What does data science have in common with astrophysics?
Both require tuning out noise to focus on the prize
Richard Masters, VP of Data and AI at Virgin Atlantic explains how the company is using data and machine learning to close the loop and improve customer experiences.
Richard Masters is VP of Data & AI at Virgin Atlantic. He also has a PhD in astrophysics and says that data science and astrophysics share a key attribute.
“Data is mainly noise. In astrophysics, there’s loads of background radiation and you spend all your time reducing that noise down to get a good picture of a distant galaxy. So, it’s an analogy I use at Virgin to say look we've got all this data, what we care about is information and getting that insight and decision support.”
It’s a good analogy because as an airline Virgin Atlantic generates vast quantities of data, and the remit of the data team covers the whole enterprise.
“We're across operations, customer experience, finance and commercial so we oversee the whole operation for analytical data and we’re creating an enterprise data platform.”
Being able to join these different data sources has come from Virgin Atlantic’s partnership with Databricks, which began in 2018.
“Initially it was a data science platform for us,” says Masters. “We were trying to create an environment for data scientists to better pull some data in, run analyses, work in the notebook environment and start to build a bit more into that warehouse. We had a lot in various databases from various providers at that point.”
Initially the focus was on commercial data and insight but that evolved, as Masters explains.
“It evolved to let us get customer data in, anonymized, etc, within the right context, so we can then look at profiles and demographics to get insight into who was on certain flights for example. Then that kind of evolved as more capabilities came onto Databricks, we realized we could do more with our ops data and with our engineering data.
“Since I’ve been back (Masters spent some time, post pandemic working at EY developing AI models but returned to Virgin in late 2023) we've put in a concerted effort to align a lot more within Databricks including Virgin Holidays. The platform's there. It's now just about formalising and moving into the newer technology like Unity catalogue to unlock all the other capabilities.”
Data governance is a vital part of the process, and not just from the customer’s perspective.
“The customer perspective is super important of course but so is trust in our data in general. If you're saying this number of tickets were sold, or that this flight was delayed, or these number of flights were cancelled, etc, you want to be clear and consistent and Databricks helps us with that.”
Use cases
Virgin Atlantic is already realising some tangible benefits from investments in data.
“The first is the ability to know you as a customer,” Masters explains. “That feeds into our crew apps, for example. We can know your history more and get you the right tailored ready to fly emails as well. Now, it’s more descriptive like what cabin you're in. Did you book through a business channel or is it consumer? What can we tweak in that messaging? And we're working now on how to add more into that. Have you had a bad experience before? How do we close the loop there? That’s the identity piece.
“We've also just got a lot more of our core commercial airline data in there. So use cases around understanding our competitor environment and looking more broadly at that market intelligence piece. That requires data from lots of different sources and comparing it to our own data. Bringing all that together really helps our commercial teams get that insight for more immediate decisions, but also more strategic ones like where might we want to expand our network to next.
“We’ve also digitised in engineering and management. All our pilots are paperless now on the planes and all the reporting and all the data underneath that has now gone on to our Databricks platform for the core reporting side of it. All that insight from this big transformation piece, again is all there in Databricks and we can build new reports and new insight beyond core compliance and safety reporting.
“Then there’s the machine learning and data science use cases. We have sim ops tools, so we can simulate certain events, calculate the probability of a delay on the flight so we can better plan for certain scenarios. Behavioural segmentation is another use case, understanding segments and then how can interact with those from a CRM perspective, and from that going right down to an individual perspective, that hyper personalisation is what we're working towards as well. We really want to get to the point where we can know when you want that gin and tonic!”
Augmentation not replacement
Masters is careful to emphasise that data and AI is not intended to replace people or obstruct customer access to information.
“It all comes down to decision support. For me, great data science has been an ordered list. What's the thing I need to look at today and a bit of context around how I should use that. Even with generative AI, it's still that decision support, but you just give more flexible context to it. You can add to that top thing on your list additional information that helps you turn it into something actionable. It could be just a description of the customer, their likes and dislikes that then helps you the employee with your experience, turn that into something really personal.”