All businesses are not equal, and approaches will vary. Skyscanner explains why it took the middle road
Since the ‘big data is like teenage sex…’ quote from Duke University Professor of Psychology and Behavioural Economics Dan Ariely went viral in 2013, the industry has come a long way. What they call it might vary, but any company likely to survive in the world of travel tech, is doing data science in some form or other. It’s also widely agreed that the term ‘big data’ doesn’t really cut it.
Mark Shilton Skyscanner’s principal data scientist recently spoke at EyeforTravel’s Smart Data Summit in Amsterdam where he shared several insights about the company’s move to put AI-driven data science at the heart of the business.
One of the things he referred to was the Booz Allen Field Guide to Data Science, which outlines three possible routes to take.
1. Centralised. In this scenario business units bring their problems to a centralised data science team.
2. Deployed. Here small data science teams are forward deployed to new business units.
3. Diffused. Data scientists are fully embedded in the business units.
Skyscanner experimented but quickly found that the centralised approach didn’t work. “We became the bottleneck and teams were still none the wiser as to how data science could help,” Shilton explained.
The company that last year was acquired by Ctrip, China’s biggest travel company, also tried the 'diffused' approach where data scientists were assigned to specific business units. In Skyscanner-speak, these units are known as tribes [Skyscanner has adopted a Spotify-type model]. With this approach there was no central data science function, but again this wasn’t the right fit. “For us this didn’t work as we felt it could easily create silos and make individual data scientists feel isolated,” Shilton said.
Instead, the company took the second road - the ‘deployed’ route, where data scientists sit in teams with a specific data science need but with a central management function that coordinates it. Although there were ongoing challenges to ensure that people still talked, communicated and worked on the right projects, this seemed the right model for Skyscanner.
For Skyscanner what was important was to set common standards and principles which all data scientists could work with, and most importantly could be checked and measured. This helped individuals to prioritise and decide what problems to work on and how to approach them.
But as Andrei Grintchenko, head of business intelligence at IATA, put it: “Everybody has to assess their own place in the journey.”