November s Travel Meets Big Data event at Trinity College, Dublin was a mix of a data analytics contest and an educational event for people interested in travel technology.
NB: This is a viewpoint from Mark Lenahan1, independent advisor to airports, airlines, retailers, and loyalty programs.
Organised by non-profit Entrepreneurs Anonymous2, the event was designed to encourage students, coders, data scientists and entrepreneurs to find new and innovative ways to tackle some of the big data challenges facing the travel industry. Although data science is not a new idea to business, doing it collaboratively and inviting an outside point of view is.
Two significant data analytics competitions were sponsored by DAA3 and Hostelworld, both of whom also very large data sets for participants to experiment with. Additional software and hardware support were provided by Dataiku5, a provider of analytics software, Bigstep6, (a bare-metal cloud computing service for big data applications, and Fluid UI7, a UX design and prototyping tool for web and mobile.4
Dublin Airport Security Queue Prediction Challenge
The Dublin Airport challenge consisted of predicting queue times at security. Based on historical data and flight information entrants were asked to predict not only passenger numbers but the presentation profile when they would arrive in the security queue. The sample data provided included historical queue times and passenger numbers as well as full airline schedules.
Not being a data scientist / analytics person myself I found it interesting that different analysis methods came up with different answers on the significance of certain data elements.
For example, a number of teams mentioned the destination airport as being an important predictor of passenger behaviour to my mind that makes intuitive sense people travelling to LHR will behave differently than people flying to North America.
However, one team determined that it could be left out of their model altogether.
Time-of-day and date-of-year were generally considered very important predictors, along with day-of-week, but there was a broad consensus unsurprisingly that aircraft capacity was a key indicator of security queue size.
This leads me to wonder what predictive analytics could do if they also had airline data, such as revenue management predictions on flight loadings?
First place for the most accurate predictions, and the grand prize of five nights in New York, went to husband and wife team Caroline Kirrane and Barry Scott.
They ran multiple models, ultimately settling on an algorithm that relied on hour-of-day, the customer prior behaviour (passenger presentation profile) and the aircraft capacity as being the most significant indicators of future behaviour. Month-of-year and day-of-week also playing an important role in the winning entry.
Barry and I were drawn to this event for two reasons, Caroline said, firstly because there was data involved and we enjoy hacking around with data. And secondly because there were two real-world problems to work on. We feel that great companies get built and change happens when people are working on actual problems rather than brainstorming about potential products that might be difficult to sell.
Hostelworld Recommender System Challenge
The Hostelworld challenge, somewhat inspired by the famous Netflix algorithm prize, was to find a way to make better recommendations.
A sample dataset of 10,000 hostels and 1 million reviews including both review scores and text, and all of the related (anonymised) customer data was shared with participants.
A portion of data was set aside for testing their predictions. The challenge was to use the data predict which hostel a given customer would be more likely to book and/or rate positively.
The data scientists from Hostelworld interacted with the teams for the entire event but unfortunately only one team completed a solution.
Silviu Preoteasa Head of Marketing Technology in Hostelworld admitted that the challenge may simply have been too much for a weekend event and decided to extend the submissions deadline until November 28th.
It was still very useful for us. We have a better understanding of the scale and complexity, Silviu added, often in explaining a problem you come to understand it a lot better yourself .
The Conference
In parallel to the big data challenges, a number of sponsors and independent travel industry commentators made presentations over the course of the weekend.
Darren Cantwell of Dublin NDRC based startup specialising in travel agent technology Dynamic Res spoke about his experiences of starting a business in travel and serving the often challenging market of small to medium travel agencies with their solution for managing quotes and making bookings.8
Toby Houchens of Travel Recon explained the importance of travel risk management and duty of care and their unconventional approach to tracking, analysing and alerting travellers and travel managers to incidents and risks anywhere in the world.9
Breffni Horgan spoke on the evolution of mobile at Hostelworld, where mobile bookings will overtake all other channels (growing to over 50% of sales) by the end of this year.10
Silviu Preoteasa explained some of the challenges of managing marketing technology for a website with more than 6 million user per month and 1,000 bookings per hour.
