Problem Statement

Campaign can promote sales, but:

  1. How to judge the campaigns that are necessary for your case?
    • black friday? singles day? just because others do?
    • Purely based on your "own" experiences?
  2. Who to promote the campaigns to without spamming your customers too much?

We are here to help Finnair decision makers to create "smart campaigns" and do "precision marketing" for those campaigns.

  1. dynamically reflects upcoming events on global scale
  2. leverage the information that Finnair customers' really care about
  3. recommend events information and flight tickets together to those highly potential customers

Our Approach

  1. model customer preference based on their social media profile (Facebook open data, likes, posts etc)
  2. extract events/artists information from service providers (ticketmaster, spotify)
  3. match the events smartly with customers based on their preference (data mining)
  4. recommend events at the same time offer the tickets (finnair open apis) to the customers

Customer preference modeling

We use machine learning techniques (music artist, sports team, etc) to detect customers' real interest. In this demo, we use music as an example.

From a subset of One Million Song dataset, we obtained artists and queried Spotify APIs for each artist in the seed graph for similar artists.

Then, we filter the graph based on user's music preference and obtain artists that the user can be potentially interested in.

Event extraction

We extract event information global-wise and in real-time from ticketmaster

We got all Finnair destinations and filtered based on customer preference.

The bubble size of each city is determined by how many Finnair customers are interested in the events of the city.

Event re-ranking based on customers' preference

Events are re-ranked based on aggregated Finnair customers' preference

Event-customer matching

We match the customers with the events based on their preference

Ticket recommendation

Finally, we recommend the events together with tickets to the customers based on the above matching.

Built With

Share this project:
×

Updates