Evaluating mobile web search performance by taking good abandonment into account
Olga Arkhipova, Lidia Grauer · 2014
Usage of mobile devices for Web search grows rapidly in recent years. The common tendency is that users want to receive information immediately results in incorporating rich snippets and vertical results into search engine result pages (SERPs) and in increasing of good abandonment. This article provides an offline metric for quality evaluation of mobile Web search, which takes good abandonment rate into consideration. The metric is the DBN click model that allows the probability to be satisfied directly on the SERP. The model parameters are estimated from the mobile search logs of a controlled experiment. The new metric outperforms traditional ERR metric in terms of the validation dataset built using a SERP degradation technique.