News media is in a digital transformation, disrupting their existing business models. Many news media houses are looking into recommender systems as a part of their digital strategies. However, the social role of journalism, existing publishing platforms and news as a continuous data stream infer particular challenges for applying standard recommender technologies. This paper explores how news recommendation can go beyond popularity and recency and take advantage of content quality metrics and interaction patterns. This knowledge is derived through adapting process mining for usage with web logs. The proposal is evaluated on real event logs from a German news publisher, revealing encouraging results.