Quoting SimilarWeb, Silviu pointed out that more than 450,000 websites sell accommodation, accounting for 2 billion visits per month and an average of 25 web searches performed prior to every booking.
Commenting about their involvement in the event, Silviu said, Hostelworld s tag line is Meet the world and within the company we all embrace that this event was a natural opportunity for us to meet people interested in data science and share our experience with them.
Dublin Airport provided two speakers: John Brennan Head of Commercial, and Katie O Leary Head of Innovation. John focussed on Dublin Airports excellent results for 2016, with more than 27 million passenger movements project by end of year, making DUB Europe s fastest growing airport with 14% year on year growth in passenger numbers.
John also made the point that the relationship between the airport and airlines was becoming more grown up both parties having realised that cooperation is vital to improving the overall passenger experience.
Head of Innovation, Katie O Leary, presented three examples of innovation in Dublin airport:
- How DAA overhauled their car parking automation and introduced online booking on dublinairport.com as well as Aer Lingus and Ryanair.
- The introduction of automated biometric kiosks in the United States pre-clearance, handling up to 4,600 passengers per day.
- The overhaul of the retail experience including Candy Cloud the sweet shop designed by actual school children.
At the start of the event I presented, for a non-travel audience, some of the customer challenges facing the travel industry:
- Customer Identity the fact that many of the people responsible for customer experience are sorely lacking in contextual and timely customer data.
- Disruption how it impacts customer experience and loyalty for all brands involved in the journey with reference to the recent Amadeus / T2RL white paper by Ira Gershkoff (previously reported in Tnooz).11
- Travel Loyalty some commercial outcomes possible in the travel industry from trying to maximise customer loyalty and lifetime value.
NO REPRO FEE 20/11/2016 Pictured are David Belton (left) with Dublin Aiport and Harish Kumar at Travel Meets Big Data , Ireland s first travel industry focused hackathon and conference that took place in Dublin this weekend. The conference brought together 200 software developers, data scientists, business developers and marketers to analyse large sets of data or Big Data to reveal trends and patterns relating to human behaviour and interactions to build commercially viable products and services. Conference sponsors, Dublin Airport and Hostelworld challenged participants to build an efficient security screening system for Dublin Airport and a recommendation engine for users for Hostelworld. PHOTO: Mark Stedman
Conclusion
As a conference, the format was a little problematic at times. Data scientists were hard at work on solving the competitions with little time left for watching presentations.
It is also difficult to get either vendors or buyers to commit to a weekend event as a result this wasn t a typical trade show.
Given the audience in the room students and graduates, scientists from all branches (I met quite a few physicists), software developers and data analysts it was a shame to not see HR out in force from the big travel tech employers in Dublin.
In the end the event wasn t quite a hackathon (more a data analytics competition) and not quite a conference (but could have been a recruitment fair), but I learned a lot.
It served to provide students, data scientists and entrepreneurs in Dublin a crash course in the travel industry and all its glorious complexity.
NB: This is a viewpoint from Mark Lenahan12, independent advisor to airports, airlines, retailers, and loyalty programs.
Images by Mark Stedman and courtesy Entrepreneurs Anonymous & B.U.I.L.D.
References
- ^ Mark Lenahan (marklenahan.com)
- ^ Entrepreneurs Anonymous (entanon.com)
- ^ DAA (www.daa.ie)
- ^ Hostelworld (www.hostelworld.com)
- ^ Dataiku (www.dataiku.com)
- ^ Bigstep (bigstep.com)
- ^ Fluid UI (www.fluidui.com)
- ^ Dynamic Res (www.dynamicres.com)
- ^ Travel Recon (travelrecon.com)
- ^ Hostelworld (www.hostelworld.com)
- ^ previously reported in Tnooz (www.tnooz.com)
- ^ Mark Lenahan (marklenahan.com)