The Finnair challenge was to introduce new ways of collecting highly unstructured data from various external sources about disturbances (aka disturbance information), allowing Finnair operations to react on issues in timely manner to make sure the impact to the customer experience (i.e. flight schedules) in minimized. Our approach is to harness everyone - Finnair's employess, suppliers and customers - to contribute information and uncover issues (disturbance information) about scheduled flights as early as possible, so that the issues can be resolved to minimize the impact on the schedules and the customer exprience.
This team has created a platform, that turns the traditional value chain of Finnair: suppliers -> Finnair -> customers into an ecosystem, where customers, suppliers and Finnair employees can contribute to the travel experience that leads into better travel experience and service for all parties. It has a dramatic impact to Finnair's business, where you stop maintaining walled gardens between suppliers and customers, and turn them into an ecosystem aiming together at creating extraordinary travel experience for customers, and on the other hand more profitable business for Finnair and its' suppliers.
We want to turn frequent flyers into frequent contributors , where you are rewarded for making it a better flight for everyone, instead of miles you spend on sitting on your seat. The idea behind the platform is the same as with tech companies offering rewards for bugs & vulnerabilities: the system will reward customers, suppliers and its own employees through how many issues you have contributed (the bigger the issue - especially if not spotted by anyone else - the higher the reward). For customers, the platform offers an excellent channel for providing feedback if something is wrong with their flying experience, instead of publishing negative reviews or complaints in the social media. For suppliers and Finnair employees, the platform offers an immediate channel to report issues, that may prevent flight departure for example.
What it does
It's a product concept that has both web and mobile user interfaces. It allows passengers, Finnairs' suppliers and collaborators, Finnair employees and other 3rd parties to contribute live streams of information through mobile application. The information becomes immediately visible within the dashboard for the Finnair employees. The dashboard can be used to provide "points" ("miles") back to contributors (i.e. people who send valuable information related to flights) so that the get immediate feedback / gratification for sending information.
The information collected from different channels is used to calculation delay prediction.
How I built it
The system is based on distributed computing platform Spark, which is enables integration of different data feeds. The mobile channels have been integrated as an example, as well as a channel generating mock-up data (for the purposes of the demo).
Each contributor (passenger, employee, collaborator, supplier etc) has a trustlevel within the system, i.e. a score that indicates how good quality information they have provided and much their information should be trusted when calculating the predictions. The delay predictions are based on a model that takes into account the trustlevels of different sources, computing final prediction using the information supplier from various feeds.
As the backend is based on scalable distributed computing platform, it can be easily extended to cover massive amounts of data. In addition, integration of new data feeds is easy.
Challenges I ran into
It would have been nice to have information about actually delayed flights - this way also learning algorithms could be used a part of the prediction cycle: the trustlevel could be adjusted automatically by creating a model which correlates the information provided from each channel and how well they predict the actual delays.
Accomplishments that I'm proud of
I was a lot of fun doing it!
What I learned
Got to know new people during the weekend.
What's next for Finnair-FrequentContributor
Either extend the concept within this business domain, or apply the same model into different business domains.