Ben Dias, easyJet’s Director of Data Science and Analytics, joined data science leaders from Janssen Pharmaceuticals and PointRight for a panel discussion on how data science, IT, and business leaders can work together to build a better enterprise. Their conversation was moderated by Domino’s Chief Data Scientist Josh Poduska and guest speaker Forrester Principal Analyst Dr. Kjell Carlsson. This post provides highlights from Ben’s comments along with a link to the full webinar.
Data Science at easyJet
easyJet is a low-cost European airline that offers flights between Europe’s leading airports. By focusing on routes that people want to travel and optimizing the frequency of flights for those routes, easyJet has made a name for itself in air travel. In fact, easyJet was awarded the best low-cost airline in Europe by Skytrax in 2019.
However, it’s no surprise that the airline industry has been among the hardest hit industries amid the COVID-19 pandemic as community lockdowns, travel restrictions, and consumer fears have kept travelers close to home. Airlines have had to become more nimble and during the panel Ben shared how the pandemic has transformed the way the company uses data, including increasing use of:
- Automation to support increased workloads, such as the need to update flight schedules on a monthly (and sometimes even weekly basis). Previously, the company typically published one winter and one summer schedule each year.
- External data that can signal market changes and evolving customer needs.
- Reinforcement learning to learn from the new data sets in real-time so the organization can respond to shifts in demand much faster.
The vast majority of organizations today recognize the importance of data science and its potential for making a real difference in how every industry operates and how every company performs. However, succeeding with data science requires strong partnerships. The relationship between IT and data science teams, as well as with business teams, often determines the overall value and extent to which data science can permeate an organization. For example, data scientists often face headwinds in getting access to the infrastructure, tooling, and data they need, which can stifle creativity and slow innovation.
During the panel discussion, Ben shared how easyJet is tackling this challenge.
He set the stage first by highlighting the differences in how data scientists and data engineers approach things—which can create friction from the outset. Ben’s background is in data science, but he ran both data science and data engineering teams together in a previous role. It was a tremendous learning opportunity, he said, that showed him just how differently the two groups work and think as they stood up a new Hadoop cluster. He describes it as akin to having teams speaking different languages.
If you understand why they [data engineering] need that control, then you can explain that makes sense for this part of the process and not the other part.
What three practices does he recommend to close the communication gap?
- Explain the “why,” for instance, why IT needs specific controls and governance in place. By doing so, they can agree on which parts of the process this makes sense for and which parts it doesn’t.
- Adopt a Lean Startup approach that progressively increases controls the closer a project is to production. For example, with this approach, easyJet gives data scientists the freedom to explore different ideas and tooling in a somewhat controlled environment using the same templates, languages, and tooling wherever possible. Still, they have flexibility to go outside this during the early stages of development, and are provided a clear understanding of the process and gates required as model development progresses. This is critical in reducing the uncertainty, so data scientists know from the outset what has to be done and when to harden the solution for running successfully in a production environment.
- Reuse existing processes when possible, such as current IT and operational guidelines for service design, service transition, and governance, when bringing new models into production rather than creating an entirely new process. Certainly, he said there would be exceptions, but in general, those should come in the early project days, as previously mentioned, when data scientists are in the exploratory phase. And there will be areas that current software deployments typically don’t cover, model monitoring being the most significant, which teams will have to put in place (it can come from either side of the aisle) both in terms of processes and people.
You should be able to use the same service design, service transition, all the governance processes...Model monitoring is the key different thing that you have to put in place.
About Ben Dias
Ben is currently leading the delivery of easyJet’s data strategy aimed at realizing the company’s ambition of becoming the world’s leading data-driven airline. Previously having worked at Royal Mail, Tesco and Unilever, he has over 15 years industry experience in solving real-world problems in an industry setting. More recently, Ben has focussed on building and leading Data Science teams and applying the Lean Start-Up approach to Data Science within large organizations. He is experienced in setting up and managing industrial research collaborations with academic and business partners. He is also actively engaged with the UK Mathematics community and very enthusiastic about inspiring the next generation of Mathematicians and Data Scientists. Ben holds a PhD in Computer Vision and an MSci in Mathematics and Astronomy, both from University College London.
To learn more
Watch the Webinar “Reaching across the aisle” to learn best practices to unlock innovation of data science teams while enabling IT to scale.