan to combine, and utilized artificial intelligence to optimize the service. Hopper, a discount flight platform, has launched an “AI driven travel agent” that allows users to input preferences about their ideal trip (i.e. duration, time of year) and get recommended flight deals. The platform learns from which deals the user does or does not accept. We plan to utilize AI in a similar way, but through connecting a user’s social media and allowing for more preference information, can offer a more customized result, beyond just flight deals. Mezi, a corporate travel website acquired by American Express, uses AI to learn about a traveller with every additional trip – when they book, what type of ticket they prefer, etc, and provides a platform for booking and managing work travel. Our goal is to provide a similar platform for booking and managing consumer travel and combining it with recommendations for activities, hotels, etc while on the vacation.
While feasibility of the product is reasonable given we have seen companies utilize different aspects of the technology, and are just looking to aggregate these features, there are risks associated with Duffle. Travel is a popular industry with many companies looking to disrupt or own the booking and planning process. Given that, we will need to rely on network effects and push marketing while continuing to innovate and improve the product.
Sources:
[1] World Travel and Tourism Council. Total Contribution of Travel and Tourism to the Global Economy from 2006 to 2017. Reported through Statista.
[2] IBIS World Industry Report. Travel Agencies in the US. September 2017.
http://www.hopper.com/corp/announcements/hopper-can-now-predict-where-youll-want-to-go-on-vacation
Team Members:
Sam Steiny
Rosie Newman
Gergana Kostadinova
Javier Rodriguez