Welcome to the first Helsinki Data Science Meetup in 2018!

The first meetup of the year was hosted by Kesko, a Finnish trading sector pioneer operating in the grocery retail space, the building and technical trade and the automobile market.

DAIN Studios opened the event by some news regarding activating the Helsinki Data Science community. A survey was conducted in January among the LinkedIn group members with the purpose of knowing how the community and the meetups should be developed further. The group clearly enjoys the hands-on examples and specific use cases presented by fellow data scientists!

One of the findings regarding the meetups was that the members would enjoy even more discussions and facilitated networking during the meetups, and that is something that was piloted already at this Kesko event.

To increase activity also in between meetups, a Slack channel was launched, you can join via this link

Then it was time for our hosts at K Digital to take stand.

As trading sector operations are affected by several global megatrends ranging from the digital revolution to climate change, Kesko has responded by creating new value propositions around digital stores and services. Kesko wants to offer an unparalleled customer experience by making use of mobile services, online services and digital marketing, and understand individual customer behaviour by making extensive use of customer data.

Kesko has been nominated as the world’s most responsible trading company, and for K Digital this means they are handle customer data with great respect.

Sanni Holm from Onninen, the building and technical trade arm of Kesko, presented the pricing engine they have been developing to align customer prices with customer expectations. The challenge is in volumes: 100k products, thousands of customers in several countries. Sofar the pilots have been very successful and using the engine, Onninen has been able to increase margins without losing business.

Niilo Latva-Pukkila from Kesko AI unit then explained how they have been predicting online behaviour with offline data, when creating recommendations on repices based on what the consumer has purchased. Some of the cool things Niilo explained was non-negative matrix factorization: expressing one large matrix by a product of two matrices. Doing this they have been able to reduce the 100k items in the receipts into 200 features. This is a method that also works for sparse matrices. In test campaigns they have been able to double the click rates for recipes displayed based on purchasing history vs most popular recipes.

Caroline Aderinwale then talked about programmatic marketing, and first presented the concept and the different players involved in the most clear and understandable way that it was really enjoyable. There is the common mantra of right context, right time, right content – but execution is not that simple. You need to have a lot of enablers in place. . (see photo)

Programmatic marketing allows Kesko to use the huge amount of collected data for delivering the right message, and moving towards the segment of one.

Once inspired by the presentation, the audience split into teams and came up with more or less futuristic concepts of grocery shopping in 20 years. The ideas ranged from logistics (eg drone delivery to fridge/home/nearby), to recipe recommendations (eg manage my health via good diet), to shopping assistance while in store (eg automatic check-out, how to find produts in store) .

Perhaps something we will see in the future when visiting our local K Market?