Millions of publishers are competing for the favour of users – and for ad revenues. Thereby, visitors’ length of stay, page impressions and monetization depend largely on the quality of publisher’s content and ad recommendations. In order to prove the excellence of the plista Recommendation Technology, a top tier publisher started a long-term test to compare its performance with recommendations based on semantic targeting.
The ACM Recommender System has become the most important international conference for the presentation and discussion of recommender system research.
This year the conference was organized in Hong Kong. The conference is well balanced, both between science and economics and between mathematics and computer scientists. Thats why I like this conference and why plista was going there.
Many projects with scientific backgrounds were built around twitter data and of course around movie recommendation. Researchers argued a lot about how to actually compare different algorithms. The crazy mathematics were much into tweaks for Matrix Factorization algorithms. Moreover, there were many talks about the advertising industry held by speakers with economic backgrounds.
plista offered a workshop and a challenge about News Recommendations. We organized this workshop together with the Norwegian University of Science and Technology, the University of California at Santa Barbara and the local TU Berlin. The challenge was running for a couple of weeks and had a dozen of international contributions from Potsdam to Israel.
Mitte des Monats haben sich die Entwickler von plista zusammengefunden um das zweite Mal einen internen DevDay zu veranstalten. Gestartet wurde das Projekt bereits vor einigen Monaten mit dem ersten DevDay im Juli 2012.
Doch was ist eigentlich ein DevDay?
Ziel dieses Events ist es, dass sich die Entwickler abseits vom Tagesgeschäft und den sonstigen Baustellen zusammenfinden können, um sich gegenseitig über größeren Baustellen, Visionen und auch generelle Trends informieren zu können. Wichtig ist dabei, dass es eine gute Mischung aus internen und externen Themen gibt, sowie Themen für alle verschiedenen Gruppen von Entwicklern bei plista.
Wie bereits im Januar auf unserem Blog angekündigt, war plista Entwickler Torben Brodt beim diesjährigen Camp Digital (23. – 26. Januar 2013) in Hamburg als Speaker zum Thema “Algorithmus” zu Gast. Das viertägige Schulungserlebnis mit Wissen, Inspiration und Hands-on-Erfahrungen zu digitaler Kommunikation wurde von der Good School Hamburg veranstaltet. Diese vermittelt mit regelmäßigen Angeboten (u.a. Seminare, Innovations-Workshops, On-the-Job-Trainings), Unternehmens-, Agenturchefs, Marketingmanagern, Medienmachern oder Unternehmensberatern das nötige Know How rund um digitale Medien und den digitalen Wandel.
In last week’s blog, I outlined how plista’s RecommendationAds work- namely through combining interested-based filtering and various data vectors in order to come up with the personalized content recommendations the user sees on the page he or she is browsing. To put it more accurately, we use recommender algorithms for the interest-based filtering process. We at plista believe that we currently have one of the best live recommender algorithms out there, proven time and again by the success of our ad formats; average CTR (Click Through Rate) for the RecommendationAds alone is over 7%.
However, in order to continue to improve our services, we’d like to see how our recommender algorithms would compare to any contenders. We have therefore launched a contest to find the best live recommender algorithm to compete with our own. There is a weekly prize of 100 € for the best-performing team, and an exclusive grand prize (details of which are still to be announced) which will be awarded to the best-performing team overall over an extended period of time.
The aim of the contest is to improve the performance of the on-screen recommendation widgets that plista provides, which come in the format “you might also find this interesting”. Performance is measured by the CTR- the amount of clicks per number of displays for a given set of recommendations- as well as the response time for each recommendation request.
The contest consists of two rounds for each team: the preliminary round, where teams receive a low volume of messages with a chance of winning the weekly prize and qualifying themselves for the second stage, where the teams receive a much higher volume of messages and the chance of winning the elusive grand prize. Interested? Then read the introduction and take a look at the contest Wiki too for more information. Have you got what it takes?
Spread and share with your friends on Twitter and Facebook – maybe one of them will be the winner of the grand prize!
Find out more: http://contest.plista.com/