Founded in 2013, Lightricks is the creator of popular, award-winning image and video editing applications. Since we hit the ground running with Facetune, our flagship product, we've built an arsenal of powerful creativity apps used by millions all over the world. Based in Israel with satellite offices in New York, Lightricks has grown to a team of over 140 employees working from beautiful, green campus-based offices in Givat Ram, Jerusalem. We're currently in a period of rapid growth: working on several breakthrough new products and opening a third office in London. Lightricks is profitable, with revenues of over $40 million in 2018 and is expected to continue growing aggressively in the coming years.
Facetune & Facetune2 are fun and powerful portrait editors that brings state-of-the-art tech to the masses. 50 million copies of the apps have been downloaded worldwide. Facetune was the #1 paid app in the US in 2017, #1 in the App Store paid charts in 130+ countries. The word 'Facetune' is a household term in North America, synonymous with photo editing. With partnerships like New York Fashion Week and collaborations at drag queen & beauty conferences, the Facetune brand is taking Lightricks to places you wouldn't expect a Jerusalem-based hi-tech company to go.
Want to help us at Lightricks pilot its churn prediction task for our flagship app? Then we invite you on board at DataHack. Sift through our application usage data to solve a pressing business need for Facetune2 (the hit selfie editor downloaded by millions of users worldwide) - predicting which users are likely to churn (cancel their paid subscription).
With help from Lighricks mentors, search through the noisy data, consisting of tens of thousands of users’ usage-records, to discover patterns that can build a model for predicting user churn.
The prize… a real-world, complex mission you can get your teeth into (with some sequence / time-series modeling), and a lucrative cash of ILS 6000.
To submit your churn predictions on the test data for grading and listing in the leaderboard (automatically updated based on incoming submissions), please mail the output of your model (see notebook for example of how to create one) to [email protected]. Please put your team's name in the Subject line.
We will rank submissions qualitatively by the F-1 score, the harmonic mean of precision and recall, with respect to the positive class (the users who churned) only (in other words, precision and recall with respect to the non-churning users is not part of the score). The F-1 score, as well as precision and recall, will appear in the leaderboard.
In the case of a tie among the top entries in terms of F-1 score up to a small constant (left to the discretion of Lightricks), we will subjectively evaluate the team's works in terms of training methodology, feature engineering, model interpretability and overall impression. We reserve the right to split the total prize of ILS 6000 and award it to more than one team, based on our subjective evaluation.
See HERE