FABRICE COFFRINI/AFP/Getty Images
show image

Sooraj Shah

Contributing Editor

Sooraj Shah is Contributing Editor of New Statesman Tech with a focus on C-level IT leader interviews. He is also a freelance technology journalist.

Finnair’s data chief on ML, strategy and selecting suppliers

Finnair has been operating for close to a century, but it was only two years ago that the airline created a head of data role, reporting in to the chief digital officer. Minna Karha, who had worked at Sanoma, a media group based in Helsinki, took up the position, and it was up to her to define what it meant to be head of data, and what action she would take.

“I was speaking to different stakeholders across the organisation, finding out what their challenges were and what they were currently doing and how they were using data analytics,” she tells NS Tech.

With this information she worked out which roles were required within the business and in which areas strengthening was required before putting together a team to support the business. This included data engineers to make data available, and data asset product owners who are responsible for ensuring teams use the full potential of the data they have to create business value.

Karha and her team then put together a data strategy, considering how they wanted the company to look in 2025 when it came to data usage.

“We chose to start with data availability, so making the data assets visible and building an environment where it’s scalable and secure to do data analytics – where it’s easy to combine data from various sources,” she says.

“We started to build a data platform and we chose Snowflake to be the data warehouse – so a modelled data layer where the analysts can really query and analyse the data, and then the other important tool was the data catalogue to make our data assets visible and have a place to manage the full inventory of all the data we have within the company, making it clear what kind of data we share with external parties,” she adds.

This means if Karha needs any specific information – such as market research data – it is all in one place, and there is also information on how the data can be accessed.

Snowflake was selected at the beginning of the year without a tender process, primarily because Finnair wanted to have a scalable environment where it’s easy to combine data but also the scale to really query and probe a big amount of data – something that the previous products, including an Oracle solution, had issues with.

Learning from machines and industries

Machine learning is another area the data team has worked on.

“We have built propensity models to provide more personalised offerings for our customers. For example, we have a model which enables us to target the most interesting upsells for a person travelling to a certain destination. This means we can know if a person on a particular flight is more likely to buy extra luggage, so we offer it to them,” Karha says.

“We are also working on using machine learning in our operations, so it gives us efficiency on being able to predict what impact weather conditions have on flight delays. This means that even before delays happen, we can take action and mitigate the delay. There are huge opportunities here,” she adds.

Internet of things data – such as the sensor data from aircraft – combined with machine learning is another area the company is exploring.

Karha believes machine learning will help to bring more advanced insights for Finnair’s operations, control centre, customer care and marketing, to make better decisions faster. She says that the airline is always on the lookout for new ideas within the industry and outside of it.

“We are continuously looking at what other airlines and other companies in the travel industry are doing. We’ve also hired data scientists from outside the aviation industry so they bring with them expertise from other companies and we’re continuously scanning the area of what companies in other sectors are doing,” she says.

“This area is critical for us, so we want to build our own models, and it’s important to have ownership of these models to ensure it’s transparent, as well as being able to check that the efficiencies we are making through machine learning are ethical and sustainable,” she adds.